diff --git a/swh/storage/cassandra/cql.py b/swh/storage/cassandra/cql.py
index 2d235e23..7214dbdc 100644
--- a/swh/storage/cassandra/cql.py
+++ b/swh/storage/cassandra/cql.py
@@ -1,1323 +1,1337 @@
 # Copyright (C) 2019-2020  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 from collections import Counter
 import dataclasses
 import datetime
 import functools
 import logging
 import random
 from typing import (
     Any,
     Callable,
     Dict,
     Iterable,
     Iterator,
     List,
     Optional,
     Tuple,
     Type,
     TypeVar,
     Union,
 )
 
 from cassandra import ConsistencyLevel, CoordinationFailure
 from cassandra.cluster import EXEC_PROFILE_DEFAULT, Cluster, ExecutionProfile, ResultSet
 from cassandra.policies import DCAwareRoundRobinPolicy, TokenAwarePolicy
 from cassandra.query import BoundStatement, PreparedStatement, dict_factory
 from mypy_extensions import NamedArg
 from tenacity import (
     retry,
     retry_if_exception_type,
     stop_after_attempt,
     wait_random_exponential,
 )
 
 from swh.core.utils import grouper
 from swh.model.identifiers import CoreSWHID
 from swh.model.model import (
     Content,
     Person,
     Sha1Git,
     SkippedContent,
     Timestamp,
     TimestampWithTimezone,
 )
 from swh.storage.interface import ListOrder
 
 from ..utils import remove_keys
 from .common import TOKEN_BEGIN, TOKEN_END, hash_url
 from .model import (
     MAGIC_NULL_PK,
     BaseRow,
     ContentRow,
     DirectoryEntryRow,
     DirectoryRow,
     ExtIDByTargetRow,
     ExtIDRow,
     MetadataAuthorityRow,
     MetadataFetcherRow,
     ObjectCountRow,
     OriginRow,
     OriginVisitRow,
     OriginVisitStatusRow,
     RawExtrinsicMetadataByIdRow,
     RawExtrinsicMetadataRow,
     ReleaseRow,
     RevisionParentRow,
     RevisionRow,
     SkippedContentRow,
     SnapshotBranchRow,
     SnapshotRow,
     content_index_table_name,
 )
 from .schema import CREATE_TABLES_QUERIES, HASH_ALGORITHMS
 
 PARTITION_KEY_RESTRICTION_MAX_SIZE = 100
 """Maximum number of restrictions in a single query.
 Usually this is a very low number (eg. SELECT ... FROM ... WHERE x=?),
 but some queries can request arbitrarily many (eg. SELECT ... FROM ... WHERE x IN ?).
 
 This can cause performance issues, as the node getting the query need to
 coordinate with other nodes to get the complete results.
 See <https://github.com/scylladb/scylla/pull/4797> for details and rationale.
 """
 
 
 logger = logging.getLogger(__name__)
 
 
 def get_execution_profiles(
     consistency_level: str = "ONE",
 ) -> Dict[object, ExecutionProfile]:
     if consistency_level not in ConsistencyLevel.name_to_value:
         raise ValueError(
             f"Configuration error: Unknown consistency level '{consistency_level}'"
         )
 
     return {
         EXEC_PROFILE_DEFAULT: ExecutionProfile(
             load_balancing_policy=TokenAwarePolicy(DCAwareRoundRobinPolicy()),
             row_factory=dict_factory,
             consistency_level=ConsistencyLevel.name_to_value[consistency_level],
         )
     }
 
 
 # Configuration for cassandra-driver's access to servers:
 # * hit the right server directly when sending a query (TokenAwarePolicy),
 # * if there's more than one, then pick one at random that's in the same
 #   datacenter as the client (DCAwareRoundRobinPolicy)
 
 
 def create_keyspace(
     hosts: List[str], keyspace: str, port: int = 9042, *, durable_writes=True
 ):
     cluster = Cluster(hosts, port=port, execution_profiles=get_execution_profiles())
     session = cluster.connect()
     extra_params = ""
     if not durable_writes:
         extra_params = "AND durable_writes = false"
     session.execute(
         """CREATE KEYSPACE IF NOT EXISTS "%s"
                        WITH REPLICATION = {
                            'class' : 'SimpleStrategy',
                            'replication_factor' : 1
                        } %s;
                     """
         % (keyspace, extra_params)
     )
     session.execute('USE "%s"' % keyspace)
     for query in CREATE_TABLES_QUERIES:
         session.execute(query)
 
 
 TRet = TypeVar("TRet")
 
 
 def _prepared_statement(
     query: str,
 ) -> Callable[[Callable[..., TRet]], Callable[..., TRet]]:
     """Returns a decorator usable on methods of CqlRunner, to
     inject them with a 'statement' argument, that is a prepared
     statement corresponding to the query.
 
     This only works on methods of CqlRunner, as preparing a
     statement requires a connection to a Cassandra server."""
 
     def decorator(f):
         @functools.wraps(f)
         def newf(self, *args, **kwargs) -> TRet:
             if f.__name__ not in self._prepared_statements:
                 statement: PreparedStatement = self._session.prepare(query)
                 self._prepared_statements[f.__name__] = statement
             return f(
                 self, *args, **kwargs, statement=self._prepared_statements[f.__name__]
             )
 
         return newf
 
     return decorator
 
 
 TArg = TypeVar("TArg")
 TSelf = TypeVar("TSelf")
 
 
 def _prepared_insert_statement(
     row_class: Type[BaseRow],
 ) -> Callable[
     [Callable[[TSelf, TArg, NamedArg(Any, "statement")], TRet]],  # noqa
     Callable[[TSelf, TArg], TRet],
 ]:
     """Shorthand for using `_prepared_statement` for `INSERT INTO`
     statements."""
     columns = row_class.cols()
     return _prepared_statement(
         "INSERT INTO %s (%s) VALUES (%s)"
         % (row_class.TABLE, ", ".join(columns), ", ".join("?" for _ in columns),)
     )
 
 
 def _prepared_exists_statement(
     table_name: str,
 ) -> Callable[
     [Callable[[TSelf, TArg, NamedArg(Any, "statement")], TRet]],  # noqa
     Callable[[TSelf, TArg], TRet],
 ]:
     """Shorthand for using `_prepared_statement` for queries that only
     check which ids in a list exist in the table."""
     return _prepared_statement(f"SELECT id FROM {table_name} WHERE id IN ?")
 
 
 def _prepared_select_statement(
     row_class: Type[BaseRow], clauses: str = "", cols: Optional[List[str]] = None,
 ) -> Callable[[Callable[..., TRet]], Callable[..., TRet]]:
     if cols is None:
         cols = row_class.cols()
 
     return _prepared_statement(
         f"SELECT {', '.join(cols)} FROM {row_class.TABLE} {clauses}"
     )
 
 
 def _prepared_select_statements(
     row_class: Type[BaseRow], queries: Dict[Any, str],
 ) -> Callable[[Callable[..., TRet]], Callable[..., TRet]]:
     """Like _prepared_statement, but supports multiple statements, passed a dict,
     and passes a dict of prepared statements to the decorated method"""
     cols = row_class.cols()
 
     statement_start = f"SELECT {', '.join(cols)} FROM {row_class.TABLE} "
 
     def decorator(f):
         @functools.wraps(f)
         def newf(self, *args, **kwargs) -> TRet:
             if f.__name__ not in self._prepared_statements:
                 self._prepared_statements[f.__name__] = {
                     key: self._session.prepare(statement_start + query)
                     for (key, query) in queries.items()
                 }
             return f(
                 self, *args, **kwargs, statements=self._prepared_statements[f.__name__]
             )
 
         return newf
 
     return decorator
 
 
 def _next_bytes_value(value: bytes) -> bytes:
     """Returns the next bytes value by incrementing the integer
     representation of the provided value and converting it back
     to bytes.
 
     For instance when prefix is b"abcd", it returns b"abce".
     """
     next_value_int = int.from_bytes(value, byteorder="big") + 1
     return next_value_int.to_bytes(
         (next_value_int.bit_length() + 7) // 8, byteorder="big"
     )
 
 
 class CqlRunner:
     """Class managing prepared statements and building queries to be sent
     to Cassandra."""
 
     def __init__(
         self, hosts: List[str], keyspace: str, port: int, consistency_level: str
     ):
         self._cluster = Cluster(
             hosts,
             port=port,
             execution_profiles=get_execution_profiles(consistency_level),
         )
         self._session = self._cluster.connect(keyspace)
         self._cluster.register_user_type(
             keyspace, "microtimestamp_with_timezone", TimestampWithTimezone
         )
         self._cluster.register_user_type(keyspace, "microtimestamp", Timestamp)
         self._cluster.register_user_type(keyspace, "person", Person)
 
         # directly a PreparedStatement for methods decorated with
         # @_prepared_statements (and its wrappers, _prepared_insert_statement,
         # _prepared_exists_statement, and _prepared_select_statement);
         # and a dict of PreparedStatements with @_prepared_select_statements
         self._prepared_statements: Dict[
             str, Union[PreparedStatement, Dict[Any, PreparedStatement]]
         ] = {}
 
     ##########################
     # Common utility functions
     ##########################
 
     MAX_RETRIES = 3
 
     @retry(
         wait=wait_random_exponential(multiplier=1, max=10),
         stop=stop_after_attempt(MAX_RETRIES),
         retry=retry_if_exception_type(CoordinationFailure),
     )
     def _execute_with_retries(self, statement, args) -> ResultSet:
         return self._session.execute(statement, args, timeout=1000.0)
 
     @_prepared_statement(
         "UPDATE object_count SET count = count + ? "
         "WHERE partition_key = 0 AND object_type = ?"
     )
     def _increment_counter(
         self, object_type: str, nb: int, *, statement: PreparedStatement
     ) -> None:
         self._execute_with_retries(statement, [nb, object_type])
 
     def _add_one(self, statement, obj: BaseRow) -> None:
         self._increment_counter(obj.TABLE, 1)
         self._execute_with_retries(statement, dataclasses.astuple(obj))
 
     _T = TypeVar("_T", bound=BaseRow)
 
     def _get_random_row(self, row_class: Type[_T], statement) -> Optional[_T]:  # noqa
         """Takes a prepared statement of the form
         "SELECT * FROM <table> WHERE token(<keys>) > ? LIMIT 1"
         and uses it to return a random row"""
         token = random.randint(TOKEN_BEGIN, TOKEN_END)
         rows = self._execute_with_retries(statement, [token])
         if not rows:
             # There are no row with a greater token; wrap around to get
             # the row with the smallest token
             rows = self._execute_with_retries(statement, [TOKEN_BEGIN])
         if rows:
             return row_class.from_dict(rows.one())  # type: ignore
         else:
             return None
 
     def _missing(self, statement, ids):
         found_ids = set()
         for id_group in grouper(ids, PARTITION_KEY_RESTRICTION_MAX_SIZE):
             rows = self._execute_with_retries(statement, [list(id_group)])
             found_ids.update(row["id"] for row in rows)
         return [id_ for id_ in ids if id_ not in found_ids]
 
     ##########################
     # 'content' table
     ##########################
 
     def _content_add_finalize(self, statement: BoundStatement) -> None:
         """Returned currified by content_add_prepare, to be called when the
         content row should be added to the primary table."""
         self._execute_with_retries(statement, None)
         self._increment_counter("content", 1)
 
     @_prepared_insert_statement(ContentRow)
     def content_add_prepare(
         self, content: ContentRow, *, statement
     ) -> Tuple[int, Callable[[], None]]:
         """Prepares insertion of a Content to the main 'content' table.
         Returns a token (to be used in secondary tables), and a function to be
         called to perform the insertion in the main table."""
         statement = statement.bind(dataclasses.astuple(content))
 
         # Type used for hashing keys (usually, it will be
         # cassandra.metadata.Murmur3Token)
         token_class = self._cluster.metadata.token_map.token_class
 
         # Token of the row when it will be inserted. This is equivalent to
         # "SELECT token({', '.join(ContentRow.PARTITION_KEY)}) FROM content WHERE ..."
         # after the row is inserted; but we need the token to insert in the
         # index tables *before* inserting to the main 'content' table
         token = token_class.from_key(statement.routing_key).value
         assert TOKEN_BEGIN <= token <= TOKEN_END
 
         # Function to be called after the indexes contain their respective
         # row
         finalizer = functools.partial(self._content_add_finalize, statement)
 
         return (token, finalizer)
 
     @_prepared_select_statement(
         ContentRow, f"WHERE {' AND '.join(map('%s = ?'.__mod__, HASH_ALGORITHMS))}"
     )
     def content_get_from_pk(
         self, content_hashes: Dict[str, bytes], *, statement
     ) -> Optional[ContentRow]:
         rows = list(
             self._execute_with_retries(
                 statement, [content_hashes[algo] for algo in HASH_ALGORITHMS]
             )
         )
         assert len(rows) <= 1
         if rows:
             return ContentRow(**rows[0])
         else:
             return None
 
     def content_missing_from_hashes(
         self, contents_hashes: List[Dict[str, bytes]]
     ) -> Iterator[Dict[str, bytes]]:
         for group in grouper(contents_hashes, PARTITION_KEY_RESTRICTION_MAX_SIZE):
             group = list(group)
 
             # Get all contents that share a sha256 with one of the contents in the group
             present = set(
                 self._content_get_hashes_from_sha256(
                     [content["sha256"] for content in group]
                 )
             )
 
             for content in group:
                 for algo in HASH_ALGORITHMS:
                     assert content.get(algo) is not None, (
                         "content_missing_from_hashes must not be called with "
                         "partial hashes."
                     )
                 if tuple(content[algo] for algo in HASH_ALGORITHMS) not in present:
                     yield content
 
     @_prepared_statement(
         f"SELECT {', '.join(HASH_ALGORITHMS)} FROM content WHERE sha256 IN ?"
     )
     def _content_get_hashes_from_sha256(
         self, ids: List[bytes], *, statement
     ) -> Iterator[Tuple[bytes, bytes, bytes, bytes]]:
         for row in self._execute_with_retries(statement, [ids]):
             yield tuple(row[algo] for algo in HASH_ALGORITHMS)  # type: ignore
 
     @_prepared_select_statement(
         ContentRow, f"WHERE token({', '.join(ContentRow.PARTITION_KEY)}) = ?"
     )
     def content_get_from_token(self, token, *, statement) -> Iterable[ContentRow]:
         return map(ContentRow.from_dict, self._execute_with_retries(statement, [token]))
 
     @_prepared_select_statement(
         ContentRow, f"WHERE token({', '.join(ContentRow.PARTITION_KEY)}) > ? LIMIT 1"
     )
     def content_get_random(self, *, statement) -> Optional[ContentRow]:
         return self._get_random_row(ContentRow, statement)
 
     @_prepared_statement(
         """
         SELECT token({pk}) AS tok, {cols} FROM {table}
         WHERE token({pk}) >= ? AND token({pk}) <= ? LIMIT ?
         """.format(
             pk=", ".join(ContentRow.PARTITION_KEY),
             cols=", ".join(ContentRow.cols()),
             table=ContentRow.TABLE,
         )
     )
     def content_get_token_range(
         self, start: int, end: int, limit: int, *, statement
     ) -> Iterable[Tuple[int, ContentRow]]:
         """Returns an iterable of (token, row)"""
         return (
             (row["tok"], ContentRow.from_dict(remove_keys(row, ("tok",))))
             for row in self._execute_with_retries(statement, [start, end, limit])
         )
 
     ##########################
     # 'content_by_*' tables
     ##########################
 
     @_prepared_statement(
         f"""
         SELECT sha1_git AS id
         FROM {content_index_table_name("sha1_git", skipped_content=False)}
         WHERE sha1_git IN ?
         """
     )
     def content_missing_by_sha1_git(
         self, ids: List[bytes], *, statement
     ) -> List[bytes]:
         return self._missing(statement, ids)
 
     def content_index_add_one(self, algo: str, content: Content, token: int) -> None:
         """Adds a row mapping content[algo] to the token of the Content in
         the main 'content' table."""
         query = f"""
             INSERT INTO {content_index_table_name(algo, skipped_content=False)}
             ({algo}, target_token)
             VALUES (%s, %s)
         """
         self._execute_with_retries(query, [content.get_hash(algo), token])
 
     def content_get_tokens_from_single_hash(
         self, algo: str, hash_: bytes
     ) -> Iterable[int]:
         assert algo in HASH_ALGORITHMS
         query = f"""
             SELECT target_token
             FROM {content_index_table_name(algo, skipped_content=False)}
             WHERE {algo} = %s
         """
         return (
             row["target_token"] for row in self._execute_with_retries(query, [hash_])
         )
 
     ##########################
     # 'skipped_content' table
     ##########################
 
     def _skipped_content_add_finalize(self, statement: BoundStatement) -> None:
         """Returned currified by skipped_content_add_prepare, to be called
         when the content row should be added to the primary table."""
         self._execute_with_retries(statement, None)
         self._increment_counter("skipped_content", 1)
 
     @_prepared_insert_statement(SkippedContentRow)
     def skipped_content_add_prepare(
         self, content, *, statement
     ) -> Tuple[int, Callable[[], None]]:
         """Prepares insertion of a Content to the main 'skipped_content' table.
         Returns a token (to be used in secondary tables), and a function to be
         called to perform the insertion in the main table."""
 
         # Replace NULLs (which are not allowed in the partition key) with
         # an empty byte string
         for key in SkippedContentRow.PARTITION_KEY:
             if getattr(content, key) is None:
                 setattr(content, key, MAGIC_NULL_PK)
 
         statement = statement.bind(dataclasses.astuple(content))
 
         # Type used for hashing keys (usually, it will be
         # cassandra.metadata.Murmur3Token)
         token_class = self._cluster.metadata.token_map.token_class
 
         # Token of the row when it will be inserted. This is equivalent to
         # "SELECT token({', '.join(SkippedContentRow.PARTITION_KEY)})
         #  FROM skipped_content WHERE ..."
         # after the row is inserted; but we need the token to insert in the
         # index tables *before* inserting to the main 'skipped_content' table
         token = token_class.from_key(statement.routing_key).value
         assert TOKEN_BEGIN <= token <= TOKEN_END
 
         # Function to be called after the indexes contain their respective
         # row
         finalizer = functools.partial(self._skipped_content_add_finalize, statement)
 
         return (token, finalizer)
 
     @_prepared_select_statement(
         SkippedContentRow,
         f"WHERE {' AND '.join(map('%s = ?'.__mod__, HASH_ALGORITHMS))}",
     )
     def skipped_content_get_from_pk(
         self, content_hashes: Dict[str, bytes], *, statement
     ) -> Optional[SkippedContentRow]:
         rows = list(
             self._execute_with_retries(
                 statement,
                 [content_hashes[algo] or MAGIC_NULL_PK for algo in HASH_ALGORITHMS],
             )
         )
         assert len(rows) <= 1
         if rows:
             return SkippedContentRow.from_dict(rows[0])
         else:
             return None
 
     @_prepared_select_statement(
         SkippedContentRow,
         f"WHERE token({', '.join(SkippedContentRow.PARTITION_KEY)}) = ?",
     )
     def skipped_content_get_from_token(
         self, token, *, statement
     ) -> Iterable[SkippedContentRow]:
         return map(
             SkippedContentRow.from_dict, self._execute_with_retries(statement, [token])
         )
 
     ##########################
     # 'skipped_content_by_*' tables
     ##########################
 
     def skipped_content_index_add_one(
         self, algo: str, content: SkippedContent, token: int
     ) -> None:
         """Adds a row mapping content[algo] to the token of the SkippedContent
         in the main 'skipped_content' table."""
         query = (
             f"INSERT INTO skipped_content_by_{algo} ({algo}, target_token) "
             f"VALUES (%s, %s)"
         )
         self._execute_with_retries(
             query, [content.get_hash(algo) or MAGIC_NULL_PK, token]
         )
 
     def skipped_content_get_tokens_from_single_hash(
         self, algo: str, hash_: bytes
     ) -> Iterable[int]:
         assert algo in HASH_ALGORITHMS
         query = f"""
             SELECT target_token
             FROM {content_index_table_name(algo, skipped_content=True)}
             WHERE {algo} = %s
         """
         return (
             row["target_token"] for row in self._execute_with_retries(query, [hash_])
         )
 
     ##########################
     # 'revision' table
     ##########################
 
     @_prepared_exists_statement("revision")
     def revision_missing(self, ids: List[bytes], *, statement) -> List[bytes]:
         return self._missing(statement, ids)
 
     @_prepared_insert_statement(RevisionRow)
     def revision_add_one(self, revision: RevisionRow, *, statement) -> None:
         self._add_one(statement, revision)
 
     @_prepared_statement(f"SELECT id FROM {RevisionRow.TABLE} WHERE id IN ?")
     def revision_get_ids(self, revision_ids, *, statement) -> Iterable[int]:
         return (
             row["id"] for row in self._execute_with_retries(statement, [revision_ids])
         )
 
     @_prepared_select_statement(RevisionRow, "WHERE id IN ?")
     def revision_get(
         self, revision_ids: List[Sha1Git], *, statement
     ) -> Iterable[RevisionRow]:
         return map(
             RevisionRow.from_dict, self._execute_with_retries(statement, [revision_ids])
         )
 
     @_prepared_select_statement(RevisionRow, "WHERE token(id) > ? LIMIT 1")
     def revision_get_random(self, *, statement) -> Optional[RevisionRow]:
         return self._get_random_row(RevisionRow, statement)
 
     ##########################
     # 'revision_parent' table
     ##########################
 
     @_prepared_insert_statement(RevisionParentRow)
     def revision_parent_add_one(
         self, revision_parent: RevisionParentRow, *, statement
     ) -> None:
         self._add_one(statement, revision_parent)
 
     @_prepared_statement(
         f"SELECT parent_id FROM {RevisionParentRow.TABLE} WHERE id = ?"
     )
     def revision_parent_get(
         self, revision_id: Sha1Git, *, statement
     ) -> Iterable[bytes]:
         return (
             row["parent_id"]
             for row in self._execute_with_retries(statement, [revision_id])
         )
 
     ##########################
     # 'release' table
     ##########################
 
     @_prepared_exists_statement("release")
     def release_missing(self, ids: List[bytes], *, statement) -> List[bytes]:
         return self._missing(statement, ids)
 
     @_prepared_insert_statement(ReleaseRow)
     def release_add_one(self, release: ReleaseRow, *, statement) -> None:
         self._add_one(statement, release)
 
     @_prepared_select_statement(ReleaseRow, "WHERE id in ?")
     def release_get(self, release_ids: List[str], *, statement) -> Iterable[ReleaseRow]:
         return map(
             ReleaseRow.from_dict, self._execute_with_retries(statement, [release_ids])
         )
 
     @_prepared_select_statement(ReleaseRow, "WHERE token(id) > ? LIMIT 1")
     def release_get_random(self, *, statement) -> Optional[ReleaseRow]:
         return self._get_random_row(ReleaseRow, statement)
 
     ##########################
     # 'directory' table
     ##########################
 
     @_prepared_exists_statement("directory")
     def directory_missing(self, ids: List[bytes], *, statement) -> List[bytes]:
         return self._missing(statement, ids)
 
     @_prepared_insert_statement(DirectoryRow)
     def directory_add_one(self, directory: DirectoryRow, *, statement) -> None:
         """Called after all calls to directory_entry_add_one, to
         commit/finalize the directory."""
         self._add_one(statement, directory)
 
     @_prepared_select_statement(DirectoryRow, "WHERE token(id) > ? LIMIT 1")
     def directory_get_random(self, *, statement) -> Optional[DirectoryRow]:
         return self._get_random_row(DirectoryRow, statement)
 
     ##########################
     # 'directory_entry' table
     ##########################
 
     @_prepared_insert_statement(DirectoryEntryRow)
     def directory_entry_add_one(self, entry: DirectoryEntryRow, *, statement) -> None:
         self._add_one(statement, entry)
 
     @_prepared_select_statement(DirectoryEntryRow, "WHERE directory_id IN ?")
     def directory_entry_get(
         self, directory_ids, *, statement
     ) -> Iterable[DirectoryEntryRow]:
         return map(
             DirectoryEntryRow.from_dict,
             self._execute_with_retries(statement, [directory_ids]),
         )
 
     @_prepared_select_statement(
         DirectoryEntryRow, "WHERE directory_id = ? AND name >= ? LIMIT ?"
     )
     def directory_entry_get_from_name(
         self, directory_id: Sha1Git, from_: bytes, limit: int, *, statement
     ) -> Iterable[DirectoryEntryRow]:
         return map(
             DirectoryEntryRow.from_dict,
             self._execute_with_retries(statement, [directory_id, from_, limit]),
         )
 
     ##########################
     # 'snapshot' table
     ##########################
 
     @_prepared_exists_statement("snapshot")
     def snapshot_missing(self, ids: List[bytes], *, statement) -> List[bytes]:
         return self._missing(statement, ids)
 
     @_prepared_insert_statement(SnapshotRow)
     def snapshot_add_one(self, snapshot: SnapshotRow, *, statement) -> None:
         self._add_one(statement, snapshot)
 
     @_prepared_select_statement(SnapshotRow, "WHERE token(id) > ? LIMIT 1")
     def snapshot_get_random(self, *, statement) -> Optional[SnapshotRow]:
         return self._get_random_row(SnapshotRow, statement)
 
     ##########################
     # 'snapshot_branch' table
     ##########################
 
     @_prepared_insert_statement(SnapshotBranchRow)
     def snapshot_branch_add_one(self, branch: SnapshotBranchRow, *, statement) -> None:
         self._add_one(statement, branch)
 
     @_prepared_statement(
         f"""
         SELECT ascii_bins_count(target_type) AS counts
         FROM {SnapshotBranchRow.TABLE}
         WHERE snapshot_id = ? AND name >= ?
         """
     )
     def snapshot_count_branches_from_name(
         self, snapshot_id: Sha1Git, from_: bytes, *, statement
     ) -> Dict[Optional[str], int]:
         row = self._execute_with_retries(statement, [snapshot_id, from_]).one()
         (nb_none, counts) = row["counts"]
         return {None: nb_none, **counts}
 
     @_prepared_statement(
         f"""
         SELECT ascii_bins_count(target_type) AS counts
         FROM {SnapshotBranchRow.TABLE}
         WHERE snapshot_id = ? AND name < ?
         """
     )
     def snapshot_count_branches_before_name(
         self, snapshot_id: Sha1Git, before: bytes, *, statement,
     ) -> Dict[Optional[str], int]:
         row = self._execute_with_retries(statement, [snapshot_id, before]).one()
         (nb_none, counts) = row["counts"]
         return {None: nb_none, **counts}
 
     def snapshot_count_branches(
         self, snapshot_id: Sha1Git, branch_name_exclude_prefix: Optional[bytes] = None,
     ) -> Dict[Optional[str], int]:
         """Returns a dictionary from type names to the number of branches
         of that type."""
         prefix = branch_name_exclude_prefix
         if prefix is None:
             return self.snapshot_count_branches_from_name(snapshot_id, b"")
         else:
             # counts branches before exclude prefix
             counts = Counter(
                 self.snapshot_count_branches_before_name(snapshot_id, prefix)
             )
 
             # no need to execute that part if each bit of the prefix equals 1
             if prefix.replace(b"\xff", b"") != b"":
                 # counts branches after exclude prefix and update counters
                 counts.update(
                     self.snapshot_count_branches_from_name(
                         snapshot_id, _next_bytes_value(prefix)
                     )
                 )
             return counts
 
     @_prepared_select_statement(
         SnapshotBranchRow, "WHERE snapshot_id = ? AND name >= ? LIMIT ?"
     )
     def snapshot_branch_get_from_name(
         self, snapshot_id: Sha1Git, from_: bytes, limit: int, *, statement
     ) -> Iterable[SnapshotBranchRow]:
         return map(
             SnapshotBranchRow.from_dict,
             self._execute_with_retries(statement, [snapshot_id, from_, limit]),
         )
 
     @_prepared_select_statement(
         SnapshotBranchRow, "WHERE snapshot_id = ? AND name >= ? AND name < ? LIMIT ?"
     )
     def snapshot_branch_get_range(
         self,
         snapshot_id: Sha1Git,
         from_: bytes,
         before: bytes,
         limit: int,
         *,
         statement,
     ) -> Iterable[SnapshotBranchRow]:
         return map(
             SnapshotBranchRow.from_dict,
             self._execute_with_retries(statement, [snapshot_id, from_, before, limit]),
         )
 
     def snapshot_branch_get(
         self,
         snapshot_id: Sha1Git,
         from_: bytes,
         limit: int,
         branch_name_exclude_prefix: Optional[bytes] = None,
     ) -> Iterable[SnapshotBranchRow]:
         prefix = branch_name_exclude_prefix
         if prefix is None:
             return self.snapshot_branch_get_from_name(snapshot_id, from_, limit)
         else:
             # get branches before the exclude prefix
             branches = list(
                 self.snapshot_branch_get_range(snapshot_id, from_, prefix, limit)
             )
             nb_branches = len(branches)
             # no need to execute that part if limit is reached
             # or if each bit of the prefix equals 1
             if nb_branches < limit and prefix.replace(b"\xff", b"") != b"":
                 # get branches after the exclude prefix and update list to return
                 branches.extend(
                     self.snapshot_branch_get_from_name(
                         snapshot_id, _next_bytes_value(prefix), limit - nb_branches
                     )
                 )
             return branches
 
     ##########################
     # 'origin' table
     ##########################
 
     @_prepared_insert_statement(OriginRow)
     def origin_add_one(self, origin: OriginRow, *, statement) -> None:
         self._add_one(statement, origin)
 
     @_prepared_select_statement(OriginRow, "WHERE sha1 = ?")
     def origin_get_by_sha1(self, sha1: bytes, *, statement) -> Iterable[OriginRow]:
         return map(OriginRow.from_dict, self._execute_with_retries(statement, [sha1]))
 
     def origin_get_by_url(self, url: str) -> Iterable[OriginRow]:
         return self.origin_get_by_sha1(hash_url(url))
 
     @_prepared_statement(
         f"""
         SELECT token(sha1) AS tok, {", ".join(OriginRow.cols())}
         FROM {OriginRow.TABLE}
         WHERE token(sha1) >= ? LIMIT ?
         """
     )
     def origin_list(
         self, start_token: int, limit: int, *, statement
     ) -> Iterable[Tuple[int, OriginRow]]:
         """Returns an iterable of (token, origin)"""
         return (
             (row["tok"], OriginRow.from_dict(remove_keys(row, ("tok",))))
             for row in self._execute_with_retries(statement, [start_token, limit])
         )
 
     @_prepared_select_statement(OriginRow)
     def origin_iter_all(self, *, statement) -> Iterable[OriginRow]:
         return map(OriginRow.from_dict, self._execute_with_retries(statement, []))
 
+    @_prepared_statement(
+        f"""
+        UPDATE {OriginRow.TABLE}
+        SET next_visit_id=?
+        WHERE sha1 = ? IF next_visit_id<?
+        """
+    )
+    def origin_bump_next_visit_id(
+        self, origin_url: str, visit_id: int, *, statement
+    ) -> None:
+        origin_sha1 = hash_url(origin_url)
+        next_id = visit_id + 1
+        self._execute_with_retries(statement, [next_id, origin_sha1, next_id])
+
     @_prepared_statement(f"SELECT next_visit_id FROM {OriginRow.TABLE} WHERE sha1 = ?")
     def _origin_get_next_visit_id(self, origin_sha1: bytes, *, statement) -> int:
         rows = list(self._execute_with_retries(statement, [origin_sha1]))
         assert len(rows) == 1  # TODO: error handling
         return rows[0]["next_visit_id"]
 
     @_prepared_statement(
         f"""
         UPDATE {OriginRow.TABLE}
         SET next_visit_id=?
         WHERE sha1 = ? IF next_visit_id=?
         """
     )
     def origin_generate_unique_visit_id(self, origin_url: str, *, statement) -> int:
         origin_sha1 = hash_url(origin_url)
         next_id = self._origin_get_next_visit_id(origin_sha1)
         while True:
             res = list(
                 self._execute_with_retries(
                     statement, [next_id + 1, origin_sha1, next_id]
                 )
             )
             assert len(res) == 1
             if res[0]["[applied]"]:
                 # No data race
                 return next_id
             else:
                 # Someone else updated it before we did, let's try again
                 next_id = res[0]["next_visit_id"]
                 # TODO: abort after too many attempts
 
         return next_id
 
     ##########################
     # 'origin_visit' table
     ##########################
 
     @_prepared_select_statements(
         OriginVisitRow,
         {
             (True, ListOrder.ASC): (
                 "WHERE origin = ? AND visit > ? ORDER BY visit ASC LIMIT ?"
             ),
             (True, ListOrder.DESC): (
                 "WHERE origin = ? AND visit < ? ORDER BY visit DESC LIMIT ?"
             ),
             (False, ListOrder.ASC): "WHERE origin = ? ORDER BY visit ASC LIMIT ?",
             (False, ListOrder.DESC): "WHERE origin = ? ORDER BY visit DESC LIMIT ?",
         },
     )
     def origin_visit_get(
         self,
         origin_url: str,
         last_visit: Optional[int],
         limit: int,
         order: ListOrder,
         *,
         statements,
     ) -> Iterable[OriginVisitRow]:
         args: List[Any] = [origin_url]
 
         if last_visit is not None:
             args.append(last_visit)
 
         args.append(limit)
 
         statement = statements[(last_visit is not None, order)]
         return map(
             OriginVisitRow.from_dict, self._execute_with_retries(statement, args)
         )
 
     @_prepared_insert_statement(OriginVisitRow)
     def origin_visit_add_one(self, visit: OriginVisitRow, *, statement) -> None:
         self._add_one(statement, visit)
 
     @_prepared_select_statement(OriginVisitRow, "WHERE origin = ? AND visit = ?")
     def origin_visit_get_one(
         self, origin_url: str, visit_id: int, *, statement
     ) -> Optional[OriginVisitRow]:
         # TODO: error handling
         rows = list(self._execute_with_retries(statement, [origin_url, visit_id]))
         if rows:
             return OriginVisitRow.from_dict(rows[0])
         else:
             return None
 
     @_prepared_select_statement(OriginVisitRow, "WHERE origin = ?")
     def origin_visit_get_all(
         self, origin_url: str, *, statement
     ) -> Iterable[OriginVisitRow]:
         return map(
             OriginVisitRow.from_dict,
             self._execute_with_retries(statement, [origin_url]),
         )
 
     @_prepared_select_statement(OriginVisitRow, "WHERE token(origin) >= ?")
     def _origin_visit_iter_from(
         self, min_token: int, *, statement
     ) -> Iterable[OriginVisitRow]:
         return map(
             OriginVisitRow.from_dict, self._execute_with_retries(statement, [min_token])
         )
 
     @_prepared_select_statement(OriginVisitRow, "WHERE token(origin) < ?")
     def _origin_visit_iter_to(
         self, max_token: int, *, statement
     ) -> Iterable[OriginVisitRow]:
         return map(
             OriginVisitRow.from_dict, self._execute_with_retries(statement, [max_token])
         )
 
     def origin_visit_iter(self, start_token: int) -> Iterator[OriginVisitRow]:
         """Returns all origin visits in order from this token,
         and wraps around the token space."""
         yield from self._origin_visit_iter_from(start_token)
         yield from self._origin_visit_iter_to(start_token)
 
     ##########################
     # 'origin_visit_status' table
     ##########################
 
     @_prepared_select_statements(
         OriginVisitStatusRow,
         {
             (True, ListOrder.ASC): (
                 "WHERE origin = ? AND visit = ? AND date >= ? "
                 "ORDER BY visit ASC LIMIT ?"
             ),
             (True, ListOrder.DESC): (
                 "WHERE origin = ? AND visit = ? AND date <= ? "
                 "ORDER BY visit DESC LIMIT ?"
             ),
             (False, ListOrder.ASC): (
                 "WHERE origin = ? AND visit = ? ORDER BY visit ASC LIMIT ?"
             ),
             (False, ListOrder.DESC): (
                 "WHERE origin = ? AND visit = ? ORDER BY visit DESC LIMIT ?"
             ),
         },
     )
     def origin_visit_status_get_range(
         self,
         origin: str,
         visit: int,
         date_from: Optional[datetime.datetime],
         limit: int,
         order: ListOrder,
         *,
         statements,
     ) -> Iterable[OriginVisitStatusRow]:
         args: List[Any] = [origin, visit]
 
         if date_from is not None:
             args.append(date_from)
 
         args.append(limit)
 
         statement = statements[(date_from is not None, order)]
 
         return map(
             OriginVisitStatusRow.from_dict, self._execute_with_retries(statement, args)
         )
 
     @_prepared_insert_statement(OriginVisitStatusRow)
     def origin_visit_status_add_one(
         self, visit_update: OriginVisitStatusRow, *, statement
     ) -> None:
         self._add_one(statement, visit_update)
 
     def origin_visit_status_get_latest(
         self, origin: str, visit: int,
     ) -> Optional[OriginVisitStatusRow]:
         """Given an origin visit id, return its latest origin_visit_status
 
          """
         return next(self.origin_visit_status_get(origin, visit), None)
 
     @_prepared_select_statement(
         OriginVisitStatusRow,
         # 'visit DESC,' is optional with Cassandra 4, but ScyllaDB needs it
         "WHERE origin = ? AND visit = ? ORDER BY visit DESC, date DESC",
     )
     def origin_visit_status_get(
         self, origin: str, visit: int, *, statement,
     ) -> Iterator[OriginVisitStatusRow]:
         """Return all origin visit statuses for a given visit
 
         """
         return map(
             OriginVisitStatusRow.from_dict,
             self._execute_with_retries(statement, [origin, visit]),
         )
 
     ##########################
     # 'metadata_authority' table
     ##########################
 
     @_prepared_insert_statement(MetadataAuthorityRow)
     def metadata_authority_add(self, authority: MetadataAuthorityRow, *, statement):
         self._add_one(statement, authority)
 
     @_prepared_select_statement(MetadataAuthorityRow, "WHERE type = ? AND url = ?")
     def metadata_authority_get(
         self, type, url, *, statement
     ) -> Optional[MetadataAuthorityRow]:
         rows = list(self._execute_with_retries(statement, [type, url]))
         if rows:
             return MetadataAuthorityRow.from_dict(rows[0])
         else:
             return None
 
     ##########################
     # 'metadata_fetcher' table
     ##########################
 
     @_prepared_insert_statement(MetadataFetcherRow)
     def metadata_fetcher_add(self, fetcher, *, statement):
         self._add_one(statement, fetcher)
 
     @_prepared_select_statement(MetadataFetcherRow, "WHERE name = ? AND version = ?")
     def metadata_fetcher_get(
         self, name, version, *, statement
     ) -> Optional[MetadataFetcherRow]:
         rows = list(self._execute_with_retries(statement, [name, version]))
         if rows:
             return MetadataFetcherRow.from_dict(rows[0])
         else:
             return None
 
     #########################
     # 'raw_extrinsic_metadata_by_id' table
     #########################
 
     @_prepared_insert_statement(RawExtrinsicMetadataByIdRow)
     def raw_extrinsic_metadata_by_id_add(self, row, *, statement):
         self._add_one(statement, row)
 
     @_prepared_select_statement(RawExtrinsicMetadataByIdRow, "WHERE id IN ?")
     def raw_extrinsic_metadata_get_by_ids(
         self, ids: List[Sha1Git], *, statement
     ) -> Iterable[RawExtrinsicMetadataByIdRow]:
         return map(
             RawExtrinsicMetadataByIdRow.from_dict,
             self._execute_with_retries(statement, [ids]),
         )
 
     #########################
     # 'raw_extrinsic_metadata' table
     #########################
 
     @_prepared_insert_statement(RawExtrinsicMetadataRow)
     def raw_extrinsic_metadata_add(self, raw_extrinsic_metadata, *, statement):
         self._add_one(statement, raw_extrinsic_metadata)
 
     @_prepared_select_statement(
         RawExtrinsicMetadataRow,
         "WHERE target=? AND authority_url=? AND discovery_date>? AND authority_type=?",
     )
     def raw_extrinsic_metadata_get_after_date(
         self,
         target: str,
         authority_type: str,
         authority_url: str,
         after: datetime.datetime,
         *,
         statement,
     ) -> Iterable[RawExtrinsicMetadataRow]:
         return map(
             RawExtrinsicMetadataRow.from_dict,
             self._execute_with_retries(
                 statement, [target, authority_url, after, authority_type]
             ),
         )
 
     @_prepared_select_statement(
         RawExtrinsicMetadataRow,
         # This is equivalent to:
         #   WHERE target=? AND authority_type = ? AND authority_url = ? "
         #   AND (discovery_date, id) > (?, ?)"
         # but it needs to be written this way to work with ScyllaDB.
         "WHERE target=? AND (authority_type, authority_url) <= (?, ?) "
         "AND (authority_type, authority_url, discovery_date, id) > (?, ?, ?, ?)",
     )
     def raw_extrinsic_metadata_get_after_date_and_id(
         self,
         target: str,
         authority_type: str,
         authority_url: str,
         after_date: datetime.datetime,
         after_id: bytes,
         *,
         statement,
     ) -> Iterable[RawExtrinsicMetadataRow]:
         return map(
             RawExtrinsicMetadataRow.from_dict,
             self._execute_with_retries(
                 statement,
                 [
                     target,
                     authority_type,
                     authority_url,
                     authority_type,
                     authority_url,
                     after_date,
                     after_id,
                 ],
             ),
         )
 
     @_prepared_select_statement(
         RawExtrinsicMetadataRow,
         "WHERE target=? AND authority_url=? AND authority_type=?",
     )
     def raw_extrinsic_metadata_get(
         self, target: str, authority_type: str, authority_url: str, *, statement
     ) -> Iterable[RawExtrinsicMetadataRow]:
         return map(
             RawExtrinsicMetadataRow.from_dict,
             self._execute_with_retries(
                 statement, [target, authority_url, authority_type]
             ),
         )
 
     @_prepared_statement(
         "SELECT authority_type, authority_url FROM raw_extrinsic_metadata "
         "WHERE target = ?"
     )
     def raw_extrinsic_metadata_get_authorities(
         self, target: str, *, statement
     ) -> Iterable[Tuple[str, str]]:
         return (
             (entry["authority_type"], entry["authority_url"])
             for entry in self._execute_with_retries(statement, [target])
         )
 
     ##########################
     # 'extid' table
     ##########################
     def _extid_add_finalize(self, statement: BoundStatement) -> None:
         """Returned currified by extid_add_prepare, to be called when the
         extid row should be added to the primary table."""
         self._execute_with_retries(statement, None)
         self._increment_counter("extid", 1)
 
     @_prepared_insert_statement(ExtIDRow)
     def extid_add_prepare(
         self, extid: ExtIDRow, *, statement
     ) -> Tuple[int, Callable[[], None]]:
         statement = statement.bind(dataclasses.astuple(extid))
         token_class = self._cluster.metadata.token_map.token_class
         token = token_class.from_key(statement.routing_key).value
         assert TOKEN_BEGIN <= token <= TOKEN_END
 
         # Function to be called after the indexes contain their respective
         # row
         finalizer = functools.partial(self._extid_add_finalize, statement)
 
         return (token, finalizer)
 
     @_prepared_select_statement(
         ExtIDRow,
         "WHERE extid_type=? AND extid=? AND extid_version=? "
         "AND target_type=? AND target=?",
     )
     def extid_get_from_pk(
         self,
         extid_type: str,
         extid: bytes,
         extid_version: int,
         target: CoreSWHID,
         *,
         statement,
     ) -> Optional[ExtIDRow]:
         rows = list(
             self._execute_with_retries(
                 statement,
                 [
                     extid_type,
                     extid,
                     extid_version,
                     target.object_type.value,
                     target.object_id,
                 ],
             ),
         )
         assert len(rows) <= 1
         if rows:
             return ExtIDRow(**rows[0])
         else:
             return None
 
     @_prepared_select_statement(
         ExtIDRow, "WHERE token(extid_type, extid) = ?",
     )
     def extid_get_from_token(self, token: int, *, statement) -> Iterable[ExtIDRow]:
         return map(ExtIDRow.from_dict, self._execute_with_retries(statement, [token]),)
 
     @_prepared_select_statement(
         ExtIDRow, "WHERE extid_type=? AND extid=?",
     )
     def extid_get_from_extid(
         self, extid_type: str, extid: bytes, *, statement
     ) -> Iterable[ExtIDRow]:
         return map(
             ExtIDRow.from_dict,
             self._execute_with_retries(statement, [extid_type, extid]),
         )
 
     def extid_get_from_target(
         self, target_type: str, target: bytes
     ) -> Iterable[ExtIDRow]:
         for token in self._extid_get_tokens_from_target(target_type, target):
             if token is not None:
                 for extid in self.extid_get_from_token(token):
                     # re-check the extid against target (in case of murmur3 collision)
                     if (
                         extid is not None
                         and extid.target_type == target_type
                         and extid.target == target
                     ):
                         yield extid
 
     ##########################
     # 'extid_by_target' table
     ##########################
 
     @_prepared_insert_statement(ExtIDByTargetRow)
     def extid_index_add_one(self, row: ExtIDByTargetRow, *, statement) -> None:
         """Adds a row mapping extid[target_type, target] to the token of the ExtID in
         the main 'extid' table."""
         self._add_one(statement, row)
 
     @_prepared_statement(
         f"""
         SELECT target_token
         FROM {ExtIDByTargetRow.TABLE}
         WHERE target_type = ? AND target = ?
         """
     )
     def _extid_get_tokens_from_target(
         self, target_type: str, target: bytes, *, statement
     ) -> Iterable[int]:
         return (
             row["target_token"]
             for row in self._execute_with_retries(statement, [target_type, target])
         )
 
     ##########################
     # Miscellaneous
     ##########################
 
     @_prepared_statement("SELECT uuid() FROM revision LIMIT 1;")
     def check_read(self, *, statement):
         self._execute_with_retries(statement, [])
 
     @_prepared_select_statement(ObjectCountRow, "WHERE partition_key=0")
     def stat_counters(self, *, statement) -> Iterable[ObjectCountRow]:
         return map(ObjectCountRow.from_dict, self._execute_with_retries(statement, []))
diff --git a/swh/storage/cassandra/storage.py b/swh/storage/cassandra/storage.py
index d3061b43..b8f1f3dd 100644
--- a/swh/storage/cassandra/storage.py
+++ b/swh/storage/cassandra/storage.py
@@ -1,1551 +1,1555 @@
 # Copyright (C) 2019-2020  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 import base64
 import datetime
 import itertools
 import operator
 import random
 import re
 from typing import (
     Any,
     Callable,
     Counter,
     Dict,
     Iterable,
     List,
     Optional,
     Sequence,
     Set,
     Tuple,
     Union,
 )
 
 import attr
 
 from swh.core.api.classes import stream_results
 from swh.core.api.serializers import msgpack_dumps, msgpack_loads
 from swh.model.hashutil import DEFAULT_ALGORITHMS, hash_to_hex
 from swh.model.identifiers import CoreSWHID, ExtendedObjectType, ExtendedSWHID
 from swh.model.identifiers import ObjectType as SwhidObjectType
 from swh.model.model import (
     Content,
     Directory,
     DirectoryEntry,
     ExtID,
     MetadataAuthority,
     MetadataAuthorityType,
     MetadataFetcher,
     Origin,
     OriginVisit,
     OriginVisitStatus,
     RawExtrinsicMetadata,
     Release,
     Revision,
     Sha1Git,
     SkippedContent,
     Snapshot,
     SnapshotBranch,
     TargetType,
 )
 from swh.storage.interface import (
     VISIT_STATUSES,
     ListOrder,
     PagedResult,
     PartialBranches,
     Sha1,
 )
 from swh.storage.objstorage import ObjStorage
 from swh.storage.utils import map_optional, now
 from swh.storage.writer import JournalWriter
 
 from . import converters
 from ..exc import HashCollision, StorageArgumentException
 from ..utils import remove_keys
 from .common import TOKEN_BEGIN, TOKEN_END, hash_url
 from .cql import CqlRunner
 from .model import (
     ContentRow,
     DirectoryEntryRow,
     DirectoryRow,
     ExtIDByTargetRow,
     ExtIDRow,
     MetadataAuthorityRow,
     MetadataFetcherRow,
     OriginRow,
     OriginVisitRow,
     OriginVisitStatusRow,
     RawExtrinsicMetadataByIdRow,
     RawExtrinsicMetadataRow,
     RevisionParentRow,
     SkippedContentRow,
     SnapshotBranchRow,
     SnapshotRow,
 )
 from .schema import HASH_ALGORITHMS
 
 # Max block size of contents to return
 BULK_BLOCK_CONTENT_LEN_MAX = 10000
 
 
 class CassandraStorage:
     def __init__(
         self,
         hosts,
         keyspace,
         objstorage,
         port=9042,
         journal_writer=None,
         allow_overwrite=False,
         consistency_level="ONE",
     ):
         """
         A backend of swh-storage backed by Cassandra
 
         Args:
             hosts: Seed Cassandra nodes, to start connecting to the cluster
             keyspace: Name of the Cassandra database to use
             objstorage: Passed as argument to :class:`ObjStorage`
             port: Cassandra port
             journal_writer: Passed as argument to :class:`JournalWriter`
             allow_overwrite: Whether ``*_add`` functions will check if an object
                 already exists in the database before sending it in an INSERT.
                 ``False`` is the default as it is more efficient when there is
                 a moderately high probability the object is already known,
                 but ``True`` can be useful to overwrite existing objects
                 (eg. when applying a schema update),
                 or when the database is known to be mostly empty.
                 Note that a ``False`` value does not guarantee there won't be
                 any overwrite.
             consistency_level: The default read/write consistency to use
         """
         self._hosts = hosts
         self._keyspace = keyspace
         self._port = port
         self._consistency_level = consistency_level
         self._set_cql_runner()
         self.journal_writer: JournalWriter = JournalWriter(journal_writer)
         self.objstorage: ObjStorage = ObjStorage(objstorage)
         self._allow_overwrite = allow_overwrite
 
     def _set_cql_runner(self):
         """Used by tests when they need to reset the CqlRunner"""
         self._cql_runner: CqlRunner = CqlRunner(
             self._hosts, self._keyspace, self._port, self._consistency_level
         )
 
     def check_config(self, *, check_write: bool) -> bool:
         self._cql_runner.check_read()
 
         return True
 
     def _content_get_from_hash(self, algo, hash_) -> Iterable:
         """From the name of a hash algorithm and a value of that hash,
         looks up the "hash -> token" secondary table (content_by_{algo})
         to get tokens.
         Then, looks up the main table (content) to get all contents with
         that token, and filters out contents whose hash doesn't match."""
         found_tokens = self._cql_runner.content_get_tokens_from_single_hash(algo, hash_)
 
         for token in found_tokens:
             assert isinstance(token, int), found_tokens
             # Query the main table ('content').
             res = self._cql_runner.content_get_from_token(token)
 
             for row in res:
                 # re-check the the hash (in case of murmur3 collision)
                 if getattr(row, algo) == hash_:
                     yield row
 
     def _content_add(self, contents: List[Content], with_data: bool) -> Dict[str, int]:
         # Filter-out content already in the database.
         if not self._allow_overwrite:
             contents = [
                 c
                 for c in contents
                 if not self._cql_runner.content_get_from_pk(c.to_dict())
             ]
 
         if with_data:
             # First insert to the objstorage, if the endpoint is
             # `content_add` (as opposed to `content_add_metadata`).
 
             # Must add to the objstorage before the DB and journal. Otherwise:
             # 1. in case of a crash the DB may "believe" we have the content, but
             #    we didn't have time to write to the objstorage before the crash
             # 2. the objstorage mirroring, which reads from the journal, may attempt to
             #    read from the objstorage before we finished writing it
             summary = self.objstorage.content_add(
                 c for c in contents if c.status != "absent"
             )
             content_add_bytes = summary["content:add:bytes"]
 
         self.journal_writer.content_add(contents)
 
         content_add = 0
         for content in contents:
             content_add += 1
 
             # Check for sha1 or sha1_git collisions. This test is not atomic
             # with the insertion, so it won't detect a collision if both
             # contents are inserted at the same time, but it's good enough.
             #
             # The proper way to do it would probably be a BATCH, but this
             # would be inefficient because of the number of partitions we
             # need to affect (len(HASH_ALGORITHMS)+1, which is currently 5)
             if not self._allow_overwrite:
                 for algo in {"sha1", "sha1_git"}:
                     collisions = []
                     # Get tokens of 'content' rows with the same value for
                     # sha1/sha1_git
                     rows = self._content_get_from_hash(algo, content.get_hash(algo))
                     for row in rows:
                         if getattr(row, algo) != content.get_hash(algo):
                             # collision of token(partition key), ignore this
                             # row
                             continue
 
                         for other_algo in HASH_ALGORITHMS:
                             if getattr(row, other_algo) != content.get_hash(other_algo):
                                 # This hash didn't match; discard the row.
                                 collisions.append(
                                     {k: getattr(row, k) for k in HASH_ALGORITHMS}
                                 )
 
                     if collisions:
                         collisions.append(content.hashes())
                         raise HashCollision(algo, content.get_hash(algo), collisions)
 
             (token, insertion_finalizer) = self._cql_runner.content_add_prepare(
                 ContentRow(**remove_keys(content.to_dict(), ("data",)))
             )
 
             # Then add to index tables
             for algo in HASH_ALGORITHMS:
                 self._cql_runner.content_index_add_one(algo, content, token)
 
             # Then to the main table
             insertion_finalizer()
 
         summary = {
             "content:add": content_add,
         }
 
         if with_data:
             summary["content:add:bytes"] = content_add_bytes
 
         return summary
 
     def content_add(self, content: List[Content]) -> Dict[str, int]:
         to_add = {
             (c.sha1, c.sha1_git, c.sha256, c.blake2s256): c for c in content
         }.values()
         contents = [attr.evolve(c, ctime=now()) for c in to_add]
         return self._content_add(list(contents), with_data=True)
 
     def content_update(
         self, contents: List[Dict[str, Any]], keys: List[str] = []
     ) -> None:
         raise NotImplementedError(
             "content_update is not supported by the Cassandra backend"
         )
 
     def content_add_metadata(self, content: List[Content]) -> Dict[str, int]:
         return self._content_add(content, with_data=False)
 
     def content_get_data(self, content: Sha1) -> Optional[bytes]:
         # FIXME: Make this method support slicing the `data`
         return self.objstorage.content_get(content)
 
     def content_get_partition(
         self,
         partition_id: int,
         nb_partitions: int,
         page_token: Optional[str] = None,
         limit: int = 1000,
     ) -> PagedResult[Content]:
         if limit is None:
             raise StorageArgumentException("limit should not be None")
 
         # Compute start and end of the range of tokens covered by the
         # requested partition
         partition_size = (TOKEN_END - TOKEN_BEGIN) // nb_partitions
         range_start = TOKEN_BEGIN + partition_id * partition_size
         range_end = TOKEN_BEGIN + (partition_id + 1) * partition_size
 
         # offset the range start according to the `page_token`.
         if page_token is not None:
             if not (range_start <= int(page_token) <= range_end):
                 raise StorageArgumentException("Invalid page_token.")
             range_start = int(page_token)
 
         next_page_token: Optional[str] = None
 
         rows = self._cql_runner.content_get_token_range(
             range_start, range_end, limit + 1
         )
         contents = []
         for counter, (tok, row) in enumerate(rows):
             if row.status == "absent":
                 continue
             row_d = row.to_dict()
             if counter >= limit:
                 next_page_token = str(tok)
                 break
             row_d.pop("ctime")
             contents.append(Content(**row_d))
 
         assert len(contents) <= limit
         return PagedResult(results=contents, next_page_token=next_page_token)
 
     def content_get(
         self, contents: List[bytes], algo: str = "sha1"
     ) -> List[Optional[Content]]:
         if algo not in DEFAULT_ALGORITHMS:
             raise StorageArgumentException(
                 "algo should be one of {','.join(DEFAULT_ALGORITHMS)}"
             )
 
         key = operator.attrgetter(algo)
         contents_by_hash: Dict[Sha1, Optional[Content]] = {}
         for hash_ in contents:
             # Get all (sha1, sha1_git, sha256, blake2s256) whose sha1/sha1_git
             # matches the argument, from the index table ('content_by_*')
             for row in self._content_get_from_hash(algo, hash_):
                 row_d = row.to_dict()
                 row_d.pop("ctime")
                 content = Content(**row_d)
                 contents_by_hash[key(content)] = content
         return [contents_by_hash.get(hash_) for hash_ in contents]
 
     def content_find(self, content: Dict[str, Any]) -> List[Content]:
         # Find an algorithm that is common to all the requested contents.
         # It will be used to do an initial filtering efficiently.
         filter_algos = list(set(content).intersection(HASH_ALGORITHMS))
         if not filter_algos:
             raise StorageArgumentException(
                 "content keys must contain at least one "
                 f"of: {', '.join(sorted(HASH_ALGORITHMS))}"
             )
         common_algo = filter_algos[0]
 
         results = []
         rows = self._content_get_from_hash(common_algo, content[common_algo])
         for row in rows:
             # Re-check all the hashes, in case of collisions (either of the
             # hash of the partition key, or the hashes in it)
             for algo in HASH_ALGORITHMS:
                 if content.get(algo) and getattr(row, algo) != content[algo]:
                     # This hash didn't match; discard the row.
                     break
             else:
                 # All hashes match, keep this row.
                 row_d = row.to_dict()
                 row_d["ctime"] = row.ctime.replace(tzinfo=datetime.timezone.utc)
                 results.append(Content(**row_d))
         return results
 
     def content_missing(
         self, contents: List[Dict[str, Any]], key_hash: str = "sha1"
     ) -> Iterable[bytes]:
         if key_hash not in DEFAULT_ALGORITHMS:
             raise StorageArgumentException(
                 "key_hash should be one of {','.join(DEFAULT_ALGORITHMS)}"
             )
 
         contents_with_all_hashes = []
         contents_with_missing_hashes = []
         for content in contents:
             if DEFAULT_ALGORITHMS <= set(content):
                 contents_with_all_hashes.append(content)
             else:
                 contents_with_missing_hashes.append(content)
 
         # These contents can be queried efficiently directly in the main table
         for content in self._cql_runner.content_missing_from_hashes(
             contents_with_all_hashes
         ):
             yield content[key_hash]
 
         # For these, we need the expensive index lookups + main table.
         for content in contents_with_missing_hashes:
             res = self.content_find(content)
             if not res:
                 yield content[key_hash]
 
     def content_missing_per_sha1(self, contents: List[bytes]) -> Iterable[bytes]:
         return self.content_missing([{"sha1": c} for c in contents])
 
     def content_missing_per_sha1_git(
         self, contents: List[Sha1Git]
     ) -> Iterable[Sha1Git]:
         return self.content_missing(
             [{"sha1_git": c} for c in contents], key_hash="sha1_git"
         )
 
     def content_get_random(self) -> Sha1Git:
         content = self._cql_runner.content_get_random()
         assert content, "Could not find any content"
         return content.sha1_git
 
     def _skipped_content_add(self, contents: List[SkippedContent]) -> Dict[str, int]:
         # Filter-out content already in the database.
         if not self._allow_overwrite:
             contents = [
                 c
                 for c in contents
                 if not self._cql_runner.skipped_content_get_from_pk(c.to_dict())
             ]
 
         self.journal_writer.skipped_content_add(contents)
 
         for content in contents:
             # Compute token of the row in the main table
             (token, insertion_finalizer) = self._cql_runner.skipped_content_add_prepare(
                 SkippedContentRow.from_dict({"origin": None, **content.to_dict()})
             )
 
             # Then add to index tables
             for algo in HASH_ALGORITHMS:
                 self._cql_runner.skipped_content_index_add_one(algo, content, token)
 
             # Then to the main table
             insertion_finalizer()
 
         return {"skipped_content:add": len(contents)}
 
     def skipped_content_add(self, content: List[SkippedContent]) -> Dict[str, int]:
         contents = [attr.evolve(c, ctime=now()) for c in content]
         return self._skipped_content_add(contents)
 
     def skipped_content_missing(
         self, contents: List[Dict[str, Any]]
     ) -> Iterable[Dict[str, Any]]:
         for content in contents:
             if not self._cql_runner.skipped_content_get_from_pk(content):
                 yield {algo: content[algo] for algo in DEFAULT_ALGORITHMS}
 
     def directory_add(self, directories: List[Directory]) -> Dict[str, int]:
         to_add = {d.id: d for d in directories}.values()
         if not self._allow_overwrite:
             # Filter out directories that are already inserted.
             missing = self.directory_missing([dir_.id for dir_ in to_add])
             directories = [dir_ for dir_ in directories if dir_.id in missing]
 
         self.journal_writer.directory_add(directories)
 
         for directory in directories:
             # Add directory entries to the 'directory_entry' table
             for entry in directory.entries:
                 self._cql_runner.directory_entry_add_one(
                     DirectoryEntryRow(directory_id=directory.id, **entry.to_dict())
                 )
 
             # Add the directory *after* adding all the entries, so someone
             # calling snapshot_get_branch in the meantime won't end up
             # with half the entries.
             self._cql_runner.directory_add_one(DirectoryRow(id=directory.id))
 
         return {"directory:add": len(directories)}
 
     def directory_missing(self, directories: List[Sha1Git]) -> Iterable[Sha1Git]:
         return self._cql_runner.directory_missing(directories)
 
     def _join_dentry_to_content(self, dentry: DirectoryEntry) -> Dict[str, Any]:
         contents: Union[List[Content], List[SkippedContentRow]]
         keys = (
             "status",
             "sha1",
             "sha1_git",
             "sha256",
             "length",
         )
         ret = dict.fromkeys(keys)
         ret.update(dentry.to_dict())
         if ret["type"] == "file":
             contents = self.content_find({"sha1_git": ret["target"]})
             if not contents:
                 tokens = list(
                     self._cql_runner.skipped_content_get_tokens_from_single_hash(
                         "sha1_git", ret["target"]
                     )
                 )
                 if tokens:
                     contents = list(
                         self._cql_runner.skipped_content_get_from_token(tokens[0])
                     )
             if contents:
                 content = contents[0]
                 for key in keys:
                     ret[key] = getattr(content, key)
         return ret
 
     def _directory_ls(
         self, directory_id: Sha1Git, recursive: bool, prefix: bytes = b""
     ) -> Iterable[Dict[str, Any]]:
         if self.directory_missing([directory_id]):
             return
         rows = list(self._cql_runner.directory_entry_get([directory_id]))
 
         for row in rows:
             entry_d = row.to_dict()
             # Build and yield the directory entry dict
             del entry_d["directory_id"]
             entry = DirectoryEntry.from_dict(entry_d)
             ret = self._join_dentry_to_content(entry)
             ret["name"] = prefix + ret["name"]
             ret["dir_id"] = directory_id
             yield ret
 
             if recursive and ret["type"] == "dir":
                 yield from self._directory_ls(
                     ret["target"], True, prefix + ret["name"] + b"/"
                 )
 
     def directory_entry_get_by_path(
         self, directory: Sha1Git, paths: List[bytes]
     ) -> Optional[Dict[str, Any]]:
         return self._directory_entry_get_by_path(directory, paths, b"")
 
     def _directory_entry_get_by_path(
         self, directory: Sha1Git, paths: List[bytes], prefix: bytes
     ) -> Optional[Dict[str, Any]]:
         if not paths:
             return None
 
         contents = list(self.directory_ls(directory))
 
         if not contents:
             return None
 
         def _get_entry(entries, name):
             """Finds the entry with the requested name, prepends the
             prefix (to get its full path), and returns it.
 
             If no entry has that name, returns None."""
             for entry in entries:
                 if entry["name"] == name:
                     entry = entry.copy()
                     entry["name"] = prefix + entry["name"]
                     return entry
 
         first_item = _get_entry(contents, paths[0])
 
         if len(paths) == 1:
             return first_item
 
         if not first_item or first_item["type"] != "dir":
             return None
 
         return self._directory_entry_get_by_path(
             first_item["target"], paths[1:], prefix + paths[0] + b"/"
         )
 
     def directory_ls(
         self, directory: Sha1Git, recursive: bool = False
     ) -> Iterable[Dict[str, Any]]:
         yield from self._directory_ls(directory, recursive)
 
     def directory_get_entries(
         self,
         directory_id: Sha1Git,
         page_token: Optional[bytes] = None,
         limit: int = 1000,
     ) -> Optional[PagedResult[DirectoryEntry]]:
         if self.directory_missing([directory_id]):
             return None
 
         entries_from: bytes = page_token or b""
         rows = self._cql_runner.directory_entry_get_from_name(
             directory_id, entries_from, limit + 1
         )
         entries = [
             DirectoryEntry.from_dict(remove_keys(row.to_dict(), ("directory_id",)))
             for row in rows
         ]
         if len(entries) > limit:
             last_entry = entries.pop()
             next_page_token = last_entry.name
         else:
             next_page_token = None
         return PagedResult(results=entries, next_page_token=next_page_token)
 
     def directory_get_random(self) -> Sha1Git:
         directory = self._cql_runner.directory_get_random()
         assert directory, "Could not find any directory"
         return directory.id
 
     def revision_add(self, revisions: List[Revision]) -> Dict[str, int]:
         # Filter-out revisions already in the database
         if not self._allow_overwrite:
             to_add = {r.id: r for r in revisions}.values()
             missing = self.revision_missing([rev.id for rev in to_add])
             revisions = [rev for rev in revisions if rev.id in missing]
         self.journal_writer.revision_add(revisions)
 
         for revision in revisions:
             revobject = converters.revision_to_db(revision)
             if revobject:
                 # Add parents first
                 for (rank, parent) in enumerate(revision.parents):
                     self._cql_runner.revision_parent_add_one(
                         RevisionParentRow(
                             id=revobject.id, parent_rank=rank, parent_id=parent
                         )
                     )
 
                 # Then write the main revision row.
                 # Writing this after all parents were written ensures that
                 # read endpoints don't return a partial view while writing
                 # the parents
                 self._cql_runner.revision_add_one(revobject)
 
         return {"revision:add": len(revisions)}
 
     def revision_missing(self, revisions: List[Sha1Git]) -> Iterable[Sha1Git]:
         return self._cql_runner.revision_missing(revisions)
 
     def revision_get(self, revision_ids: List[Sha1Git]) -> List[Optional[Revision]]:
         rows = self._cql_runner.revision_get(revision_ids)
         revisions: Dict[Sha1Git, Revision] = {}
         for row in rows:
             # TODO: use a single query to get all parents?
             # (it might have lower latency, but requires more code and more
             # bandwidth, because revision id would be part of each returned
             # row)
             parents = tuple(self._cql_runner.revision_parent_get(row.id))
             # parent_rank is the clustering key, so results are already
             # sorted by rank.
             rev = converters.revision_from_db(row, parents=parents)
             revisions[rev.id] = rev
 
         return [revisions.get(rev_id) for rev_id in revision_ids]
 
     def _get_parent_revs(
         self,
         rev_ids: Iterable[Sha1Git],
         seen: Set[Sha1Git],
         limit: Optional[int],
         short: bool,
     ) -> Union[
         Iterable[Dict[str, Any]], Iterable[Tuple[Sha1Git, Tuple[Sha1Git, ...]]],
     ]:
         if limit and len(seen) >= limit:
             return
         rev_ids = [id_ for id_ in rev_ids if id_ not in seen]
         if not rev_ids:
             return
         seen |= set(rev_ids)
 
         # We need this query, even if short=True, to return consistent
         # results (ie. not return only a subset of a revision's parents
         # if it is being written)
         if short:
             ids = self._cql_runner.revision_get_ids(rev_ids)
             for id_ in ids:
                 # TODO: use a single query to get all parents?
                 # (it might have less latency, but requires less code and more
                 # bandwidth (because revision id would be part of each returned
                 # row)
                 parents = tuple(self._cql_runner.revision_parent_get(id_))
 
                 # parent_rank is the clustering key, so results are already
                 # sorted by rank.
 
                 yield (id_, parents)
                 yield from self._get_parent_revs(parents, seen, limit, short)
         else:
             rows = self._cql_runner.revision_get(rev_ids)
 
             for row in rows:
                 # TODO: use a single query to get all parents?
                 # (it might have less latency, but requires less code and more
                 # bandwidth (because revision id would be part of each returned
                 # row)
                 parents = tuple(self._cql_runner.revision_parent_get(row.id))
 
                 # parent_rank is the clustering key, so results are already
                 # sorted by rank.
 
                 rev = converters.revision_from_db(row, parents=parents)
                 yield rev.to_dict()
                 yield from self._get_parent_revs(parents, seen, limit, short)
 
     def revision_log(
         self, revisions: List[Sha1Git], limit: Optional[int] = None
     ) -> Iterable[Optional[Dict[str, Any]]]:
         seen: Set[Sha1Git] = set()
         yield from self._get_parent_revs(revisions, seen, limit, False)
 
     def revision_shortlog(
         self, revisions: List[Sha1Git], limit: Optional[int] = None
     ) -> Iterable[Optional[Tuple[Sha1Git, Tuple[Sha1Git, ...]]]]:
         seen: Set[Sha1Git] = set()
         yield from self._get_parent_revs(revisions, seen, limit, True)
 
     def revision_get_random(self) -> Sha1Git:
         revision = self._cql_runner.revision_get_random()
         assert revision, "Could not find any revision"
         return revision.id
 
     def release_add(self, releases: List[Release]) -> Dict[str, int]:
         if not self._allow_overwrite:
             to_add = {r.id: r for r in releases}.values()
             missing = set(self.release_missing([rel.id for rel in to_add]))
             releases = [rel for rel in to_add if rel.id in missing]
         self.journal_writer.release_add(releases)
 
         for release in releases:
             if release:
                 self._cql_runner.release_add_one(converters.release_to_db(release))
 
         return {"release:add": len(releases)}
 
     def release_missing(self, releases: List[Sha1Git]) -> Iterable[Sha1Git]:
         return self._cql_runner.release_missing(releases)
 
     def release_get(self, releases: List[Sha1Git]) -> List[Optional[Release]]:
         rows = self._cql_runner.release_get(releases)
         rels: Dict[Sha1Git, Release] = {}
         for row in rows:
             release = converters.release_from_db(row)
             rels[row.id] = release
 
         return [rels.get(rel_id) for rel_id in releases]
 
     def release_get_random(self) -> Sha1Git:
         release = self._cql_runner.release_get_random()
         assert release, "Could not find any release"
         return release.id
 
     def snapshot_add(self, snapshots: List[Snapshot]) -> Dict[str, int]:
         if not self._allow_overwrite:
             to_add = {s.id: s for s in snapshots}.values()
             missing = self._cql_runner.snapshot_missing([snp.id for snp in to_add])
             snapshots = [snp for snp in snapshots if snp.id in missing]
 
         for snapshot in snapshots:
             self.journal_writer.snapshot_add([snapshot])
 
             # Add branches
             for (branch_name, branch) in snapshot.branches.items():
                 if branch is None:
                     target_type: Optional[str] = None
                     target: Optional[bytes] = None
                 else:
                     target_type = branch.target_type.value
                     target = branch.target
                 self._cql_runner.snapshot_branch_add_one(
                     SnapshotBranchRow(
                         snapshot_id=snapshot.id,
                         name=branch_name,
                         target_type=target_type,
                         target=target,
                     )
                 )
 
             # Add the snapshot *after* adding all the branches, so someone
             # calling snapshot_get_branch in the meantime won't end up
             # with half the branches.
             self._cql_runner.snapshot_add_one(SnapshotRow(id=snapshot.id))
 
         return {"snapshot:add": len(snapshots)}
 
     def snapshot_missing(self, snapshots: List[Sha1Git]) -> Iterable[Sha1Git]:
         return self._cql_runner.snapshot_missing(snapshots)
 
     def snapshot_get(self, snapshot_id: Sha1Git) -> Optional[Dict[str, Any]]:
         d = self.snapshot_get_branches(snapshot_id)
         if d is None:
             return None
         return {
             "id": d["id"],
             "branches": {
                 name: branch.to_dict() if branch else None
                 for (name, branch) in d["branches"].items()
             },
             "next_branch": d["next_branch"],
         }
 
     def snapshot_count_branches(
         self, snapshot_id: Sha1Git, branch_name_exclude_prefix: Optional[bytes] = None,
     ) -> Optional[Dict[Optional[str], int]]:
         if self._cql_runner.snapshot_missing([snapshot_id]):
             # Makes sure we don't fetch branches for a snapshot that is
             # being added.
             return None
 
         return self._cql_runner.snapshot_count_branches(
             snapshot_id, branch_name_exclude_prefix
         )
 
     def snapshot_get_branches(
         self,
         snapshot_id: Sha1Git,
         branches_from: bytes = b"",
         branches_count: int = 1000,
         target_types: Optional[List[str]] = None,
         branch_name_include_substring: Optional[bytes] = None,
         branch_name_exclude_prefix: Optional[bytes] = None,
     ) -> Optional[PartialBranches]:
         if self._cql_runner.snapshot_missing([snapshot_id]):
             # Makes sure we don't fetch branches for a snapshot that is
             # being added.
             return None
 
         branches: List = []
         while len(branches) < branches_count + 1:
             new_branches = list(
                 self._cql_runner.snapshot_branch_get(
                     snapshot_id,
                     branches_from,
                     branches_count + 1,
                     branch_name_exclude_prefix,
                 )
             )
 
             if not new_branches:
                 break
 
             branches_from = new_branches[-1].name
 
             new_branches_filtered = new_branches
 
             # Filter by target_type
             if target_types:
                 new_branches_filtered = [
                     branch
                     for branch in new_branches_filtered
                     if branch.target is not None and branch.target_type in target_types
                 ]
 
             # Filter by branches_name_pattern
             if branch_name_include_substring:
                 new_branches_filtered = [
                     branch
                     for branch in new_branches_filtered
                     if branch.name is not None
                     and (
                         branch_name_include_substring is None
                         or branch_name_include_substring in branch.name
                     )
                 ]
 
             branches.extend(new_branches_filtered)
 
             if len(new_branches) < branches_count + 1:
                 break
 
         if len(branches) > branches_count:
             last_branch = branches.pop(-1).name
         else:
             last_branch = None
 
         return PartialBranches(
             id=snapshot_id,
             branches={
                 branch.name: None
                 if branch.target is None
                 else SnapshotBranch(
                     target=branch.target, target_type=TargetType(branch.target_type)
                 )
                 for branch in branches
             },
             next_branch=last_branch,
         )
 
     def snapshot_get_random(self) -> Sha1Git:
         snapshot = self._cql_runner.snapshot_get_random()
         assert snapshot, "Could not find any snapshot"
         return snapshot.id
 
     def object_find_by_sha1_git(self, ids: List[Sha1Git]) -> Dict[Sha1Git, List[Dict]]:
         results: Dict[Sha1Git, List[Dict]] = {id_: [] for id_ in ids}
         missing_ids = set(ids)
 
         # Mind the order, revision is the most likely one for a given ID,
         # so we check revisions first.
         queries: List[Tuple[str, Callable[[List[Sha1Git]], List[Sha1Git]]]] = [
             ("revision", self._cql_runner.revision_missing),
             ("release", self._cql_runner.release_missing),
             ("content", self._cql_runner.content_missing_by_sha1_git),
             ("directory", self._cql_runner.directory_missing),
         ]
 
         for (object_type, query_fn) in queries:
             found_ids = missing_ids - set(query_fn(list(missing_ids)))
             for sha1_git in found_ids:
                 results[sha1_git].append(
                     {"sha1_git": sha1_git, "type": object_type,}
                 )
                 missing_ids.remove(sha1_git)
 
             if not missing_ids:
                 # We found everything, skipping the next queries.
                 break
 
         return results
 
     def origin_get(self, origins: List[str]) -> Iterable[Optional[Origin]]:
         return [self.origin_get_one(origin) for origin in origins]
 
     def origin_get_one(self, origin_url: str) -> Optional[Origin]:
         """Given an origin url, return the origin if it exists, None otherwise
 
         """
         rows = list(self._cql_runner.origin_get_by_url(origin_url))
         if rows:
             assert len(rows) == 1
             return Origin(url=rows[0].url)
         else:
             return None
 
     def origin_get_by_sha1(self, sha1s: List[bytes]) -> List[Optional[Dict[str, Any]]]:
         results = []
         for sha1 in sha1s:
             rows = list(self._cql_runner.origin_get_by_sha1(sha1))
             origin = {"url": rows[0].url} if rows else None
             results.append(origin)
         return results
 
     def origin_list(
         self, page_token: Optional[str] = None, limit: int = 100
     ) -> PagedResult[Origin]:
         # Compute what token to begin the listing from
         start_token = TOKEN_BEGIN
         if page_token:
             start_token = int(page_token)
             if not (TOKEN_BEGIN <= start_token <= TOKEN_END):
                 raise StorageArgumentException("Invalid page_token.")
         next_page_token = None
 
         origins = []
         # Take one more origin so we can reuse it as the next page token if any
         for (tok, row) in self._cql_runner.origin_list(start_token, limit + 1):
             origins.append(Origin(url=row.url))
             # keep reference of the last id for pagination purposes
             last_id = tok
 
         if len(origins) > limit:
             # last origin id is the next page token
             next_page_token = str(last_id)
             # excluding that origin from the result to respect the limit size
             origins = origins[:limit]
 
         assert len(origins) <= limit
 
         return PagedResult(results=origins, next_page_token=next_page_token)
 
     def origin_search(
         self,
         url_pattern: str,
         page_token: Optional[str] = None,
         limit: int = 50,
         regexp: bool = False,
         with_visit: bool = False,
         visit_types: Optional[List[str]] = None,
     ) -> PagedResult[Origin]:
         # TODO: remove this endpoint, swh-search should be used instead.
         next_page_token = None
         offset = int(page_token) if page_token else 0
 
         origin_rows = [row for row in self._cql_runner.origin_iter_all()]
         if regexp:
             pat = re.compile(url_pattern)
             origin_rows = [row for row in origin_rows if pat.search(row.url)]
         else:
             origin_rows = [row for row in origin_rows if url_pattern in row.url]
 
         if with_visit:
             origin_rows = [row for row in origin_rows if row.next_visit_id > 1]
 
         if visit_types:
 
             def _has_visit_types(origin, visit_types):
                 for origin_visit in stream_results(self.origin_visit_get, origin):
                     if origin_visit.type in visit_types:
                         return True
                 return False
 
             origin_rows = [
                 row for row in origin_rows if _has_visit_types(row.url, visit_types)
             ]
 
         origins = [Origin(url=row.url) for row in origin_rows]
 
         origins = origins[offset : offset + limit + 1]
         if len(origins) > limit:
             # next offset
             next_page_token = str(offset + limit)
             # excluding that origin from the result to respect the limit size
             origins = origins[:limit]
 
         assert len(origins) <= limit
         return PagedResult(results=origins, next_page_token=next_page_token)
 
     def origin_count(
         self, url_pattern: str, regexp: bool = False, with_visit: bool = False
     ) -> int:
         raise NotImplementedError(
             "The Cassandra backend does not implement origin_count"
         )
 
     def origin_add(self, origins: List[Origin]) -> Dict[str, int]:
         if not self._allow_overwrite:
             to_add = {o.url: o for o in origins}.values()
             origins = [ori for ori in to_add if self.origin_get_one(ori.url) is None]
 
         self.journal_writer.origin_add(origins)
         for origin in origins:
             self._cql_runner.origin_add_one(
                 OriginRow(sha1=hash_url(origin.url), url=origin.url, next_visit_id=1)
             )
         return {"origin:add": len(origins)}
 
     def origin_visit_add(self, visits: List[OriginVisit]) -> Iterable[OriginVisit]:
         for visit in visits:
             origin = self.origin_get_one(visit.origin)
             if not origin:  # Cannot add a visit without an origin
                 raise StorageArgumentException("Unknown origin %s", visit.origin)
 
         all_visits = []
         nb_visits = 0
         for visit in visits:
             nb_visits += 1
-            if not visit.visit:
+            if visit.visit:
+                # Set origin.next_visit_id = max(origin.next_visit_id, visit.visit+1)
+                # so the next loader run does not reuse the id.
+                self._cql_runner.origin_bump_next_visit_id(visit.origin, visit.visit)
+            else:
                 visit_id = self._cql_runner.origin_generate_unique_visit_id(
                     visit.origin
                 )
                 visit = attr.evolve(visit, visit=visit_id)
             self.journal_writer.origin_visit_add([visit])
             self._cql_runner.origin_visit_add_one(OriginVisitRow(**visit.to_dict()))
             assert visit.visit is not None
             all_visits.append(visit)
             self._origin_visit_status_add(
                 OriginVisitStatus(
                     origin=visit.origin,
                     visit=visit.visit,
                     date=visit.date,
                     type=visit.type,
                     status="created",
                     snapshot=None,
                 )
             )
 
         return all_visits
 
     def _origin_visit_status_add(self, visit_status: OriginVisitStatus) -> None:
         """Add an origin visit status"""
         if visit_status.type is None:
             visit_row = self._cql_runner.origin_visit_get_one(
                 visit_status.origin, visit_status.visit
             )
             if visit_row is None:
                 raise StorageArgumentException(
                     f"Unknown origin visit {visit_status.visit} "
                     f"of origin {visit_status.origin}"
                 )
             visit_status = attr.evolve(visit_status, type=visit_row.type)
 
         self.journal_writer.origin_visit_status_add([visit_status])
         self._cql_runner.origin_visit_status_add_one(
             converters.visit_status_to_row(visit_status)
         )
 
     def origin_visit_status_add(
         self, visit_statuses: List[OriginVisitStatus]
     ) -> Dict[str, int]:
         # First round to check existence (fail early if any is ko)
         for visit_status in visit_statuses:
             origin_url = self.origin_get_one(visit_status.origin)
             if not origin_url:
                 raise StorageArgumentException(f"Unknown origin {visit_status.origin}")
 
         for visit_status in visit_statuses:
             self._origin_visit_status_add(visit_status)
         return {"origin_visit_status:add": len(visit_statuses)}
 
     def _origin_visit_apply_status(
         self, visit: Dict[str, Any], visit_status: OriginVisitStatusRow
     ) -> Dict[str, Any]:
         """Retrieve the latest visit status information for the origin visit.
         Then merge it with the visit and return it.
 
         """
         return {
             # default to the values in visit
             **visit,
             # override with the last update
             **visit_status.to_dict(),
             # visit['origin'] is the URL (via a join), while
             # visit_status['origin'] is only an id.
             "origin": visit["origin"],
             # but keep the date of the creation of the origin visit
             "date": visit["date"],
             # We use the visit type from origin visit
             # if it's not present on the origin visit status
             "type": visit_status.type or visit["type"],
         }
 
     def _origin_visit_get_latest_status(self, visit: OriginVisit) -> OriginVisitStatus:
         """Retrieve the latest visit status information for the origin visit object.
 
         """
         assert visit.visit
         row = self._cql_runner.origin_visit_status_get_latest(visit.origin, visit.visit)
         assert row is not None
         visit_status = converters.row_to_visit_status(row)
         return attr.evolve(visit_status, origin=visit.origin)
 
     @staticmethod
     def _format_origin_visit_row(visit):
         return {
             **visit.to_dict(),
             "origin": visit.origin,
             "date": visit.date.replace(tzinfo=datetime.timezone.utc),
         }
 
     def origin_visit_get(
         self,
         origin: str,
         page_token: Optional[str] = None,
         order: ListOrder = ListOrder.ASC,
         limit: int = 10,
     ) -> PagedResult[OriginVisit]:
         if not isinstance(order, ListOrder):
             raise StorageArgumentException("order must be a ListOrder value")
         if page_token and not isinstance(page_token, str):
             raise StorageArgumentException("page_token must be a string.")
 
         next_page_token = None
         visit_from = None if page_token is None else int(page_token)
         visits: List[OriginVisit] = []
         extra_limit = limit + 1
 
         rows = self._cql_runner.origin_visit_get(origin, visit_from, extra_limit, order)
         for row in rows:
             visits.append(converters.row_to_visit(row))
 
         assert len(visits) <= extra_limit
         if len(visits) == extra_limit:
             visits = visits[:limit]
             next_page_token = str(visits[-1].visit)
 
         return PagedResult(results=visits, next_page_token=next_page_token)
 
     def origin_visit_status_get(
         self,
         origin: str,
         visit: int,
         page_token: Optional[str] = None,
         order: ListOrder = ListOrder.ASC,
         limit: int = 10,
     ) -> PagedResult[OriginVisitStatus]:
         next_page_token = None
         date_from = None
         if page_token is not None:
             date_from = datetime.datetime.fromisoformat(page_token)
 
         # Take one more visit status so we can reuse it as the next page token if any
         rows = self._cql_runner.origin_visit_status_get_range(
             origin, visit, date_from, limit + 1, order
         )
         visit_statuses = [converters.row_to_visit_status(row) for row in rows]
         if len(visit_statuses) > limit:
             # last visit status date is the next page token
             next_page_token = str(visit_statuses[-1].date)
             # excluding that visit status from the result to respect the limit size
             visit_statuses = visit_statuses[:limit]
 
         return PagedResult(results=visit_statuses, next_page_token=next_page_token)
 
     def origin_visit_find_by_date(
         self, origin: str, visit_date: datetime.datetime
     ) -> Optional[OriginVisit]:
         # Iterator over all the visits of the origin
         # This should be ok for now, as there aren't too many visits
         # per origin.
         rows = list(self._cql_runner.origin_visit_get_all(origin))
 
         def key(visit):
             dt = visit.date.replace(tzinfo=datetime.timezone.utc) - visit_date
             return (abs(dt), -visit.visit)
 
         if rows:
             return converters.row_to_visit(min(rows, key=key))
         return None
 
     def origin_visit_get_by(self, origin: str, visit: int) -> Optional[OriginVisit]:
         row = self._cql_runner.origin_visit_get_one(origin, visit)
         if row:
             return converters.row_to_visit(row)
         return None
 
     def origin_visit_get_latest(
         self,
         origin: str,
         type: Optional[str] = None,
         allowed_statuses: Optional[List[str]] = None,
         require_snapshot: bool = False,
     ) -> Optional[OriginVisit]:
         if allowed_statuses and not set(allowed_statuses).intersection(VISIT_STATUSES):
             raise StorageArgumentException(
                 f"Unknown allowed statuses {','.join(allowed_statuses)}, only "
                 f"{','.join(VISIT_STATUSES)} authorized"
             )
         # TODO: Do not fetch all visits
         rows = self._cql_runner.origin_visit_get_all(origin)
         latest_visit = None
         for row in rows:
             visit = self._format_origin_visit_row(row)
             for status_row in self._cql_runner.origin_visit_status_get(
                 origin, visit["visit"]
             ):
                 updated_visit = self._origin_visit_apply_status(visit, status_row)
                 if type is not None and updated_visit["type"] != type:
                     continue
                 if allowed_statuses and updated_visit["status"] not in allowed_statuses:
                     continue
                 if require_snapshot and updated_visit["snapshot"] is None:
                     continue
 
                 # updated_visit is a candidate
                 if latest_visit is not None:
                     if updated_visit["date"] < latest_visit["date"]:
                         continue
                     if updated_visit["visit"] < latest_visit["visit"]:
                         continue
 
                 latest_visit = updated_visit
 
         if latest_visit is None:
             return None
         return OriginVisit(
             origin=latest_visit["origin"],
             visit=latest_visit["visit"],
             date=latest_visit["date"],
             type=latest_visit["type"],
         )
 
     def origin_visit_status_get_latest(
         self,
         origin_url: str,
         visit: int,
         allowed_statuses: Optional[List[str]] = None,
         require_snapshot: bool = False,
     ) -> Optional[OriginVisitStatus]:
         if allowed_statuses and not set(allowed_statuses).intersection(VISIT_STATUSES):
             raise StorageArgumentException(
                 f"Unknown allowed statuses {','.join(allowed_statuses)}, only "
                 f"{','.join(VISIT_STATUSES)} authorized"
             )
         rows = list(self._cql_runner.origin_visit_status_get(origin_url, visit))
         # filtering is done python side as we cannot do it server side
         if allowed_statuses:
             rows = [row for row in rows if row.status in allowed_statuses]
         if require_snapshot:
             rows = [row for row in rows if row.snapshot is not None]
         if not rows:
             return None
         return converters.row_to_visit_status(rows[0])
 
     def origin_visit_status_get_random(self, type: str) -> Optional[OriginVisitStatus]:
         back_in_the_day = now() - datetime.timedelta(weeks=12)  # 3 months back
 
         # Random position to start iteration at
         start_token = random.randint(TOKEN_BEGIN, TOKEN_END)
 
         # Iterator over all visits, ordered by token(origins) then visit_id
         rows = self._cql_runner.origin_visit_iter(start_token)
         for row in rows:
             visit = converters.row_to_visit(row)
             visit_status = self._origin_visit_get_latest_status(visit)
             if visit.date > back_in_the_day and visit_status.status == "full":
                 return visit_status
         return None
 
     def stat_counters(self):
         rows = self._cql_runner.stat_counters()
         keys = (
             "content",
             "directory",
             "origin",
             "origin_visit",
             "release",
             "revision",
             "skipped_content",
             "snapshot",
         )
         stats = {key: 0 for key in keys}
         stats.update({row.object_type: row.count for row in rows})
         return stats
 
     def refresh_stat_counters(self):
         pass
 
     def raw_extrinsic_metadata_add(
         self, metadata: List[RawExtrinsicMetadata]
     ) -> Dict[str, int]:
         self.journal_writer.raw_extrinsic_metadata_add(metadata)
         counter = Counter[ExtendedObjectType]()
         for metadata_entry in metadata:
             if not self._cql_runner.metadata_authority_get(
                 metadata_entry.authority.type.value, metadata_entry.authority.url
             ):
                 raise StorageArgumentException(
                     f"Unknown authority {metadata_entry.authority}"
                 )
             if not self._cql_runner.metadata_fetcher_get(
                 metadata_entry.fetcher.name, metadata_entry.fetcher.version
             ):
                 raise StorageArgumentException(
                     f"Unknown fetcher {metadata_entry.fetcher}"
                 )
 
             try:
                 row = RawExtrinsicMetadataRow(
                     id=metadata_entry.id,
                     type=metadata_entry.target.object_type.name.lower(),
                     target=str(metadata_entry.target),
                     authority_type=metadata_entry.authority.type.value,
                     authority_url=metadata_entry.authority.url,
                     discovery_date=metadata_entry.discovery_date,
                     fetcher_name=metadata_entry.fetcher.name,
                     fetcher_version=metadata_entry.fetcher.version,
                     format=metadata_entry.format,
                     metadata=metadata_entry.metadata,
                     origin=metadata_entry.origin,
                     visit=metadata_entry.visit,
                     snapshot=map_optional(str, metadata_entry.snapshot),
                     release=map_optional(str, metadata_entry.release),
                     revision=map_optional(str, metadata_entry.revision),
                     path=metadata_entry.path,
                     directory=map_optional(str, metadata_entry.directory),
                 )
             except TypeError as e:
                 raise StorageArgumentException(*e.args)
 
             # Add to the index first
             self._cql_runner.raw_extrinsic_metadata_by_id_add(
                 RawExtrinsicMetadataByIdRow(
                     id=row.id,
                     target=row.target,
                     authority_type=row.authority_type,
                     authority_url=row.authority_url,
                 )
             )
 
             # Then to the main table
             self._cql_runner.raw_extrinsic_metadata_add(row)
             counter[metadata_entry.target.object_type] += 1
         return {
             f"{type.value}_metadata:add": count for (type, count) in counter.items()
         }
 
     def raw_extrinsic_metadata_get(
         self,
         target: ExtendedSWHID,
         authority: MetadataAuthority,
         after: Optional[datetime.datetime] = None,
         page_token: Optional[bytes] = None,
         limit: int = 1000,
     ) -> PagedResult[RawExtrinsicMetadata]:
         if page_token is not None:
             (after_date, id_) = msgpack_loads(base64.b64decode(page_token))
             if after and after_date < after:
                 raise StorageArgumentException(
                     "page_token is inconsistent with the value of 'after'."
                 )
             entries = self._cql_runner.raw_extrinsic_metadata_get_after_date_and_id(
                 str(target), authority.type.value, authority.url, after_date, id_,
             )
         elif after is not None:
             entries = self._cql_runner.raw_extrinsic_metadata_get_after_date(
                 str(target), authority.type.value, authority.url, after
             )
         else:
             entries = self._cql_runner.raw_extrinsic_metadata_get(
                 str(target), authority.type.value, authority.url
             )
 
         if limit:
             entries = itertools.islice(entries, 0, limit + 1)
 
         results = []
         for entry in entries:
             assert str(target) == entry.target
 
             results.append(converters.row_to_raw_extrinsic_metadata(entry))
 
         if len(results) > limit:
             results.pop()
             assert len(results) == limit
             last_result = results[-1]
             next_page_token: Optional[str] = base64.b64encode(
                 msgpack_dumps((last_result.discovery_date, last_result.id,))
             ).decode()
         else:
             next_page_token = None
 
         return PagedResult(next_page_token=next_page_token, results=results,)
 
     def raw_extrinsic_metadata_get_by_ids(
         self, ids: List[Sha1Git]
     ) -> List[RawExtrinsicMetadata]:
         keys = self._cql_runner.raw_extrinsic_metadata_get_by_ids(ids)
 
         results: Set[RawExtrinsicMetadata] = set()
         for key in keys:
             candidates = self._cql_runner.raw_extrinsic_metadata_get(
                 key.target, key.authority_type, key.authority_url
             )
             candidates = [
                 candidate for candidate in candidates if candidate.id == key.id
             ]
             if len(candidates) > 1:
                 raise Exception(
                     "Found multiple RawExtrinsicMetadata objects with the same id: "
                     + hash_to_hex(key.id)
                 )
             results.update(map(converters.row_to_raw_extrinsic_metadata, candidates))
 
         return list(results)
 
     def raw_extrinsic_metadata_get_authorities(
         self, target: ExtendedSWHID
     ) -> List[MetadataAuthority]:
         return [
             MetadataAuthority(
                 type=MetadataAuthorityType(authority_type), url=authority_url
             )
             for (authority_type, authority_url) in set(
                 self._cql_runner.raw_extrinsic_metadata_get_authorities(str(target))
             )
         ]
 
     def metadata_fetcher_add(self, fetchers: List[MetadataFetcher]) -> Dict[str, int]:
         self.journal_writer.metadata_fetcher_add(fetchers)
         for fetcher in fetchers:
             self._cql_runner.metadata_fetcher_add(
                 MetadataFetcherRow(name=fetcher.name, version=fetcher.version,)
             )
         return {"metadata_fetcher:add": len(fetchers)}
 
     def metadata_fetcher_get(
         self, name: str, version: str
     ) -> Optional[MetadataFetcher]:
         fetcher = self._cql_runner.metadata_fetcher_get(name, version)
         if fetcher:
             return MetadataFetcher(name=fetcher.name, version=fetcher.version,)
         else:
             return None
 
     def metadata_authority_add(
         self, authorities: List[MetadataAuthority]
     ) -> Dict[str, int]:
         self.journal_writer.metadata_authority_add(authorities)
         for authority in authorities:
             self._cql_runner.metadata_authority_add(
                 MetadataAuthorityRow(url=authority.url, type=authority.type.value,)
             )
         return {"metadata_authority:add": len(authorities)}
 
     def metadata_authority_get(
         self, type: MetadataAuthorityType, url: str
     ) -> Optional[MetadataAuthority]:
         authority = self._cql_runner.metadata_authority_get(type.value, url)
         if authority:
             return MetadataAuthority(
                 type=MetadataAuthorityType(authority.type), url=authority.url,
             )
         else:
             return None
 
     # ExtID tables
     def extid_add(self, ids: List[ExtID]) -> Dict[str, int]:
         if not self._allow_overwrite:
             extids = [
                 extid
                 for extid in ids
                 if not self._cql_runner.extid_get_from_pk(
                     extid_type=extid.extid_type,
                     extid_version=extid.extid_version,
                     extid=extid.extid,
                     target=extid.target,
                 )
             ]
         else:
             extids = list(ids)
 
         self.journal_writer.extid_add(extids)
 
         inserted = 0
         for extid in extids:
             target_type = extid.target.object_type.value
             target = extid.target.object_id
             extidrow = ExtIDRow(
                 extid_type=extid.extid_type,
                 extid_version=extid.extid_version,
                 extid=extid.extid,
                 target_type=target_type,
                 target=target,
             )
             (token, insertion_finalizer) = self._cql_runner.extid_add_prepare(extidrow)
             indexrow = ExtIDByTargetRow(
                 target_type=target_type, target=target, target_token=token,
             )
             self._cql_runner.extid_index_add_one(indexrow)
             insertion_finalizer()
             inserted += 1
         return {"extid:add": inserted}
 
     def extid_get_from_extid(self, id_type: str, ids: List[bytes]) -> List[ExtID]:
         result: List[ExtID] = []
         for extid in ids:
             extidrows = list(self._cql_runner.extid_get_from_extid(id_type, extid))
             result.extend(
                 ExtID(
                     extid_type=extidrow.extid_type,
                     extid_version=extidrow.extid_version,
                     extid=extidrow.extid,
                     target=CoreSWHID(
                         object_type=extidrow.target_type, object_id=extidrow.target,
                     ),
                 )
                 for extidrow in extidrows
             )
         return result
 
     def extid_get_from_target(
         self, target_type: SwhidObjectType, ids: List[Sha1Git]
     ) -> List[ExtID]:
         result: List[ExtID] = []
         for target in ids:
             extidrows = list(
                 self._cql_runner.extid_get_from_target(target_type.value, target)
             )
             result.extend(
                 ExtID(
                     extid_type=extidrow.extid_type,
                     extid_version=extidrow.extid_version,
                     extid=extidrow.extid,
                     target=CoreSWHID(
                         object_type=SwhidObjectType(extidrow.target_type),
                         object_id=extidrow.target,
                     ),
                 )
                 for extidrow in extidrows
             )
         return result
 
     # Misc
     def clear_buffers(self, object_types: Sequence[str] = ()) -> None:
         """Do nothing
 
         """
         return None
 
     def flush(self, object_types: Sequence[str] = ()) -> Dict[str, int]:
         return {}
diff --git a/swh/storage/in_memory.py b/swh/storage/in_memory.py
index b9b3734a..951f2e47 100644
--- a/swh/storage/in_memory.py
+++ b/swh/storage/in_memory.py
@@ -1,734 +1,738 @@
 # Copyright (C) 2015-2020  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 from collections import defaultdict
 import datetime
 import functools
 import itertools
 import random
 from typing import (
     Any,
     Dict,
     Generic,
     Iterable,
     Iterator,
     List,
     Optional,
     Tuple,
     Type,
     TypeVar,
     Union,
 )
 
 from swh.model.identifiers import ExtendedSWHID
 from swh.model.model import Content, Sha1Git, SkippedContent
 from swh.storage.cassandra import CassandraStorage
 from swh.storage.cassandra.model import (
     BaseRow,
     ContentRow,
     DirectoryEntryRow,
     DirectoryRow,
     ExtIDByTargetRow,
     ExtIDRow,
     MetadataAuthorityRow,
     MetadataFetcherRow,
     ObjectCountRow,
     OriginRow,
     OriginVisitRow,
     OriginVisitStatusRow,
     RawExtrinsicMetadataByIdRow,
     RawExtrinsicMetadataRow,
     ReleaseRow,
     RevisionParentRow,
     RevisionRow,
     SkippedContentRow,
     SnapshotBranchRow,
     SnapshotRow,
 )
 from swh.storage.interface import ListOrder
 from swh.storage.objstorage import ObjStorage
 
 from .common import origin_url_to_sha1
 from .writer import JournalWriter
 
 TRow = TypeVar("TRow", bound=BaseRow)
 
 
 class Table(Generic[TRow]):
     def __init__(self, row_class: Type[TRow]):
         self.row_class = row_class
         self.primary_key_cols = row_class.PARTITION_KEY + row_class.CLUSTERING_KEY
 
         # Map from tokens to clustering keys to rows
         # These are not actually partitions (or rather, there is one partition
         # for each token) and they aren't sorted.
         # But it is good enough if we don't care about performance;
         # and makes the code a lot simpler.
         self.data: Dict[int, Dict[Tuple, TRow]] = defaultdict(dict)
 
     def __repr__(self):
         return f"<__module__.Table[{self.row_class.__name__}] object>"
 
     def partition_key(self, row: Union[TRow, Dict[str, Any]]) -> Tuple:
         """Returns the partition key of a row (ie. the cells which get hashed
         into the token."""
         if isinstance(row, dict):
             row_d = row
         else:
             row_d = row.to_dict()
         return tuple(row_d[col] for col in self.row_class.PARTITION_KEY)
 
     def clustering_key(self, row: Union[TRow, Dict[str, Any]]) -> Tuple:
         """Returns the clustering key of a row (ie. the cells which are used
         for sorting rows within a partition."""
         if isinstance(row, dict):
             row_d = row
         else:
             row_d = row.to_dict()
         return tuple(row_d[col] for col in self.row_class.CLUSTERING_KEY)
 
     def primary_key(self, row):
         return self.partition_key(row) + self.clustering_key(row)
 
     def primary_key_from_dict(self, d: Dict[str, Any]) -> Tuple:
         """Returns the primary key (ie. concatenation of partition key and
         clustering key) of the given dictionary interpreted as a row."""
         return tuple(d[col] for col in self.primary_key_cols)
 
     def token(self, key: Tuple):
         """Returns the token of a row (ie. the hash of its partition key)."""
         return hash(key)
 
     def get_partition(self, token: int) -> Dict[Tuple, TRow]:
         """Returns the partition that contains this token."""
         return self.data[token]
 
     def insert(self, row: TRow):
         partition = self.data[self.token(self.partition_key(row))]
         partition[self.clustering_key(row)] = row
 
     def split_primary_key(self, key: Tuple) -> Tuple[Tuple, Tuple]:
         """Returns (partition_key, clustering_key) from a partition key"""
         assert len(key) == len(self.primary_key_cols)
 
         partition_key = key[0 : len(self.row_class.PARTITION_KEY)]
         clustering_key = key[len(self.row_class.PARTITION_KEY) :]
 
         return (partition_key, clustering_key)
 
     def get_from_partition_key(self, partition_key: Tuple) -> Iterable[TRow]:
         """Returns at most one row, from its partition key."""
         token = self.token(partition_key)
         for row in self.get_from_token(token):
             if self.partition_key(row) == partition_key:
                 yield row
 
     def get_from_primary_key(self, primary_key: Tuple) -> Optional[TRow]:
         """Returns at most one row, from its primary key."""
         (partition_key, clustering_key) = self.split_primary_key(primary_key)
 
         token = self.token(partition_key)
         partition = self.get_partition(token)
 
         return partition.get(clustering_key)
 
     def get_from_token(self, token: int) -> Iterable[TRow]:
         """Returns all rows whose token (ie. non-cryptographic hash of the
         partition key) is the one passed as argument."""
         return (v for (k, v) in sorted(self.get_partition(token).items()))
 
     def iter_all(self) -> Iterator[Tuple[Tuple, TRow]]:
         return (
             (self.primary_key(row), row)
             for (token, partition) in self.data.items()
             for (clustering_key, row) in partition.items()
         )
 
     def get_random(self) -> Optional[TRow]:
         return random.choice([row for (pk, row) in self.iter_all()])
 
 
 class InMemoryCqlRunner:
     def __init__(self):
         self._contents = Table(ContentRow)
         self._content_indexes = defaultdict(lambda: defaultdict(set))
         self._skipped_contents = Table(ContentRow)
         self._skipped_content_indexes = defaultdict(lambda: defaultdict(set))
         self._directories = Table(DirectoryRow)
         self._directory_entries = Table(DirectoryEntryRow)
         self._revisions = Table(RevisionRow)
         self._revision_parents = Table(RevisionParentRow)
         self._releases = Table(ReleaseRow)
         self._snapshots = Table(SnapshotRow)
         self._snapshot_branches = Table(SnapshotBranchRow)
         self._origins = Table(OriginRow)
         self._origin_visits = Table(OriginVisitRow)
         self._origin_visit_statuses = Table(OriginVisitStatusRow)
         self._metadata_authorities = Table(MetadataAuthorityRow)
         self._metadata_fetchers = Table(MetadataFetcherRow)
         self._raw_extrinsic_metadata = Table(RawExtrinsicMetadataRow)
         self._raw_extrinsic_metadata_by_id = Table(RawExtrinsicMetadataByIdRow)
         self._extid = Table(ExtIDRow)
         self._stat_counters = defaultdict(int)
 
     def increment_counter(self, object_type: str, nb: int):
         self._stat_counters[object_type] += nb
 
     def stat_counters(self) -> Iterable[ObjectCountRow]:
         for (object_type, count) in self._stat_counters.items():
             yield ObjectCountRow(partition_key=0, object_type=object_type, count=count)
 
     ##########################
     # 'content' table
     ##########################
 
     def _content_add_finalize(self, content: ContentRow) -> None:
         self._contents.insert(content)
         self.increment_counter("content", 1)
 
     def content_add_prepare(self, content: ContentRow):
         finalizer = functools.partial(self._content_add_finalize, content)
         return (self._contents.token(self._contents.partition_key(content)), finalizer)
 
     def content_get_from_pk(
         self, content_hashes: Dict[str, bytes]
     ) -> Optional[ContentRow]:
         primary_key = self._contents.primary_key_from_dict(content_hashes)
         return self._contents.get_from_primary_key(primary_key)
 
     def content_get_from_token(self, token: int) -> Iterable[ContentRow]:
         return self._contents.get_from_token(token)
 
     def content_get_random(self) -> Optional[ContentRow]:
         return self._contents.get_random()
 
     def content_get_token_range(
         self, start: int, end: int, limit: int,
     ) -> Iterable[Tuple[int, ContentRow]]:
         matches = [
             (token, row)
             for (token, partition) in self._contents.data.items()
             for (clustering_key, row) in partition.items()
             if start <= token <= end
         ]
         matches.sort()
         return matches[0:limit]
 
     def content_missing_from_hashes(
         self, contents_hashes: List[Dict[str, bytes]]
     ) -> Iterator[Dict[str, bytes]]:
         for content_hashes in contents_hashes:
             if not self.content_get_from_pk(content_hashes):
                 yield content_hashes
 
     ##########################
     # 'content_by_*' tables
     ##########################
 
     def content_missing_by_sha1_git(self, ids: List[bytes]) -> List[bytes]:
         missing = []
         for id_ in ids:
             if id_ not in self._content_indexes["sha1_git"]:
                 missing.append(id_)
         return missing
 
     def content_index_add_one(self, algo: str, content: Content, token: int) -> None:
         self._content_indexes[algo][content.get_hash(algo)].add(token)
 
     def content_get_tokens_from_single_hash(
         self, algo: str, hash_: bytes
     ) -> Iterable[int]:
         return self._content_indexes[algo][hash_]
 
     ##########################
     # 'skipped_content' table
     ##########################
 
     def _skipped_content_add_finalize(self, content: SkippedContentRow) -> None:
         self._skipped_contents.insert(content)
         self.increment_counter("skipped_content", 1)
 
     def skipped_content_add_prepare(self, content: SkippedContentRow):
         finalizer = functools.partial(self._skipped_content_add_finalize, content)
         return (
             self._skipped_contents.token(self._contents.partition_key(content)),
             finalizer,
         )
 
     def skipped_content_get_from_pk(
         self, content_hashes: Dict[str, bytes]
     ) -> Optional[SkippedContentRow]:
         primary_key = self._skipped_contents.primary_key_from_dict(content_hashes)
         return self._skipped_contents.get_from_primary_key(primary_key)
 
     def skipped_content_get_from_token(self, token: int) -> Iterable[SkippedContentRow]:
         return self._skipped_contents.get_from_token(token)
 
     ##########################
     # 'skipped_content_by_*' tables
     ##########################
 
     def skipped_content_index_add_one(
         self, algo: str, content: SkippedContent, token: int
     ) -> None:
         self._skipped_content_indexes[algo][content.get_hash(algo)].add(token)
 
     def skipped_content_get_tokens_from_single_hash(
         self, algo: str, hash_: bytes
     ) -> Iterable[int]:
         return self._skipped_content_indexes[algo][hash_]
 
     ##########################
     # 'directory' table
     ##########################
 
     def directory_missing(self, ids: List[bytes]) -> List[bytes]:
         missing = []
         for id_ in ids:
             if self._directories.get_from_primary_key((id_,)) is None:
                 missing.append(id_)
         return missing
 
     def directory_add_one(self, directory: DirectoryRow) -> None:
         self._directories.insert(directory)
         self.increment_counter("directory", 1)
 
     def directory_get_random(self) -> Optional[DirectoryRow]:
         return self._directories.get_random()
 
     ##########################
     # 'directory_entry' table
     ##########################
 
     def directory_entry_add_one(self, entry: DirectoryEntryRow) -> None:
         self._directory_entries.insert(entry)
 
     def directory_entry_get(
         self, directory_ids: List[Sha1Git]
     ) -> Iterable[DirectoryEntryRow]:
         for id_ in directory_ids:
             yield from self._directory_entries.get_from_partition_key((id_,))
 
     def directory_entry_get_from_name(
         self, directory_id: Sha1Git, from_: bytes, limit: int
     ) -> Iterable[DirectoryEntryRow]:
         # Get all entries
         entries = self._directory_entries.get_from_partition_key((directory_id,))
         # Filter out the ones before from_
         entries = itertools.dropwhile(lambda entry: entry.name < from_, entries)
         # Apply limit
         return itertools.islice(entries, limit)
 
     ##########################
     # 'revision' table
     ##########################
 
     def revision_missing(self, ids: List[bytes]) -> Iterable[bytes]:
         missing = []
         for id_ in ids:
             if self._revisions.get_from_primary_key((id_,)) is None:
                 missing.append(id_)
         return missing
 
     def revision_add_one(self, revision: RevisionRow) -> None:
         self._revisions.insert(revision)
         self.increment_counter("revision", 1)
 
     def revision_get_ids(self, revision_ids) -> Iterable[int]:
         for id_ in revision_ids:
             if self._revisions.get_from_primary_key((id_,)) is not None:
                 yield id_
 
     def revision_get(self, revision_ids: List[Sha1Git]) -> Iterable[RevisionRow]:
         for id_ in revision_ids:
             row = self._revisions.get_from_primary_key((id_,))
             if row:
                 yield row
 
     def revision_get_random(self) -> Optional[RevisionRow]:
         return self._revisions.get_random()
 
     ##########################
     # 'revision_parent' table
     ##########################
 
     def revision_parent_add_one(self, revision_parent: RevisionParentRow) -> None:
         self._revision_parents.insert(revision_parent)
 
     def revision_parent_get(self, revision_id: Sha1Git) -> Iterable[bytes]:
         for parent in self._revision_parents.get_from_partition_key((revision_id,)):
             yield parent.parent_id
 
     ##########################
     # 'release' table
     ##########################
 
     def release_missing(self, ids: List[bytes]) -> List[bytes]:
         missing = []
         for id_ in ids:
             if self._releases.get_from_primary_key((id_,)) is None:
                 missing.append(id_)
         return missing
 
     def release_add_one(self, release: ReleaseRow) -> None:
         self._releases.insert(release)
         self.increment_counter("release", 1)
 
     def release_get(self, release_ids: List[str]) -> Iterable[ReleaseRow]:
         for id_ in release_ids:
             row = self._releases.get_from_primary_key((id_,))
             if row:
                 yield row
 
     def release_get_random(self) -> Optional[ReleaseRow]:
         return self._releases.get_random()
 
     ##########################
     # 'snapshot' table
     ##########################
 
     def snapshot_missing(self, ids: List[bytes]) -> List[bytes]:
         missing = []
         for id_ in ids:
             if self._snapshots.get_from_primary_key((id_,)) is None:
                 missing.append(id_)
         return missing
 
     def snapshot_add_one(self, snapshot: SnapshotRow) -> None:
         self._snapshots.insert(snapshot)
         self.increment_counter("snapshot", 1)
 
     def snapshot_get_random(self) -> Optional[SnapshotRow]:
         return self._snapshots.get_random()
 
     ##########################
     # 'snapshot_branch' table
     ##########################
 
     def snapshot_branch_add_one(self, branch: SnapshotBranchRow) -> None:
         self._snapshot_branches.insert(branch)
 
     def snapshot_count_branches(
         self, snapshot_id: Sha1Git, branch_name_exclude_prefix: Optional[bytes] = None,
     ) -> Dict[Optional[str], int]:
         """Returns a dictionary from type names to the number of branches
         of that type."""
         counts: Dict[Optional[str], int] = defaultdict(int)
         for branch in self._snapshot_branches.get_from_partition_key((snapshot_id,)):
             if branch_name_exclude_prefix and branch.name.startswith(
                 branch_name_exclude_prefix
             ):
                 continue
             if branch.target_type is None:
                 target_type = None
             else:
                 target_type = branch.target_type
             counts[target_type] += 1
         return counts
 
     def snapshot_branch_get(
         self,
         snapshot_id: Sha1Git,
         from_: bytes,
         limit: int,
         branch_name_exclude_prefix: Optional[bytes] = None,
     ) -> Iterable[SnapshotBranchRow]:
         count = 0
         for branch in self._snapshot_branches.get_from_partition_key((snapshot_id,)):
             prefix = branch_name_exclude_prefix
             if branch.name >= from_ and (
                 prefix is None or not branch.name.startswith(prefix)
             ):
                 count += 1
                 yield branch
             if count >= limit:
                 break
 
     ##########################
     # 'origin' table
     ##########################
 
     def origin_add_one(self, origin: OriginRow) -> None:
         self._origins.insert(origin)
         self.increment_counter("origin", 1)
 
     def origin_get_by_sha1(self, sha1: bytes) -> Iterable[OriginRow]:
         return self._origins.get_from_partition_key((sha1,))
 
     def origin_get_by_url(self, url: str) -> Iterable[OriginRow]:
         return self.origin_get_by_sha1(origin_url_to_sha1(url))
 
     def origin_list(
         self, start_token: int, limit: int
     ) -> Iterable[Tuple[int, OriginRow]]:
         """Returns an iterable of (token, origin)"""
         matches = [
             (token, row)
             for (token, partition) in self._origins.data.items()
             for (clustering_key, row) in partition.items()
             if token >= start_token
         ]
         matches.sort()
         return matches[0:limit]
 
     def origin_iter_all(self) -> Iterable[OriginRow]:
         return (
             row
             for (token, partition) in self._origins.data.items()
             for (clustering_key, row) in partition.items()
         )
 
+    def origin_bump_next_visit_id(self, origin_url: str, visit_id: int) -> None:
+        origin = list(self.origin_get_by_url(origin_url))[0]
+        origin.next_visit_id = max(origin.next_visit_id, visit_id + 1)
+
     def origin_generate_unique_visit_id(self, origin_url: str) -> int:
         origin = list(self.origin_get_by_url(origin_url))[0]
         visit_id = origin.next_visit_id
         origin.next_visit_id += 1
         return visit_id
 
     ##########################
     # 'origin_visit' table
     ##########################
 
     def origin_visit_get(
         self, origin_url: str, last_visit: Optional[int], limit: int, order: ListOrder,
     ) -> Iterable[OriginVisitRow]:
         visits = list(self._origin_visits.get_from_partition_key((origin_url,)))
 
         if last_visit is not None:
             if order == ListOrder.ASC:
                 visits = [v for v in visits if v.visit > last_visit]
             else:
                 visits = [v for v in visits if v.visit < last_visit]
 
         visits.sort(key=lambda v: v.visit, reverse=order == ListOrder.DESC)
 
         visits = visits[0:limit]
 
         return visits
 
     def origin_visit_add_one(self, visit: OriginVisitRow) -> None:
         self._origin_visits.insert(visit)
         self.increment_counter("origin_visit", 1)
 
     def origin_visit_get_one(
         self, origin_url: str, visit_id: int
     ) -> Optional[OriginVisitRow]:
         return self._origin_visits.get_from_primary_key((origin_url, visit_id))
 
     def origin_visit_get_all(self, origin_url: str) -> Iterable[OriginVisitRow]:
         return self._origin_visits.get_from_partition_key((origin_url,))
 
     def origin_visit_iter(self, start_token: int) -> Iterator[OriginVisitRow]:
         """Returns all origin visits in order from this token,
         and wraps around the token space."""
         return (
             row
             for (token, partition) in self._origin_visits.data.items()
             for (clustering_key, row) in partition.items()
         )
 
     ##########################
     # 'origin_visit_status' table
     ##########################
 
     def origin_visit_status_get_range(
         self,
         origin: str,
         visit: int,
         date_from: Optional[datetime.datetime],
         limit: int,
         order: ListOrder,
     ) -> Iterable[OriginVisitStatusRow]:
         statuses = list(self.origin_visit_status_get(origin, visit))
 
         if date_from is not None:
             if order == ListOrder.ASC:
                 statuses = [s for s in statuses if s.date >= date_from]
             else:
                 statuses = [s for s in statuses if s.date <= date_from]
 
         statuses.sort(key=lambda s: s.date, reverse=order == ListOrder.DESC)
 
         return statuses[0:limit]
 
     def origin_visit_status_add_one(self, visit_update: OriginVisitStatusRow) -> None:
         self._origin_visit_statuses.insert(visit_update)
         self.increment_counter("origin_visit_status", 1)
 
     def origin_visit_status_get_latest(
         self, origin: str, visit: int,
     ) -> Optional[OriginVisitStatusRow]:
         """Given an origin visit id, return its latest origin_visit_status
 
          """
         return next(self.origin_visit_status_get(origin, visit), None)
 
     def origin_visit_status_get(
         self, origin: str, visit: int,
     ) -> Iterator[OriginVisitStatusRow]:
         """Return all origin visit statuses for a given visit
 
         """
         statuses = [
             s
             for s in self._origin_visit_statuses.get_from_partition_key((origin,))
             if s.visit == visit
         ]
         statuses.sort(key=lambda s: s.date, reverse=True)
         return iter(statuses)
 
     ##########################
     # 'metadata_authority' table
     ##########################
 
     def metadata_authority_add(self, authority: MetadataAuthorityRow):
         self._metadata_authorities.insert(authority)
         self.increment_counter("metadata_authority", 1)
 
     def metadata_authority_get(self, type, url) -> Optional[MetadataAuthorityRow]:
         return self._metadata_authorities.get_from_primary_key((url, type))
 
     ##########################
     # 'metadata_fetcher' table
     ##########################
 
     def metadata_fetcher_add(self, fetcher: MetadataFetcherRow):
         self._metadata_fetchers.insert(fetcher)
         self.increment_counter("metadata_fetcher", 1)
 
     def metadata_fetcher_get(self, name, version) -> Optional[MetadataAuthorityRow]:
         return self._metadata_fetchers.get_from_primary_key((name, version))
 
     #########################
     # 'raw_extrinsic_metadata_by_id' table
     #########################
 
     def raw_extrinsic_metadata_by_id_add(
         self, row: RawExtrinsicMetadataByIdRow
     ) -> None:
         self._raw_extrinsic_metadata_by_id.insert(row)
 
     def raw_extrinsic_metadata_get_by_ids(
         self, ids
     ) -> List[RawExtrinsicMetadataByIdRow]:
         results = []
         for id_ in ids:
             result = self._raw_extrinsic_metadata_by_id.get_from_primary_key((id_,))
             if result:
                 results.append(result)
         return results
 
     #########################
     # 'raw_extrinsic_metadata' table
     #########################
 
     def raw_extrinsic_metadata_add(self, raw_extrinsic_metadata):
         self._raw_extrinsic_metadata.insert(raw_extrinsic_metadata)
         self.increment_counter("raw_extrinsic_metadata", 1)
 
     def raw_extrinsic_metadata_get_after_date(
         self,
         target: str,
         authority_type: str,
         authority_url: str,
         after: datetime.datetime,
     ) -> Iterable[RawExtrinsicMetadataRow]:
         metadata = self.raw_extrinsic_metadata_get(
             target, authority_type, authority_url
         )
         return (m for m in metadata if m.discovery_date > after)
 
     def raw_extrinsic_metadata_get_after_date_and_id(
         self,
         target: str,
         authority_type: str,
         authority_url: str,
         after_date: datetime.datetime,
         after_id: bytes,
     ) -> Iterable[RawExtrinsicMetadataRow]:
         metadata = self._raw_extrinsic_metadata.get_from_partition_key((target,))
         after_tuple = (after_date, after_id)
         return (
             m
             for m in metadata
             if m.authority_type == authority_type
             and m.authority_url == authority_url
             and (m.discovery_date, m.id) > after_tuple
         )
 
     def raw_extrinsic_metadata_get(
         self, target: str, authority_type: str, authority_url: str
     ) -> Iterable[RawExtrinsicMetadataRow]:
         metadata = self._raw_extrinsic_metadata.get_from_partition_key((target,))
         return (
             m
             for m in metadata
             if m.authority_type == authority_type and m.authority_url == authority_url
         )
 
     def raw_extrinsic_metadata_get_authorities(
         self, target: str
     ) -> Iterable[Tuple[str, str]]:
         metadata = self._raw_extrinsic_metadata.get_from_partition_key((target,))
         return ((m.authority_type, m.authority_url) for m in metadata)
 
     #########################
     # 'extid' table
     #########################
     def _extid_add_finalize(self, extid: ExtIDRow) -> None:
         self._extid.insert(extid)
         self.increment_counter("extid", 1)
 
     def extid_add_prepare(self, extid: ExtIDRow):
         finalizer = functools.partial(self._extid_add_finalize, extid)
         return (self._extid.token(self._extid.partition_key(extid)), finalizer)
 
     def extid_index_add_one(self, row: ExtIDByTargetRow) -> None:
         pass
 
     def extid_get_from_pk(
         self, extid_type: str, extid: bytes, extid_version: int, target: ExtendedSWHID,
     ) -> Optional[ExtIDRow]:
         primary_key = self._extid.primary_key_from_dict(
             dict(
                 extid_type=extid_type,
                 extid=extid,
                 extid_version=extid_version,
                 target_type=target.object_type.value,
                 target=target.object_id,
             )
         )
         return self._extid.get_from_primary_key(primary_key)
 
     def extid_get_from_extid(self, extid_type: str, extid: bytes) -> Iterable[ExtIDRow]:
         return (
             row
             for pk, row in self._extid.iter_all()
             if row.extid_type == extid_type and row.extid == extid
         )
 
     def extid_get_from_target(
         self, target_type: str, target: bytes
     ) -> Iterable[ExtIDRow]:
         return (
             row
             for pk, row in self._extid.iter_all()
             if row.target_type == target_type and row.target == target
         )
 
 
 class InMemoryStorage(CassandraStorage):
     _cql_runner: InMemoryCqlRunner  # type: ignore
 
     def __init__(self, journal_writer=None):
         self.reset()
         self.journal_writer = JournalWriter(journal_writer)
         self._allow_overwrite = False
 
     def reset(self):
         self._cql_runner = InMemoryCqlRunner()
         self.objstorage = ObjStorage({"cls": "memory"})
 
     def check_config(self, *, check_write: bool) -> bool:
         return True
diff --git a/swh/storage/tests/storage_tests.py b/swh/storage/tests/storage_tests.py
index a8906cbb..246eb305 100644
--- a/swh/storage/tests/storage_tests.py
+++ b/swh/storage/tests/storage_tests.py
@@ -1,4547 +1,4595 @@
 # Copyright (C) 2015-2021  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 from collections import defaultdict
 import datetime
 from datetime import timedelta
 import inspect
 import itertools
 import math
 import random
 from typing import Any, ClassVar, Dict, Iterator, Optional
 from unittest.mock import MagicMock
 
 import attr
 from hypothesis import HealthCheck, given, settings, strategies
 import pytest
 
 from swh.core.api.classes import stream_results
 from swh.model import from_disk
 from swh.model.hashutil import DEFAULT_ALGORITHMS, hash_to_bytes
 from swh.model.hypothesis_strategies import objects
 from swh.model.identifiers import CoreSWHID, ObjectType
 from swh.model.model import (
     Content,
     Directory,
     ExtID,
     Origin,
     OriginVisit,
     OriginVisitStatus,
     Person,
     RawExtrinsicMetadata,
     Revision,
     SkippedContent,
     Snapshot,
     SnapshotBranch,
     TargetType,
 )
 from swh.storage import get_storage
 from swh.storage.common import origin_url_to_sha1 as sha1
 from swh.storage.exc import HashCollision, StorageArgumentException
 from swh.storage.interface import ListOrder, PagedResult, StorageInterface
 from swh.storage.tests.conftest import function_scoped_fixture_check
 from swh.storage.utils import (
     content_hex_hashes,
     now,
     remove_keys,
     round_to_milliseconds,
 )
 
 
 def transform_entries(
     storage: StorageInterface, dir_: Directory, *, prefix: bytes = b""
 ) -> Iterator[Dict[str, Any]]:
     """Iterate through a directory's entries, and yields the items 'directory_ls' is
        expected to return; including content metadata for file entries."""
 
     for ent in dir_.entries:
         if ent.type == "dir":
             yield {
                 "dir_id": dir_.id,
                 "type": ent.type,
                 "target": ent.target,
                 "name": prefix + ent.name,
                 "perms": ent.perms,
                 "status": None,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "length": None,
             }
         elif ent.type == "file":
             contents = storage.content_find({"sha1_git": ent.target})
             assert contents
             ent_dict = contents[0].to_dict()
             for key in ["ctime", "blake2s256"]:
                 ent_dict.pop(key, None)
             ent_dict.update(
                 {
                     "dir_id": dir_.id,
                     "type": ent.type,
                     "target": ent.target,
                     "name": prefix + ent.name,
                     "perms": ent.perms,
                 }
             )
             yield ent_dict
 
 
 def assert_contents_ok(
     expected_contents, actual_contents, keys_to_check={"sha1", "data"}
 ):
     """Assert that a given list of contents matches on a given set of keys.
 
     """
     for k in keys_to_check:
         expected_list = set([c.get(k) for c in expected_contents])
         actual_list = set([c.get(k) for c in actual_contents])
         assert actual_list == expected_list, k
 
 
 class LazyContent(Content):
     def with_data(self):
         return Content.from_dict({**self.to_dict(), "data": b"42\n"})
 
 
 class TestStorage:
     """Main class for Storage testing.
 
     This class is used as-is to test local storage (see TestLocalStorage
     below) and remote storage (see TestRemoteStorage in
     test_remote_storage.py.
 
     We need to have the two classes inherit from this base class
     separately to avoid nosetests running the tests from the base
     class twice.
     """
 
     maxDiff = None  # type: ClassVar[Optional[int]]
 
     def test_types(self, swh_storage_backend_config):
         """Checks all methods of StorageInterface are implemented by this
         backend, and that they have the same signature."""
         # Create an instance of the protocol (which cannot be instantiated
         # directly, so this creates a subclass, then instantiates it)
         interface = type("_", (StorageInterface,), {})()
         storage = get_storage(**swh_storage_backend_config)
 
         assert "content_add" in dir(interface)
 
         missing_methods = []
 
         for meth_name in dir(interface):
             if meth_name.startswith("_"):
                 continue
             interface_meth = getattr(interface, meth_name)
             try:
                 concrete_meth = getattr(storage, meth_name)
             except AttributeError:
                 if not getattr(interface_meth, "deprecated_endpoint", False):
                     # The backend is missing a (non-deprecated) endpoint
                     missing_methods.append(meth_name)
                 continue
 
             expected_signature = inspect.signature(interface_meth)
             actual_signature = inspect.signature(concrete_meth)
 
             assert expected_signature == actual_signature, meth_name
 
         assert missing_methods == []
 
         # If all the assertions above succeed, then this one should too.
         # But there's no harm in double-checking.
         # And we could replace the assertions above by this one, but unlike
         # the assertions above, it doesn't explain what is missing.
         assert isinstance(storage, StorageInterface)
 
     def test_check_config(self, swh_storage):
         assert swh_storage.check_config(check_write=True)
         assert swh_storage.check_config(check_write=False)
 
     def test_content_add(self, swh_storage, sample_data):
         cont = sample_data.content
 
         insertion_start_time = now()
         actual_result = swh_storage.content_add([cont])
         insertion_end_time = now()
 
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
 
         assert swh_storage.content_get_data(cont.sha1) == cont.data
 
         expected_cont = attr.evolve(cont, data=None)
 
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         for obj in contents:
             assert insertion_start_time <= obj.ctime
             assert obj.ctime <= insertion_end_time
             assert obj == expected_cont
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["content"] == 1
 
     def test_content_add_from_lazy_content(self, swh_storage, sample_data):
         cont = sample_data.content
         lazy_content = LazyContent.from_dict(cont.to_dict())
 
         insertion_start_time = now()
 
         actual_result = swh_storage.content_add([lazy_content])
 
         insertion_end_time = now()
 
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
 
         # the fact that we retrieve the content object from the storage with
         # the correct 'data' field ensures it has been 'called'
         assert swh_storage.content_get_data(cont.sha1) == cont.data
 
         expected_cont = attr.evolve(lazy_content, data=None, ctime=None)
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         for obj in contents:
             assert insertion_start_time <= obj.ctime
             assert obj.ctime <= insertion_end_time
             assert attr.evolve(obj, ctime=None).to_dict() == expected_cont.to_dict()
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["content"] == 1
 
     def test_content_get_data_missing(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
 
         swh_storage.content_add([cont])
 
         # Query a single missing content
         actual_content_data = swh_storage.content_get_data(cont2.sha1)
         assert actual_content_data is None
 
         # Check content_get does not abort after finding a missing content
         actual_content_data = swh_storage.content_get_data(cont.sha1)
         assert actual_content_data == cont.data
         actual_content_data = swh_storage.content_get_data(cont2.sha1)
         assert actual_content_data is None
 
     def test_content_add_different_input(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
 
         actual_result = swh_storage.content_add([cont, cont2])
         assert actual_result == {
             "content:add": 2,
             "content:add:bytes": cont.length + cont2.length,
         }
 
     def test_content_add_twice(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
 
         actual_result = swh_storage.content_add([cont])
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
         assert len(swh_storage.journal_writer.journal.objects) == 1
 
         actual_result = swh_storage.content_add([cont, cont2])
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont2.length,
         }
         assert 2 <= len(swh_storage.journal_writer.journal.objects) <= 3
 
         assert len(swh_storage.content_find(cont.to_dict())) == 1
         assert len(swh_storage.content_find(cont2.to_dict())) == 1
 
     def test_content_add_collision(self, swh_storage, sample_data):
         cont1 = sample_data.content
 
         # create (corrupted) content with same sha1{,_git} but != sha256
         sha256_array = bytearray(cont1.sha256)
         sha256_array[0] += 1
         cont1b = attr.evolve(cont1, sha256=bytes(sha256_array))
 
         with pytest.raises(HashCollision) as cm:
             swh_storage.content_add([cont1, cont1b])
 
         exc = cm.value
         actual_algo = exc.algo
         assert actual_algo in ["sha1", "sha1_git"]
         actual_id = exc.hash_id
         assert actual_id == getattr(cont1, actual_algo).hex()
         collisions = exc.args[2]
         assert len(collisions) == 2
         assert collisions == [
             content_hex_hashes(cont1.hashes()),
             content_hex_hashes(cont1b.hashes()),
         ]
         assert exc.colliding_content_hashes() == [
             cont1.hashes(),
             cont1b.hashes(),
         ]
 
     def test_content_add_duplicate(self, swh_storage, sample_data):
         cont = sample_data.content
         swh_storage.content_add([cont, cont])
 
         assert swh_storage.content_get_data(cont.sha1) == cont.data
 
     def test_content_update(self, swh_storage, sample_data):
         cont1 = sample_data.content
 
         if hasattr(swh_storage, "journal_writer"):
             swh_storage.journal_writer.journal = None  # TODO, not supported
 
         swh_storage.content_add([cont1])
 
         # alter the sha1_git for example
         cont1b = attr.evolve(
             cont1, sha1_git=hash_to_bytes("3a60a5275d0333bf13468e8b3dcab90f4046e654")
         )
 
         swh_storage.content_update([cont1b.to_dict()], keys=["sha1_git"])
 
         actual_contents = swh_storage.content_get([cont1.sha1])
         expected_content = attr.evolve(cont1b, data=None)
         assert actual_contents == [expected_content]
 
     def test_content_add_metadata(self, swh_storage, sample_data):
         cont = attr.evolve(sample_data.content, data=None, ctime=now())
 
         actual_result = swh_storage.content_add_metadata([cont])
         assert actual_result == {
             "content:add": 1,
         }
 
         expected_cont = cont
         assert swh_storage.content_get([cont.sha1]) == [expected_cont]
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         for obj in contents:
             obj = attr.evolve(obj, ctime=None)
             assert obj == cont
 
     def test_content_add_metadata_different_input(self, swh_storage, sample_data):
         contents = sample_data.contents[:2]
         cont = attr.evolve(contents[0], data=None, ctime=now())
         cont2 = attr.evolve(contents[1], data=None, ctime=now())
 
         actual_result = swh_storage.content_add_metadata([cont, cont2])
         assert actual_result == {
             "content:add": 2,
         }
 
     def test_content_add_metadata_collision(self, swh_storage, sample_data):
         cont1 = attr.evolve(sample_data.content, data=None, ctime=now())
 
         # create (corrupted) content with same sha1{,_git} but != sha256
         sha1_git_array = bytearray(cont1.sha256)
         sha1_git_array[0] += 1
         cont1b = attr.evolve(cont1, sha256=bytes(sha1_git_array))
 
         with pytest.raises(HashCollision) as cm:
             swh_storage.content_add_metadata([cont1, cont1b])
 
         exc = cm.value
         actual_algo = exc.algo
         assert actual_algo in ["sha1", "sha1_git", "blake2s256"]
         actual_id = exc.hash_id
         assert actual_id == getattr(cont1, actual_algo).hex()
         collisions = exc.args[2]
         assert len(collisions) == 2
         assert collisions == [
             content_hex_hashes(cont1.hashes()),
             content_hex_hashes(cont1b.hashes()),
         ]
         assert exc.colliding_content_hashes() == [
             cont1.hashes(),
             cont1b.hashes(),
         ]
 
     def test_content_add_objstorage_first(self, swh_storage, sample_data):
         """Tests the objstorage is written to before the DB and journal"""
         cont = sample_data.content
 
         swh_storage.objstorage.content_add = MagicMock(side_effect=Exception("Oops"))
 
         # Try to add, but the objstorage crashes
         try:
             swh_storage.content_add([cont])
         except Exception:
             pass
 
         # The DB must be written to after the objstorage, so the DB should be
         # unchanged if the objstorage crashed
         assert swh_storage.content_get_data(cont.sha1) is None
 
         # The journal too
         assert list(swh_storage.journal_writer.journal.objects) == []
 
     def test_skipped_content_add(self, swh_storage, sample_data):
         contents = sample_data.skipped_contents[:2]
         cont = contents[0]
         cont2 = attr.evolve(contents[1], blake2s256=None)
 
         contents_dict = [c.to_dict() for c in [cont, cont2]]
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
 
         assert missing == [cont.hashes(), cont2.hashes()]
 
         actual_result = swh_storage.skipped_content_add([cont, cont, cont2])
 
         assert 2 <= actual_result.pop("skipped_content:add") <= 3
         assert actual_result == {}
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert missing == []
 
     def test_skipped_content_add_missing_hashes(self, swh_storage, sample_data):
         cont, cont2 = [
             attr.evolve(c, sha1_git=None) for c in sample_data.skipped_contents[:2]
         ]
         contents_dict = [c.to_dict() for c in [cont, cont2]]
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert len(missing) == 2
 
         actual_result = swh_storage.skipped_content_add([cont, cont, cont2])
 
         assert 2 <= actual_result.pop("skipped_content:add") <= 3
         assert actual_result == {}
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert missing == []
 
     def test_skipped_content_missing_partial_hash(self, swh_storage, sample_data):
         cont = sample_data.skipped_content
         cont2 = attr.evolve(cont, sha1_git=None)
         contents_dict = [c.to_dict() for c in [cont, cont2]]
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert len(missing) == 2
 
         actual_result = swh_storage.skipped_content_add([cont])
 
         assert actual_result.pop("skipped_content:add") == 1
         assert actual_result == {}
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert missing == [cont2.hashes()]
 
     @pytest.mark.property_based
     @settings(
         deadline=None,  # this test is very slow
         suppress_health_check=function_scoped_fixture_check,
     )
     @given(
         strategies.sets(
             elements=strategies.sampled_from(["sha256", "sha1_git", "blake2s256"]),
             min_size=0,
         )
     )
     def test_content_missing(self, swh_storage, sample_data, algos):
         algos |= {"sha1"}
         content, missing_content = [sample_data.content2, sample_data.skipped_content]
         swh_storage.content_add([content])
 
         test_contents = [content.to_dict()]
         missing_per_hash = defaultdict(list)
         for i in range(256):
             test_content = missing_content.to_dict()
             for hash in algos:
                 test_content[hash] = bytes([i]) + test_content[hash][1:]
                 missing_per_hash[hash].append(test_content[hash])
             test_contents.append(test_content)
 
         assert set(swh_storage.content_missing(test_contents)) == set(
             missing_per_hash["sha1"]
         )
 
         for hash in algos:
             assert set(
                 swh_storage.content_missing(test_contents, key_hash=hash)
             ) == set(missing_per_hash[hash])
 
     @pytest.mark.property_based
     @settings(suppress_health_check=function_scoped_fixture_check,)
     @given(
         strategies.sets(
             elements=strategies.sampled_from(["sha256", "sha1_git", "blake2s256"]),
             min_size=0,
         )
     )
     def test_content_missing_unknown_algo(self, swh_storage, sample_data, algos):
         algos |= {"sha1"}
         content, missing_content = [sample_data.content2, sample_data.skipped_content]
         swh_storage.content_add([content])
 
         test_contents = [content.to_dict()]
         missing_per_hash = defaultdict(list)
         for i in range(16):
             test_content = missing_content.to_dict()
             for hash in algos:
                 test_content[hash] = bytes([i]) + test_content[hash][1:]
                 missing_per_hash[hash].append(test_content[hash])
             test_content["nonexisting_algo"] = b"\x00"
             test_contents.append(test_content)
 
         assert set(swh_storage.content_missing(test_contents)) == set(
             missing_per_hash["sha1"]
         )
 
         for hash in algos:
             assert set(
                 swh_storage.content_missing(test_contents, key_hash=hash)
             ) == set(missing_per_hash[hash])
 
     def test_content_missing_per_sha1(self, swh_storage, sample_data):
         # given
         cont = sample_data.content
         cont2 = sample_data.content2
         missing_cont = sample_data.skipped_content
         missing_cont2 = sample_data.skipped_content2
         swh_storage.content_add([cont, cont2])
 
         # when
         gen = swh_storage.content_missing_per_sha1(
             [cont.sha1, missing_cont.sha1, cont2.sha1, missing_cont2.sha1]
         )
         # then
         assert list(gen) == [missing_cont.sha1, missing_cont2.sha1]
 
     def test_content_missing_per_sha1_git(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
         missing_cont = sample_data.skipped_content
         missing_cont2 = sample_data.skipped_content2
 
         swh_storage.content_add([cont, cont2])
 
         contents = [
             cont.sha1_git,
             cont2.sha1_git,
             missing_cont.sha1_git,
             missing_cont2.sha1_git,
         ]
 
         missing_contents = swh_storage.content_missing_per_sha1_git(contents)
         assert list(missing_contents) == [missing_cont.sha1_git, missing_cont2.sha1_git]
 
         missing_contents = swh_storage.content_missing_per_sha1_git([])
         assert list(missing_contents) == []
 
     def test_content_get_partition(self, swh_storage, swh_contents):
         """content_get_partition paginates results if limit exceeded"""
         expected_contents = [
             attr.evolve(c, data=None) for c in swh_contents if c.status != "absent"
         ]
 
         actual_contents = []
         for i in range(16):
             actual_result = swh_storage.content_get_partition(i, 16)
             assert actual_result.next_page_token is None
             actual_contents.extend(actual_result.results)
 
         assert len(actual_contents) == len(expected_contents)
         for content in actual_contents:
             assert content in expected_contents
             assert content.ctime is None
 
     def test_content_get_partition_full(self, swh_storage, swh_contents):
         """content_get_partition for a single partition returns all available contents
 
         """
         expected_contents = [
             attr.evolve(c, data=None) for c in swh_contents if c.status != "absent"
         ]
 
         actual_result = swh_storage.content_get_partition(0, 1)
         assert actual_result.next_page_token is None
 
         actual_contents = actual_result.results
         assert len(actual_contents) == len(expected_contents)
         for content in actual_contents:
             assert content in expected_contents
 
     def test_content_get_partition_empty(self, swh_storage, swh_contents):
         """content_get_partition when at least one of the partitions is empty"""
         expected_contents = {
             cont.sha1 for cont in swh_contents if cont.status != "absent"
         }
         # nb_partitions = smallest power of 2 such that at least one of
         # the partitions is empty
         nb_partitions = 1 << math.floor(math.log2(len(swh_contents)) + 1)
 
         seen_sha1s = []
 
         for i in range(nb_partitions):
             actual_result = swh_storage.content_get_partition(
                 i, nb_partitions, limit=len(swh_contents) + 1
             )
 
             for content in actual_result.results:
                 seen_sha1s.append(content.sha1)
 
             # Limit is higher than the max number of results
             assert actual_result.next_page_token is None
 
         assert set(seen_sha1s) == expected_contents
 
     def test_content_get_partition_limit_none(self, swh_storage):
         """content_get_partition call with wrong limit input should fail"""
         with pytest.raises(StorageArgumentException, match="limit should not be None"):
             swh_storage.content_get_partition(1, 16, limit=None)
 
     def test_content_get_partition_pagination_generate(self, swh_storage, swh_contents):
         """content_get_partition returns contents within range provided"""
         expected_contents = [
             attr.evolve(c, data=None) for c in swh_contents if c.status != "absent"
         ]
 
         # retrieve contents
         actual_contents = []
         for i in range(4):
             page_token = None
             while True:
                 actual_result = swh_storage.content_get_partition(
                     i, 4, limit=3, page_token=page_token
                 )
                 actual_contents.extend(actual_result.results)
                 page_token = actual_result.next_page_token
 
                 if page_token is None:
                     break
 
         assert len(actual_contents) == len(expected_contents)
         for content in actual_contents:
             assert content in expected_contents
 
     @pytest.mark.parametrize("algo", sorted(DEFAULT_ALGORITHMS))
     def test_content_get(self, swh_storage, sample_data, algo):
         cont1, cont2 = sample_data.contents[:2]
 
         swh_storage.content_add([cont1, cont2])
 
         actual_contents = swh_storage.content_get(
             [getattr(cont1, algo), getattr(cont2, algo)], algo
         )
 
         # we only retrieve the metadata so no data nor ctime within
         expected_contents = [attr.evolve(c, data=None) for c in [cont1, cont2]]
 
         assert actual_contents == expected_contents
         for content in actual_contents:
             assert content.ctime is None
 
     @pytest.mark.parametrize("algo", sorted(DEFAULT_ALGORITHMS))
     def test_content_get_missing(self, swh_storage, sample_data, algo):
         cont1, cont2 = sample_data.contents[:2]
         assert cont1.sha1 != cont2.sha1
         missing_cont = sample_data.skipped_content
 
         swh_storage.content_add([cont1, cont2])
 
         actual_contents = swh_storage.content_get(
             [getattr(cont1, algo), getattr(cont2, algo), getattr(missing_cont, algo)],
             algo,
         )
 
         expected_contents = [
             attr.evolve(c, data=None) if c else None for c in [cont1, cont2, None]
         ]
         assert actual_contents == expected_contents
 
     def test_content_get_random(self, swh_storage, sample_data):
         cont, cont2, cont3 = sample_data.contents[:3]
         swh_storage.content_add([cont, cont2, cont3])
 
         assert swh_storage.content_get_random() in {
             cont.sha1_git,
             cont2.sha1_git,
             cont3.sha1_git,
         }
 
     def test_directory_add(self, swh_storage, sample_data):
         content = sample_data.content
         directory = sample_data.directories[1]
         assert directory.entries[0].target == content.sha1_git
         swh_storage.content_add([content])
 
         init_missing = list(swh_storage.directory_missing([directory.id]))
         assert [directory.id] == init_missing
 
         actual_result = swh_storage.directory_add([directory])
         assert actual_result == {"directory:add": 1}
 
         assert ("directory", directory) in list(
             swh_storage.journal_writer.journal.objects
         )
 
         actual_data = list(swh_storage.directory_ls(directory.id))
         expected_data = list(transform_entries(swh_storage, directory))
 
         for data in actual_data:
             assert data in expected_data
 
         after_missing = list(swh_storage.directory_missing([directory.id]))
         assert after_missing == []
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["directory"] == 1
 
     def test_directory_add_twice(self, swh_storage, sample_data):
         directory = sample_data.directories[1]
 
         actual_result = swh_storage.directory_add([directory])
         assert actual_result == {"directory:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", directory)
         ]
 
         actual_result = swh_storage.directory_add([directory])
         assert actual_result == {"directory:add": 0}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", directory)
         ]
 
     def test_directory_ls_recursive(self, swh_storage, sample_data):
         # create consistent dataset regarding the directories we want to list
         content, content2 = sample_data.contents[:2]
         swh_storage.content_add([content, content2])
         dir1, dir2, dir3 = sample_data.directories[:3]
 
         dir_ids = [d.id for d in [dir1, dir2, dir3]]
         init_missing = list(swh_storage.directory_missing(dir_ids))
         assert init_missing == dir_ids
 
         actual_result = swh_storage.directory_add([dir1, dir2, dir3])
         assert actual_result == {"directory:add": 3}
 
         # List directory containing one file
         actual_data = list(swh_storage.directory_ls(dir1.id, recursive=True))
         expected_data = list(transform_entries(swh_storage, dir1))
         for data in actual_data:
             assert data in expected_data
 
         # List directory containing a file and an unknown subdirectory
         actual_data = list(swh_storage.directory_ls(dir2.id, recursive=True))
         expected_data = list(transform_entries(swh_storage, dir2))
         for data in actual_data:
             assert data in expected_data
 
         # List directory containing both a known and unknown subdirectory, entries
         # should be both those of the directory and of the known subdir (up to contents)
         actual_data = list(swh_storage.directory_ls(dir3.id, recursive=True))
         expected_data = list(
             itertools.chain(
                 transform_entries(swh_storage, dir3),
                 transform_entries(swh_storage, dir2, prefix=b"subdir/"),
             )
         )
 
         for data in actual_data:
             assert data in expected_data
 
     def test_directory_ls_non_recursive(self, swh_storage, sample_data):
         # create consistent dataset regarding the directories we want to list
         content, content2 = sample_data.contents[:2]
         swh_storage.content_add([content, content2])
         dir1, dir2, dir3, _, dir5 = sample_data.directories[:5]
 
         dir_ids = [d.id for d in [dir1, dir2, dir3, dir5]]
         init_missing = list(swh_storage.directory_missing(dir_ids))
         assert init_missing == dir_ids
 
         actual_result = swh_storage.directory_add([dir1, dir2, dir3, dir5])
         assert actual_result == {"directory:add": 4}
 
         # List directory containing a file and an unknown subdirectory
         actual_data = list(swh_storage.directory_ls(dir1.id))
         expected_data = list(transform_entries(swh_storage, dir1))
         for data in actual_data:
             assert data in expected_data
 
         # List directory containing a single file
         actual_data = list(swh_storage.directory_ls(dir2.id))
         expected_data = list(transform_entries(swh_storage, dir2))
         for data in actual_data:
             assert data in expected_data
 
         # List directory containing a known subdirectory, entries should
         # only be those of the parent directory, not of the subdir
         actual_data = list(swh_storage.directory_ls(dir3.id))
         expected_data = list(transform_entries(swh_storage, dir3))
         for data in actual_data:
             assert data in expected_data
 
     def test_directory_ls_missing_content(self, swh_storage, sample_data):
         swh_storage.directory_add([sample_data.directory2])
         assert list(swh_storage.directory_ls(sample_data.directory2.id)) == [
             {
                 "dir_id": sample_data.directory2.id,
                 "length": None,
                 "name": b"oof",
                 "perms": 33188,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "target": sample_data.directory2.entries[0].target,
                 "type": "file",
             },
         ]
 
     def test_directory_ls_skipped_content(self, swh_storage, sample_data):
         swh_storage.directory_add([sample_data.directory2])
 
         cont = SkippedContent(
             sha1_git=sample_data.directory2.entries[0].target,
             sha1=b"c" * 20,
             sha256=None,
             blake2s256=None,
             length=42,
             status="absent",
             reason="You need a premium subscription to access this content",
         )
         swh_storage.skipped_content_add([cont])
 
         assert list(swh_storage.directory_ls(sample_data.directory2.id)) == [
             {
                 "dir_id": sample_data.directory2.id,
                 "length": 42,
                 "name": b"oof",
                 "perms": 33188,
                 "sha1": b"c" * 20,
                 "sha1_git": sample_data.directory2.entries[0].target,
                 "sha256": None,
                 "status": "absent",
                 "target": sample_data.directory2.entries[0].target,
                 "type": "file",
             },
         ]
 
     def test_directory_entry_get_by_path(self, swh_storage, sample_data):
         cont, content2 = sample_data.contents[:2]
         dir1, dir2, dir3, dir4, dir5 = sample_data.directories[:5]
 
         # given
         dir_ids = [d.id for d in [dir1, dir2, dir3, dir4, dir5]]
         init_missing = list(swh_storage.directory_missing(dir_ids))
         assert init_missing == dir_ids
 
         actual_result = swh_storage.directory_add([dir3, dir4])
         assert actual_result == {"directory:add": 2}
 
         expected_entries = [
             {
                 "dir_id": dir3.id,
                 "name": b"foo",
                 "type": "file",
                 "target": cont.sha1_git,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "perms": from_disk.DentryPerms.content,
                 "length": None,
             },
             {
                 "dir_id": dir3.id,
                 "name": b"subdir",
                 "type": "dir",
                 "target": dir2.id,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "perms": from_disk.DentryPerms.directory,
                 "length": None,
             },
             {
                 "dir_id": dir3.id,
                 "name": b"hello",
                 "type": "file",
                 "target": content2.sha1_git,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "perms": from_disk.DentryPerms.content,
                 "length": None,
             },
         ]
 
         # when (all must be found here)
         for entry, expected_entry in zip(dir3.entries, expected_entries):
             actual_entry = swh_storage.directory_entry_get_by_path(
                 dir3.id, [entry.name]
             )
             assert actual_entry == expected_entry
 
         # same, but deeper
         for entry, expected_entry in zip(dir3.entries, expected_entries):
             actual_entry = swh_storage.directory_entry_get_by_path(
                 dir4.id, [b"subdir1", entry.name]
             )
             expected_entry = expected_entry.copy()
             expected_entry["name"] = b"subdir1/" + expected_entry["name"]
             assert actual_entry == expected_entry
 
         # when (nothing should be found here since `dir` is not persisted.)
         for entry in dir2.entries:
             actual_entry = swh_storage.directory_entry_get_by_path(
                 dir2.id, [entry.name]
             )
             assert actual_entry is None
 
     def test_directory_get_entries_pagination(self, swh_storage, sample_data):
         # Note: this test assumes entries are returned in lexicographic order,
         # which is not actually guaranteed by the interface.
         dir_ = sample_data.directory3
         entries = sorted(dir_.entries, key=lambda entry: entry.name)
         swh_storage.directory_add(sample_data.directories)
 
         # No pagination needed
         actual_data = swh_storage.directory_get_entries(dir_.id)
         assert actual_data == PagedResult(results=entries, next_page_token=None)
 
         # A little pagination
         actual_data = swh_storage.directory_get_entries(dir_.id, limit=2)
         assert actual_data.results == entries[0:2]
         assert actual_data.next_page_token is not None
 
         actual_data = swh_storage.directory_get_entries(
             dir_.id, page_token=actual_data.next_page_token
         )
         assert actual_data == PagedResult(results=entries[2:], next_page_token=None)
 
     @pytest.mark.parametrize("limit", [1, 2, 3, 4, 5])
     def test_directory_get_entries(self, swh_storage, sample_data, limit):
         dir_ = sample_data.directory3
         swh_storage.directory_add(sample_data.directories)
 
         actual_data = list(
             stream_results(swh_storage.directory_get_entries, dir_.id, limit=limit,)
         )
         assert sorted(actual_data) == sorted(dir_.entries)
 
     def test_directory_get_random(self, swh_storage, sample_data):
         dir1, dir2, dir3 = sample_data.directories[:3]
         swh_storage.directory_add([dir1, dir2, dir3])
 
         assert swh_storage.directory_get_random() in {
             dir1.id,
             dir2.id,
             dir3.id,
         }
 
     def test_revision_add(self, swh_storage, sample_data):
         revision = sample_data.revision
         init_missing = swh_storage.revision_missing([revision.id])
         assert list(init_missing) == [revision.id]
 
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 1}
 
         end_missing = swh_storage.revision_missing([revision.id])
         assert list(end_missing) == []
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("revision", revision)
         ]
 
         # already there so nothing added
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 0}
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["revision"] == 1
 
     def test_revision_add_twice(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("revision", revision)
         ]
 
         actual_result = swh_storage.revision_add([revision, revision2])
         assert actual_result == {"revision:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("revision", revision),
             ("revision", revision2),
         ]
 
     def test_revision_add_name_clash(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         revision1 = attr.evolve(
             revision,
             author=Person(
                 fullname=b"John Doe <john.doe@example.com>",
                 name=b"John Doe",
                 email=b"john.doe@example.com",
             ),
         )
         revision2 = attr.evolve(
             revision2,
             author=Person(
                 fullname=b"John Doe <john.doe@example.com>",
                 name=b"John Doe ",
                 email=b"john.doe@example.com ",
             ),
         )
         actual_result = swh_storage.revision_add([revision1, revision2])
         assert actual_result == {"revision:add": 2}
 
     def test_revision_get_order(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         add_result = swh_storage.revision_add([revision, revision2])
         assert add_result == {"revision:add": 2}
 
         # order 1
         actual_revisions = swh_storage.revision_get([revision.id, revision2.id])
         assert actual_revisions == [revision, revision2]
 
         # order 2
         actual_revisions2 = swh_storage.revision_get([revision2.id, revision.id])
         assert actual_revisions2 == [revision2, revision]
 
     def test_revision_log(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # rev4 -is-child-of-> rev3 -> rev1, (rev2 -> rev1)
         swh_storage.revision_add([revision1, revision2, revision3, revision4])
 
         # when
         results = list(swh_storage.revision_log([revision4.id]))
 
         # for comparison purposes
         actual_results = [Revision.from_dict(r) for r in results]
         assert len(actual_results) == 4  # rev4 -child-> rev3 -> rev1, (rev2 -> rev1)
         assert actual_results == [revision4, revision3, revision1, revision2]
 
     def test_revision_log_with_limit(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # revision4 -is-child-of-> revision3
         swh_storage.revision_add([revision3, revision4])
         results = list(swh_storage.revision_log([revision4.id], 1))
 
         actual_results = [Revision.from_dict(r) for r in results]
         assert len(actual_results) == 1
         assert actual_results[0] == revision4
 
     def test_revision_log_unknown_revision(self, swh_storage, sample_data):
         revision = sample_data.revision
         rev_log = list(swh_storage.revision_log([revision.id]))
         assert rev_log == []
 
     def test_revision_shortlog(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # rev4 -is-child-of-> rev3 -> (rev1, rev2); rev2 -> rev1
         swh_storage.revision_add([revision1, revision2, revision3, revision4])
 
         results = list(swh_storage.revision_shortlog([revision4.id]))
         actual_results = [[id, tuple(parents)] for (id, parents) in results]
 
         assert len(actual_results) == 4
         assert actual_results == [
             [revision4.id, revision4.parents],
             [revision3.id, revision3.parents],
             [revision1.id, revision1.parents],
             [revision2.id, revision2.parents],
         ]
 
     def test_revision_shortlog_with_limit(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # revision4 -is-child-of-> revision3
         swh_storage.revision_add([revision1, revision2, revision3, revision4])
         results = list(swh_storage.revision_shortlog([revision4.id], 1))
         actual_results = [[id, tuple(parents)] for (id, parents) in results]
 
         assert len(actual_results) == 1
         assert list(actual_results[0]) == [revision4.id, revision4.parents]
 
     def test_revision_get(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         swh_storage.revision_add([revision])
 
         actual_revisions = swh_storage.revision_get([revision.id, revision2.id])
 
         assert len(actual_revisions) == 2
         assert actual_revisions == [revision, None]
 
     def test_revision_get_no_parents(self, swh_storage, sample_data):
         revision = sample_data.revision
         swh_storage.revision_add([revision])
 
         actual_revision = swh_storage.revision_get([revision.id])[0]
 
         assert revision.parents == ()
         assert actual_revision.parents == ()  # no parents on this one
 
     def test_revision_get_random(self, swh_storage, sample_data):
         revision1, revision2, revision3 = sample_data.revisions[:3]
 
         swh_storage.revision_add([revision1, revision2, revision3])
 
         assert swh_storage.revision_get_random() in {
             revision1.id,
             revision2.id,
             revision3.id,
         }
 
     def test_revision_missing_many(self, swh_storage, sample_data):
         """Large number of revision ids to check can cause ScyllaDB to reject
         queries."""
         revision = sample_data.revision
         ids = [bytes([b1, b2]) * 10 for b1 in range(256) for b2 in range(10)]
         ids.append(revision.id)
         ids.sort()
         init_missing = swh_storage.revision_missing(ids)
         assert set(init_missing) == set(ids)
 
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 1}
 
         end_missing = swh_storage.revision_missing(ids)
         assert set(end_missing) == set(ids) - {revision.id}
 
     def test_extid_add_git(self, swh_storage, sample_data):
 
         gitids = [
             revision.id
             for revision in sample_data.revisions
             if revision.type.value == "git"
         ]
         extids = [
             ExtID(
                 extid=gitid,
                 extid_type="git",
                 target=CoreSWHID(object_id=gitid, object_type=ObjectType.REVISION,),
             )
             for gitid in gitids
         ]
 
         assert swh_storage.extid_get_from_extid("git", gitids) == []
         assert swh_storage.extid_get_from_target(ObjectType.REVISION, gitids) == []
 
         summary = swh_storage.extid_add(extids)
         assert summary == {"extid:add": len(gitids)}
 
         assert swh_storage.extid_get_from_extid("git", gitids) == extids
         assert swh_storage.extid_get_from_target(ObjectType.REVISION, gitids) == extids
 
         assert swh_storage.extid_get_from_extid("hg", gitids) == []
         assert swh_storage.extid_get_from_target(ObjectType.RELEASE, gitids) == []
 
         # check ExtIDs have been added to the journal
         extids_in_journal = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "extid"
         ]
         assert extids == extids_in_journal
 
     def test_extid_add_hg(self, swh_storage, sample_data):
         def get_node(revision):
             node = None
             if revision.extra_headers:
                 node = dict(revision.extra_headers).get(b"node")
             if node is None and revision.metadata:
                 node = hash_to_bytes(revision.metadata.get("node"))
             return node
 
         swhids = [
             revision.id
             for revision in sample_data.revisions
             if revision.type.value == "hg"
         ]
         extids = [
             get_node(revision)
             for revision in sample_data.revisions
             if revision.type.value == "hg"
         ]
 
         assert swh_storage.extid_get_from_extid("hg", extids) == []
         assert swh_storage.extid_get_from_target(ObjectType.REVISION, swhids) == []
 
         extid_objs = [
             ExtID(
                 extid=hgid,
                 extid_type="hg",
                 extid_version=1,
                 target=CoreSWHID(object_id=swhid, object_type=ObjectType.REVISION,),
             )
             for hgid, swhid in zip(extids, swhids)
         ]
         summary = swh_storage.extid_add(extid_objs)
         assert summary == {"extid:add": len(swhids)}
 
         assert swh_storage.extid_get_from_extid("hg", extids) == extid_objs
         assert (
             swh_storage.extid_get_from_target(ObjectType.REVISION, swhids) == extid_objs
         )
 
         assert swh_storage.extid_get_from_extid("git", extids) == []
         assert swh_storage.extid_get_from_target(ObjectType.RELEASE, swhids) == []
 
         # check ExtIDs have been added to the journal
         extids_in_journal = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "extid"
         ]
         assert extid_objs == extids_in_journal
 
     def test_extid_add_twice(self, swh_storage, sample_data):
 
         gitids = [
             revision.id
             for revision in sample_data.revisions
             if revision.type.value == "git"
         ]
 
         extids = [
             ExtID(
                 extid=gitid,
                 extid_type="git",
                 target=CoreSWHID(object_id=gitid, object_type=ObjectType.REVISION,),
             )
             for gitid in gitids
         ]
         summary = swh_storage.extid_add(extids)
         assert summary == {"extid:add": len(gitids)}
 
         # add them again, should be noop
         summary = swh_storage.extid_add(extids)
         # assert summary == {"extid:add": 0}
         assert swh_storage.extid_get_from_extid("git", gitids) == extids
         assert swh_storage.extid_get_from_target(ObjectType.REVISION, gitids) == extids
 
     def test_extid_add_extid_multicity(self, swh_storage, sample_data):
 
         ids = [
             revision.id
             for revision in sample_data.revisions
             if revision.type.value == "git"
         ]
 
         extids = [
             ExtID(
                 extid=extid,
                 extid_type="git",
                 extid_version=2,
                 target=CoreSWHID(object_id=extid, object_type=ObjectType.REVISION,),
             )
             for extid in ids
         ]
         swh_storage.extid_add(extids)
 
         # try to add "modified-extid" versions, should be added
         extids2 = [
             ExtID(
                 extid=extid,
                 extid_type="hg",
                 extid_version=2,
                 target=CoreSWHID(object_id=extid, object_type=ObjectType.REVISION,),
             )
             for extid in ids
         ]
         swh_storage.extid_add(extids2)
 
         assert swh_storage.extid_get_from_extid("git", ids) == extids
         assert swh_storage.extid_get_from_extid("hg", ids) == extids2
         assert set(swh_storage.extid_get_from_target(ObjectType.REVISION, ids)) == {
             *extids,
             *extids2,
         }
 
     def test_extid_add_target_multicity(self, swh_storage, sample_data):
 
         ids = [
             revision.id
             for revision in sample_data.revisions
             if revision.type.value == "git"
         ]
 
         extids = [
             ExtID(
                 extid=extid,
                 extid_type="git",
                 target=CoreSWHID(object_id=extid, object_type=ObjectType.REVISION,),
             )
             for extid in ids
         ]
         swh_storage.extid_add(extids)
 
         # try to add "modified" versions, should be added
         extids2 = [
             ExtID(
                 extid=extid,
                 extid_type="git",
                 target=CoreSWHID(object_id=extid, object_type=ObjectType.RELEASE,),
             )
             for extid in ids
         ]
         swh_storage.extid_add(extids2)
 
         assert set(swh_storage.extid_get_from_extid("git", ids)) == {*extids, *extids2}
         assert swh_storage.extid_get_from_target(ObjectType.REVISION, ids) == extids
         assert swh_storage.extid_get_from_target(ObjectType.RELEASE, ids) == extids2
 
     def test_extid_version_behavior(self, swh_storage, sample_data):
         ids = [
             revision.id
             for revision in sample_data.revisions
             if revision.type.value == "git"
         ]
 
         # Insert extids with several different versions
         extids = [
             ExtID(
                 extid=extid,
                 extid_type="git",
                 target=CoreSWHID(object_id=extid, object_type=ObjectType.REVISION,),
             )
             for extid in ids
         ] + [
             ExtID(
                 extid=extid,
                 extid_type="git",
                 extid_version=1,
                 target=CoreSWHID(object_id=extid, object_type=ObjectType.REVISION,),
             )
             for extid in ids
         ]
         swh_storage.extid_add(extids)
 
         # Check that both versions get returned
         for git_id in ids:
             objs = swh_storage.extid_get_from_extid("git", [git_id])
             assert len(objs) == 2
             assert set(obj.extid_version for obj in objs) == {0, 1}
         for swhid in ids:
             objs = swh_storage.extid_get_from_target(ObjectType.REVISION, [swhid])
             assert len(objs) == 2
             assert set(obj.extid_version for obj in objs) == {0, 1}
 
     def test_release_add(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         init_missing = swh_storage.release_missing([release.id, release2.id])
         assert list(init_missing) == [release.id, release2.id]
 
         actual_result = swh_storage.release_add([release, release2])
         assert actual_result == {"release:add": 2}
 
         end_missing = swh_storage.release_missing([release.id, release2.id])
         assert list(end_missing) == []
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release),
             ("release", release2),
         ]
 
         # already present so nothing added
         actual_result = swh_storage.release_add([release, release2])
         assert actual_result == {"release:add": 0}
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["release"] == 2
 
     def test_release_add_no_author_date(self, swh_storage, sample_data):
         full_release = sample_data.release
 
         release = attr.evolve(full_release, author=None, date=None)
         actual_result = swh_storage.release_add([release])
         assert actual_result == {"release:add": 1}
 
         end_missing = swh_storage.release_missing([release.id])
         assert list(end_missing) == []
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release)
         ]
 
     def test_release_add_twice(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         actual_result = swh_storage.release_add([release])
         assert actual_result == {"release:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release)
         ]
 
         actual_result = swh_storage.release_add([release, release2, release, release2])
         assert actual_result == {"release:add": 1}
 
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("release", release), ("release", release2),]
         )
 
     def test_release_add_name_clash(self, swh_storage, sample_data):
         release, release2 = [
             attr.evolve(
                 c,
                 author=Person(
                     fullname=b"John Doe <john.doe@example.com>",
                     name=b"John Doe",
                     email=b"john.doe@example.com",
                 ),
             )
             for c in sample_data.releases[:2]
         ]
 
         actual_result = swh_storage.release_add([release, release2])
         assert actual_result == {"release:add": 2}
 
     def test_release_get(self, swh_storage, sample_data):
         release, release2, release3 = sample_data.releases[:3]
 
         # given
         swh_storage.release_add([release, release2])
 
         # when
         actual_releases = swh_storage.release_get([release.id, release2.id])
 
         # then
         assert actual_releases == [release, release2]
 
         unknown_releases = swh_storage.release_get([release3.id])
         assert unknown_releases[0] is None
 
     def test_release_get_order(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         add_result = swh_storage.release_add([release, release2])
         assert add_result == {"release:add": 2}
 
         # order 1
         actual_releases = swh_storage.release_get([release.id, release2.id])
         assert actual_releases == [release, release2]
 
         # order 2
         actual_releases2 = swh_storage.release_get([release2.id, release.id])
         assert actual_releases2 == [release2, release]
 
     def test_release_get_random(self, swh_storage, sample_data):
         release, release2, release3 = sample_data.releases[:3]
 
         swh_storage.release_add([release, release2, release3])
 
         assert swh_storage.release_get_random() in {
             release.id,
             release2.id,
             release3.id,
         }
 
     def test_origin_add(self, swh_storage, sample_data):
         origins = list(sample_data.origins)
         origin_urls = [o.url for o in origins]
 
         assert swh_storage.origin_get(origin_urls) == [None] * len(origins)
 
         stats = swh_storage.origin_add(origins)
         assert stats == {"origin:add": len(origin_urls)}
 
         actual_origins = swh_storage.origin_get(origin_urls)
         assert actual_origins == origins
 
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin) for origin in origins]
         )
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["origin"] == len(origins)
 
     def test_origin_add_twice(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         add1 = swh_storage.origin_add([origin, origin2])
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
         assert add1 == {"origin:add": 2}
 
         add2 = swh_storage.origin_add([origin, origin2])
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
         assert add2 == {"origin:add": 0}
 
     def test_origin_add_twice_at_once(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         add1 = swh_storage.origin_add([origin, origin2, origin, origin2])
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
         assert add1 == {"origin:add": 2}
 
         add2 = swh_storage.origin_add([origin, origin2, origin, origin2])
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
         assert add2 == {"origin:add": 0}
 
     def test_origin_get(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         assert swh_storage.origin_get([origin.url]) == [None]
         swh_storage.origin_add([origin])
 
         actual_origins = swh_storage.origin_get([origin.url])
         assert actual_origins == [origin]
 
         actual_origins = swh_storage.origin_get([origin.url, "not://exists"])
         assert actual_origins == [origin, None]
 
     def _generate_random_visits(self, nb_visits=100, start=0, end=7):
         """Generate random visits within the last 2 months (to avoid
         computations)
 
         """
         visits = []
         today = now()
         for weeks in range(nb_visits, 0, -1):
             hours = random.randint(0, 24)
             minutes = random.randint(0, 60)
             seconds = random.randint(0, 60)
             days = random.randint(0, 28)
             weeks = random.randint(start, end)
             date_visit = today - timedelta(
                 weeks=weeks, hours=hours, minutes=minutes, seconds=seconds, days=days
             )
             visits.append(date_visit)
         return visits
 
     def test_origin_visit_get__unknown_origin(self, swh_storage):
         actual_page = swh_storage.origin_visit_get("foo")
         assert actual_page.next_page_token is None
         assert actual_page.results == []
         assert actual_page == PagedResult()
 
     def test_origin_visit_get__validation_failure(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         with pytest.raises(
             StorageArgumentException, match="page_token must be a string"
         ):
             swh_storage.origin_visit_get(origin.url, page_token=10)  # not bytes
 
         with pytest.raises(
             StorageArgumentException, match="order must be a ListOrder value"
         ):
             swh_storage.origin_visit_get(origin.url, order="foobar")  # wrong order
 
     def test_origin_visit_get_all(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         ov1, ov2, ov3 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
             ]
         )
 
         # order asc, no token, no limit
         actual_page = swh_storage.origin_visit_get(origin.url)
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov1, ov2, ov3]
 
         # order asc, no token, limit
         actual_page = swh_storage.origin_visit_get(origin.url, limit=2)
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ov1, ov2]
 
         # order asc, token, no limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov3]
 
         # order asc, no token, limit
         actual_page = swh_storage.origin_visit_get(origin.url, limit=1)
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ov1]
 
         # order asc, token, no limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov2, ov3]
 
         # order asc, token, limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, limit=2
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov2, ov3]
 
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, limit=1
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ov2]
 
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, limit=1
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov3]
 
         # order desc, no token, no limit
         actual_page = swh_storage.origin_visit_get(origin.url, order=ListOrder.DESC)
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov3, ov2, ov1]
 
         # order desc, no token, limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, limit=2, order=ListOrder.DESC
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ov3, ov2]
 
         # order desc, token, no limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov1]
 
         # order desc, no token, limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, limit=1, order=ListOrder.DESC
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ov3]
 
         # order desc, token, no limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov2, ov1]
 
         # order desc, token, limit
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, order=ListOrder.DESC, limit=1
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ov2]
 
         actual_page = swh_storage.origin_visit_get(
             origin.url, page_token=next_page_token, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ov1]
 
     def test_origin_visit_status_get__unknown_cases(self, swh_storage, sample_data):
         origin = sample_data.origin
         actual_page = swh_storage.origin_visit_status_get("foobar", 1)
         assert actual_page.next_page_token is None
         assert actual_page.results == []
 
         actual_page = swh_storage.origin_visit_status_get(origin.url, 1)
         assert actual_page.next_page_token is None
         assert actual_page.results == []
 
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
             ]
         )[0]
         actual_page = swh_storage.origin_visit_status_get(origin.url, ov1.visit + 10)
         assert actual_page.next_page_token is None
         assert actual_page.results == []
 
     def test_origin_visit_status_add_unknown_type(self, swh_storage, sample_data):
         ov = OriginVisit(
             origin=sample_data.origin.url,
             date=now(),
             type=sample_data.type_visit1,
             visit=42,
         )
         ovs = OriginVisitStatus(
             origin=ov.origin,
             visit=ov.visit,
             date=now(),
             status="created",
             snapshot=None,
         )
 
         with pytest.raises(StorageArgumentException):
             swh_storage.origin_visit_status_add([ovs])
 
         swh_storage.origin_add([sample_data.origin])
 
         with pytest.raises(StorageArgumentException):
             swh_storage.origin_visit_status_add([ovs])
 
         swh_storage.origin_visit_add([ov])
 
         swh_storage.origin_visit_status_add([ovs])
 
     def test_origin_visit_status_get_all(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         date_visit3 = round_to_milliseconds(now())
         date_visit1 = date_visit3 - datetime.timedelta(hours=2)
         date_visit2 = date_visit3 - datetime.timedelta(hours=1)
         assert date_visit1 < date_visit2 < date_visit3
 
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url, date=date_visit1, type=sample_data.type_visit1,
                 ),
             ]
         )[0]
 
         ovs1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit1,
             type=ov1.type,
             status="created",
             snapshot=None,
         )
 
         ovs2 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit2,
             type=ov1.type,
             status="partial",
             snapshot=None,
         )
 
         ovs3 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit3,
             type=ov1.type,
             status="full",
             snapshot=sample_data.snapshot.id,
             metadata={},
         )
 
         swh_storage.origin_visit_status_add([ovs2, ovs3])
 
         # order asc, no token, no limit
         actual_page = swh_storage.origin_visit_status_get(origin.url, ov1.visit)
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs1, ovs2, ovs3]
 
         # order asc, no token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, limit=2
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ovs1, ovs2]
 
         # order asc, token, no limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs3]
 
         # order asc, no token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, limit=1
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ovs1]
 
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs2, ovs3]
 
         # order asc, token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token, limit=2
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs2, ovs3]
 
         # order asc, no token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, limit=2
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ovs1, ovs2]
 
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token, limit=1
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs3]
 
         # order desc, no token, no limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs3, ovs2, ovs1]
 
         # order desc, no token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, limit=2, order=ListOrder.DESC
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ovs3, ovs2]
 
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs1]
 
         # order desc, no token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, order=ListOrder.DESC, limit=1
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ovs3]
 
         # order desc, token, no limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs2, ovs1]
 
         # order desc, token, limit
         actual_page = swh_storage.origin_visit_status_get(
             origin.url,
             ov1.visit,
             page_token=next_page_token,
             order=ListOrder.DESC,
             limit=1,
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [ovs2]
 
         actual_page = swh_storage.origin_visit_status_get(
             origin.url, ov1.visit, page_token=next_page_token, order=ListOrder.DESC
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [ovs1]
 
     def test_origin_visit_status_get_random(self, swh_storage, sample_data):
         origins = sample_data.origins[:2]
         swh_storage.origin_add(origins)
 
         # Add some random visits within the selection range
         visits = self._generate_random_visits()
         visit_type = "git"
 
         # Add visits to those origins
         for origin in origins:
             for date_visit in visits:
                 visit = swh_storage.origin_visit_add(
                     [OriginVisit(origin=origin.url, date=date_visit, type=visit_type,)]
                 )[0]
                 swh_storage.origin_visit_status_add(
                     [
                         OriginVisitStatus(
                             origin=origin.url,
                             visit=visit.visit,
                             date=now(),
                             status="full",
                             snapshot=None,
                         )
                     ]
                 )
 
         swh_storage.refresh_stat_counters()
 
         stats = swh_storage.stat_counters()
         assert stats["origin"] == len(origins)
         assert stats["origin_visit"] == len(origins) * len(visits)
 
         random_ovs = swh_storage.origin_visit_status_get_random(visit_type)
         assert random_ovs
         assert random_ovs.origin is not None
         assert random_ovs.origin in [o.url for o in origins]
         assert random_ovs.type is not None
 
     def test_origin_visit_status_get_random_nothing_found(
         self, swh_storage, sample_data
     ):
         origins = sample_data.origins
         swh_storage.origin_add(origins)
         visit_type = "hg"
         # Add some visits outside of the random generation selection so nothing
         # will be found by the random selection
         visits = self._generate_random_visits(nb_visits=3, start=13, end=24)
         for origin in origins:
             for date_visit in visits:
                 visit = swh_storage.origin_visit_add(
                     [OriginVisit(origin=origin.url, date=date_visit, type=visit_type,)]
                 )[0]
                 swh_storage.origin_visit_status_add(
                     [
                         OriginVisitStatus(
                             origin=origin.url,
                             visit=visit.visit,
                             date=now(),
                             status="full",
                             snapshot=None,
                         )
                     ]
                 )
 
         random_origin_visit = swh_storage.origin_visit_status_get_random(visit_type)
         assert random_origin_visit is None
 
     def test_origin_get_by_sha1(self, swh_storage, sample_data):
         origin = sample_data.origin
         assert swh_storage.origin_get([origin.url])[0] is None
         swh_storage.origin_add([origin])
 
         origins = list(swh_storage.origin_get_by_sha1([sha1(origin.url)]))
         assert len(origins) == 1
         assert origins[0]["url"] == origin.url
 
     def test_origin_get_by_sha1_not_found(self, swh_storage, sample_data):
         unknown_origin = sample_data.origin
         assert swh_storage.origin_get([unknown_origin.url])[0] is None
         origins = list(swh_storage.origin_get_by_sha1([sha1(unknown_origin.url)]))
         assert len(origins) == 1
         assert origins[0] is None
 
     def test_origin_search_single_result(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         actual_page = swh_storage.origin_search(origin.url)
         assert actual_page.next_page_token is None
         assert actual_page.results == []
 
         actual_page = swh_storage.origin_search(origin.url, regexp=True)
         assert actual_page.next_page_token is None
         assert actual_page.results == []
 
         swh_storage.origin_add([origin])
         actual_page = swh_storage.origin_search(origin.url)
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin]
 
         actual_page = swh_storage.origin_search(f".{origin.url[1:-1]}.", regexp=True)
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin]
 
         swh_storage.origin_add([origin2])
         actual_page = swh_storage.origin_search(origin2.url)
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin2]
 
         actual_page = swh_storage.origin_search(f".{origin2.url[1:-1]}.", regexp=True)
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin2]
 
     def test_origin_search_no_regexp(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         swh_storage.origin_add([origin, origin2])
 
         # no pagination
         actual_page = swh_storage.origin_search("/")
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin, origin2]
 
         # offset=0
         actual_page = swh_storage.origin_search("/", page_token=None, limit=1)
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [origin]
 
         # offset=1
         actual_page = swh_storage.origin_search(
             "/", page_token=next_page_token, limit=1
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin2]
 
     def test_origin_search_regexp_substring(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         swh_storage.origin_add([origin, origin2])
 
         # no pagination
         actual_page = swh_storage.origin_search("/", regexp=True)
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin, origin2]
 
         # offset=0
         actual_page = swh_storage.origin_search(
             "/", page_token=None, limit=1, regexp=True
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [origin]
 
         # offset=1
         actual_page = swh_storage.origin_search(
             "/", page_token=next_page_token, limit=1, regexp=True
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin2]
 
     def test_origin_search_regexp_fullstring(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         swh_storage.origin_add([origin, origin2])
 
         # no pagination
         actual_page = swh_storage.origin_search(".*/.*", regexp=True)
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin, origin2]
 
         # offset=0
         actual_page = swh_storage.origin_search(
             ".*/.*", page_token=None, limit=1, regexp=True
         )
         next_page_token = actual_page.next_page_token
         assert next_page_token is not None
         assert actual_page.results == [origin]
 
         # offset=1
         actual_page = swh_storage.origin_search(
             ".*/.*", page_token=next_page_token, limit=1, regexp=True
         )
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin2]
 
     def test_origin_search_no_visit_types(self, swh_storage, sample_data):
         origin = sample_data.origins[0]
         swh_storage.origin_add([origin])
         actual_page = swh_storage.origin_search(origin.url, visit_types=["git"])
         assert actual_page.next_page_token is None
         assert actual_page.results == []
 
     def test_origin_search_with_visit_types(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         swh_storage.origin_add([origin, origin2])
         swh_storage.origin_visit_add(
             [
                 OriginVisit(origin=origin.url, date=now(), type="git"),
                 OriginVisit(origin=origin2.url, date=now(), type="svn"),
             ]
         )
         actual_page = swh_storage.origin_search(origin.url, visit_types=["git"])
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin]
 
         actual_page = swh_storage.origin_search(origin2.url, visit_types=["svn"])
         assert actual_page.next_page_token is None
         assert actual_page.results == [origin2]
 
     def test_origin_search_multiple_visit_types(self, swh_storage, sample_data):
         origin = sample_data.origins[0]
         swh_storage.origin_add([origin])
 
         def _add_visit_type(visit_type):
             swh_storage.origin_visit_add(
                 [OriginVisit(origin=origin.url, date=now(), type=visit_type)]
             )
 
         def _check_visit_types(visit_types):
             actual_page = swh_storage.origin_search(origin.url, visit_types=visit_types)
             assert actual_page.next_page_token is None
             assert actual_page.results == [origin]
 
         _add_visit_type("git")
         _check_visit_types(["git"])
         _check_visit_types(["git", "hg"])
 
         _add_visit_type("hg")
         _check_visit_types(["hg"])
         _check_visit_types(["git", "hg"])
 
     def test_origin_visit_add(self, swh_storage, sample_data):
         origin1 = sample_data.origins[1]
         swh_storage.origin_add([origin1])
 
         date_visit = now()
         date_visit2 = date_visit + datetime.timedelta(minutes=1)
 
         date_visit = round_to_milliseconds(date_visit)
         date_visit2 = round_to_milliseconds(date_visit2)
 
         visit1 = OriginVisit(
             origin=origin1.url, date=date_visit, type=sample_data.type_visit1,
         )
         visit2 = OriginVisit(
             origin=origin1.url, date=date_visit2, type=sample_data.type_visit2,
         )
 
         # add once
         ov1, ov2 = swh_storage.origin_visit_add([visit1, visit2])
         # then again (will be ignored as they already exist)
         origin_visit1, origin_visit2 = swh_storage.origin_visit_add([ov1, ov2])
         assert ov1 == origin_visit1
         assert ov2 == origin_visit2
 
+        assert ov1.visit == 1
+        assert ov2.visit == 2
+
         ovs1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit,
             type=ov1.type,
             status="created",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=date_visit2,
             type=ov2.type,
             status="created",
             snapshot=None,
         )
 
         actual_visits = swh_storage.origin_visit_get(origin1.url).results
         expected_visits = [ov1, ov2]
         assert len(expected_visits) == len(actual_visits)
         for visit in expected_visits:
             assert visit in actual_visits
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = list(
             [("origin", origin1)]
             + [("origin_visit", visit) for visit in expected_visits] * 2
             + [("origin_visit_status", ovs) for ovs in [ovs1, ovs2]]
         )
 
         for obj in expected_objects:
             assert obj in actual_objects
 
+    def test_origin_visit_add_replayed(self, swh_storage, sample_data):
+        """Tests adding a visit with an id makes sure the next id is higher"""
+        origin1 = sample_data.origins[1]
+        swh_storage.origin_add([origin1])
+
+        date_visit = now()
+        date_visit2 = date_visit + datetime.timedelta(minutes=1)
+
+        date_visit = round_to_milliseconds(date_visit)
+        date_visit2 = round_to_milliseconds(date_visit2)
+
+        visit1 = OriginVisit(
+            origin=origin1.url, date=date_visit, type=sample_data.type_visit1, visit=42
+        )
+        visit2 = OriginVisit(
+            origin=origin1.url, date=date_visit2, type=sample_data.type_visit2,
+        )
+
+        # add once
+        ov1, ov2 = swh_storage.origin_visit_add([visit1, visit2])
+        # then again (will be ignored as they already exist)
+        origin_visit1, origin_visit2 = swh_storage.origin_visit_add([ov1, ov2])
+        assert ov1 == origin_visit1
+        assert ov2 == origin_visit2
+
+        assert ov1.visit == 42
+        assert ov2.visit == 43
+
+        visit3 = OriginVisit(
+            origin=origin1.url, date=date_visit, type=sample_data.type_visit1, visit=12
+        )
+        visit4 = OriginVisit(
+            origin=origin1.url, date=date_visit2, type=sample_data.type_visit2,
+        )
+
+        # add once
+        ov3, ov4 = swh_storage.origin_visit_add([visit3, visit4])
+        # then again (will be ignored as they already exist)
+        origin_visit3, origin_visit4 = swh_storage.origin_visit_add([ov3, ov4])
+        assert ov3 == origin_visit3
+        assert ov4 == origin_visit4
+
+        assert ov3.visit == 12
+        assert ov4.visit == 44
+
     def test_origin_visit_add_validation(self, swh_storage, sample_data):
         """Unknown origin when adding visits should raise"""
         visit = attr.evolve(sample_data.origin_visit, origin="something-unknonw")
         with pytest.raises(StorageArgumentException, match="Unknown origin"):
             swh_storage.origin_visit_add([visit])
 
         objects = list(swh_storage.journal_writer.journal.objects)
         assert not objects
 
     def test_origin_visit_status_add_validation(self, swh_storage):
         """Wrong origin_visit_status input should raise storage argument error"""
         date_visit = now()
         visit_status1 = OriginVisitStatus(
             origin="unknown-origin-url",
             visit=10,
             date=date_visit,
             status="full",
             snapshot=None,
         )
         with pytest.raises(StorageArgumentException, match="Unknown origin"):
             swh_storage.origin_visit_status_add([visit_status1])
 
         objects = list(swh_storage.journal_writer.journal.objects)
         assert not objects
 
     def test_origin_visit_status_add(self, swh_storage, sample_data):
         """Correct origin visit statuses should add a new visit status
 
         """
         snapshot = sample_data.snapshot
         origin1 = sample_data.origins[1]
         origin2 = Origin(url="new-origin")
         swh_storage.origin_add([origin1, origin2])
 
         ov1, ov2 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
                 OriginVisit(
                     origin=origin2.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
             ]
         )
 
         ovs1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=sample_data.date_visit1,
             type=ov1.type,
             status="created",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=sample_data.date_visit2,
             type=ov2.type,
             status="created",
             snapshot=None,
         )
 
         date_visit_now = round_to_milliseconds(now())
         visit_status1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit_now,
             type=ov1.type,
             status="full",
             snapshot=snapshot.id,
         )
 
         date_visit_now = round_to_milliseconds(now())
         visit_status2 = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=date_visit_now,
             type=ov2.type,
             status="ongoing",
             snapshot=None,
             metadata={"intrinsic": "something"},
         )
         stats = swh_storage.origin_visit_status_add([visit_status1, visit_status2])
         assert stats == {"origin_visit_status:add": 2}
 
         visit = swh_storage.origin_visit_get_latest(origin1.url, require_snapshot=True)
         visit_status = swh_storage.origin_visit_status_get_latest(
             origin1.url, visit.visit, require_snapshot=True
         )
         assert visit_status == visit_status1
 
         visit = swh_storage.origin_visit_get_latest(origin2.url, require_snapshot=False)
         visit_status = swh_storage.origin_visit_status_get_latest(
             origin2.url, visit.visit, require_snapshot=False
         )
         assert origin2.url != origin1.url
         assert visit_status == visit_status2
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
 
         expected_origins = [origin1, origin2]
         expected_visits = [ov1, ov2]
         expected_visit_statuses = [ovs1, ovs2, visit_status1, visit_status2]
 
         expected_objects = (
             [("origin", o) for o in expected_origins]
             + [("origin_visit", v) for v in expected_visits]
             + [("origin_visit_status", ovs) for ovs in expected_visit_statuses]
         )
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_origin_visit_status_add_twice(self, swh_storage, sample_data):
         """Correct origin visit statuses should add a new visit status
 
         """
         snapshot = sample_data.snapshot
         origin1 = sample_data.origins[1]
         swh_storage.origin_add([origin1])
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
             ]
         )[0]
 
         ovs1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=sample_data.date_visit1,
             type=ov1.type,
             status="created",
             snapshot=None,
         )
         date_visit_now = round_to_milliseconds(now())
         visit_status1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit_now,
             type=ov1.type,
             status="full",
             snapshot=snapshot.id,
         )
 
         stats = swh_storage.origin_visit_status_add([visit_status1])
         assert stats == {"origin_visit_status:add": 1}
         # second call will ignore existing entries (will send to storage though)
         stats = swh_storage.origin_visit_status_add([visit_status1])
         # ...so the storage still returns it as an addition
         assert stats == {"origin_visit_status:add": 1}
 
         visit_status = swh_storage.origin_visit_status_get_latest(ov1.origin, ov1.visit)
         assert visit_status == visit_status1
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
 
         expected_origins = [origin1]
         expected_visits = [ov1]
         expected_visit_statuses = [ovs1, visit_status1, visit_status1]
 
         # write twice in the journal
         expected_objects = (
             [("origin", o) for o in expected_origins]
             + [("origin_visit", v) for v in expected_visits]
             + [("origin_visit_status", ovs) for ovs in expected_visit_statuses]
         )
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_origin_visit_find_by_date(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit1,
         )
         visit2 = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit3,
             type=sample_data.type_visit2,
         )
         visit3 = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit3,
         )
         ov1, ov2, ov3 = swh_storage.origin_visit_add([visit1, visit2, visit3])
 
         ovs1 = OriginVisitStatus(
             origin=origin.url,
             visit=ov1.visit,
             date=sample_data.date_visit2,
             status="ongoing",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=origin.url,
             visit=ov2.visit,
             date=sample_data.date_visit3,
             status="ongoing",
             snapshot=None,
         )
         ovs3 = OriginVisitStatus(
             origin=origin.url,
             visit=ov3.visit,
             date=sample_data.date_visit2,
             status="ongoing",
             snapshot=None,
         )
         swh_storage.origin_visit_status_add([ovs1, ovs2, ovs3])
 
         # Simple case
         actual_visit = swh_storage.origin_visit_find_by_date(
             origin.url, sample_data.date_visit3
         )
         assert actual_visit == ov2
 
         # There are two visits at the same date, the latest must be returned
         actual_visit = swh_storage.origin_visit_find_by_date(
             origin.url, sample_data.date_visit2
         )
         assert actual_visit == ov3
 
     def test_origin_visit_find_by_date__unknown_origin(self, swh_storage, sample_data):
         actual_visit = swh_storage.origin_visit_find_by_date(
             "foo", sample_data.date_visit2
         )
         assert actual_visit is None
 
     def test_origin_visit_get_by(self, swh_storage, sample_data):
         snapshot = sample_data.snapshot
         origins = sample_data.origins[:2]
         swh_storage.origin_add(origins)
         origin_url, origin_url2 = [o.url for o in origins]
 
         visit = OriginVisit(
             origin=origin_url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit2,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin_url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         # Add some other {origin, visit} entries
         visit2 = OriginVisit(
             origin=origin_url,
             date=sample_data.date_visit3,
             type=sample_data.type_visit3,
         )
         visit3 = OriginVisit(
             origin=origin_url2,
             date=sample_data.date_visit3,
             type=sample_data.type_visit3,
         )
         swh_storage.origin_visit_add([visit2, visit3])
 
         # when
         visit1_metadata = {
             "contents": 42,
             "directories": 22,
         }
 
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin_url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="full",
                     snapshot=snapshot.id,
                     metadata=visit1_metadata,
                 )
             ]
         )
 
         actual_visit = swh_storage.origin_visit_get_by(origin_url, origin_visit1.visit)
         assert actual_visit == origin_visit1
 
     def test_origin_visit_get_by__no_result(self, swh_storage, sample_data):
         actual_visit = swh_storage.origin_visit_get_by("unknown", 10)  # unknown origin
         assert actual_visit is None
 
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         actual_visit = swh_storage.origin_visit_get_by(origin.url, 999)  # unknown visit
         assert actual_visit is None
 
     def test_origin_visit_get_latest_edge_cases(self, swh_storage, sample_data):
         # unknown origin so no result
         assert swh_storage.origin_visit_get_latest("unknown-origin") is None
 
         # unknown type so no result
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         assert swh_storage.origin_visit_get_latest(origin.url, type="unknown") is None
 
         # unknown allowed statuses should raise
         with pytest.raises(StorageArgumentException, match="Unknown allowed statuses"):
             swh_storage.origin_visit_get_latest(
                 origin.url, allowed_statuses=["unknown"]
             )
 
     def test_origin_visit_get_latest_filter_type(self, swh_storage, sample_data):
         """Filtering origin visit get latest with filter type should be ok
 
         """
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="git",
         )
         visit2 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit2, type="hg",
         )
         date_now = round_to_milliseconds(now())
         visit3 = OriginVisit(origin=origin.url, date=date_now, type="hg",)
         assert sample_data.date_visit1 < sample_data.date_visit2
         assert sample_data.date_visit2 < date_now
 
         ov1, ov2, ov3 = swh_storage.origin_visit_add([visit1, visit2, visit3])
 
         # Check type filter is ok
         actual_visit = swh_storage.origin_visit_get_latest(origin.url, type="git")
         assert actual_visit == ov1
         actual_visit = swh_storage.origin_visit_get_latest(origin.url, type="hg")
         assert actual_visit == ov3
         actual_visit_unknown_type = swh_storage.origin_visit_get_latest(
             origin.url, type="npm",  # no visit matching that type
         )
         assert actual_visit_unknown_type is None
 
     def test_origin_visit_get_latest(self, swh_storage, sample_data):
         empty_snapshot, complete_snapshot = sample_data.snapshots[1:3]
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="git",
         )
         visit2 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit2, type="hg",
         )
         date_now = round_to_milliseconds(now())
         visit3 = OriginVisit(origin=origin.url, date=date_now, type="hg",)
         assert visit1.date < visit2.date
         assert visit2.date < visit3.date
 
         ov1, ov2, ov3 = swh_storage.origin_visit_add([visit1, visit2, visit3])
 
         # no filters, latest visit is the last one (whose date is most recent)
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # 3 visits, none has snapshot so nothing is returned
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         assert actual_visit is None
 
         # visit are created with "created" status, so nothing will get returned
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["partial"]
         )
         assert actual_visit is None
 
         # visit are created with "created" status, so most recent again
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["created"]
         )
         assert actual_visit == ov3
 
         # Add snapshot to visit1; require_snapshot=True makes it return first visit
         swh_storage.snapshot_add([complete_snapshot])
         visit_status_with_snapshot = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=round_to_milliseconds(now()),
             type=ov1.type,
             status="ongoing",
             snapshot=complete_snapshot.id,
         )
         swh_storage.origin_visit_status_add([visit_status_with_snapshot])
         # only the first visit has a snapshot now
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         assert actual_visit == ov1
 
         # only the first visit has a status ongoing now
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["ongoing"]
         )
         assert actual_visit == ov1
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov1.visit, require_snapshot=True
         )
         assert actual_visit_status == visit_status_with_snapshot
 
         # ... and require_snapshot=False (defaults) still returns latest visit (3rd)
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=False
         )
         assert actual_visit == ov3
         # no specific filter, this returns as before the latest visit
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # Status filter: all three visits are status=ongoing, so no visit
         # returned
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"]
         )
         assert actual_visit is None
 
         visit_status1_full = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=round_to_milliseconds(now()),
             type=ov1.type,
             status="full",
             snapshot=complete_snapshot.id,
         )
         # Mark the first visit as completed and check status filter again
         swh_storage.origin_visit_status_add([visit_status1_full])
 
         # only the first visit has the full status
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"]
         )
         assert actual_visit == ov1
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov1.visit, allowed_statuses=["full"]
         )
         assert actual_visit_status == visit_status1_full
 
         # no specific filter, this returns as before the latest visit
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # Add snapshot to visit2 and check that the new snapshot is returned
         swh_storage.snapshot_add([empty_snapshot])
 
         visit_status2_full = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=round_to_milliseconds(now()),
             type=ov2.type,
             status="ongoing",
             snapshot=empty_snapshot.id,
         )
         swh_storage.origin_visit_status_add([visit_status2_full])
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         # 2nd visit is most recent with a snapshot
         assert actual_visit == ov2
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov2.visit, require_snapshot=True
         )
         assert actual_visit_status == visit_status2_full
 
         # no specific filter, this returns as before the latest visit, 3rd one
         actual_origin = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_origin == ov3
 
         # full status is still the first visit
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"]
         )
         assert actual_visit == ov1
 
         # Add snapshot to visit3 (same date as visit2)
         visit_status3_with_snapshot = OriginVisitStatus(
             origin=ov3.origin,
             visit=ov3.visit,
             date=round_to_milliseconds(now()),
             type=ov3.type,
             status="ongoing",
             snapshot=complete_snapshot.id,
         )
         swh_storage.origin_visit_status_add([visit_status3_with_snapshot])
 
         # full status is still the first visit
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"], require_snapshot=True,
         )
         assert actual_visit == ov1
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url,
             visit=actual_visit.visit,
             allowed_statuses=["full"],
             require_snapshot=True,
         )
         assert actual_visit_status == visit_status1_full
 
         # most recent is still the 3rd visit
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # 3rd visit has a snapshot now, so it's elected
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         assert actual_visit == ov3
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov3.visit, require_snapshot=True
         )
         assert actual_visit_status == visit_status3_with_snapshot
 
     def test_origin_visit_get_latest__same_date(self, swh_storage, sample_data):
         empty_snapshot, complete_snapshot = sample_data.snapshots[1:3]
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="git",
         )
         visit2 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="hg",
         )
 
         ov1, ov2 = swh_storage.origin_visit_add([visit1, visit2])
 
         # ties should be broken by using the visit id
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov2
 
     def test_origin_visit_get_latest__not_last(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
 
         visit1, visit2 = sample_data.origin_visits[:2]
         assert visit1.origin == origin.url
 
         swh_storage.origin_visit_add([visit1])
         ov1 = swh_storage.origin_visit_get_latest(origin.url)
 
         # Add snapshot to visit1, latest snapshot = visit 1 snapshot
         complete_snapshot = sample_data.snapshots[2]
         swh_storage.snapshot_add([complete_snapshot])
 
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov1.visit,
                     date=visit2.date,
                     status="partial",
                     snapshot=None,
                 )
             ]
         )
         assert visit1.date < visit2.date
 
         # no snapshot associated to the visit, so None
         visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["partial"], require_snapshot=True,
         )
         assert visit is None
 
         date_now = now()
         assert visit2.date < date_now
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov1.visit,
                     date=date_now,
                     status="full",
                     snapshot=complete_snapshot.id,
                 )
             ]
         )
 
         swh_storage.origin_visit_add(
             [OriginVisit(origin=origin.url, date=now(), type=visit1.type,)]
         )
 
         visit = swh_storage.origin_visit_get_latest(origin.url, require_snapshot=True)
         assert visit is not None
 
     def test_origin_visit_status_get_latest__validation(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="git",
         )
 
         # unknown allowed statuses should raise
         with pytest.raises(StorageArgumentException, match="Unknown allowed statuses"):
             swh_storage.origin_visit_status_get_latest(
                 origin.url, visit1.visit, allowed_statuses=["unknown"]
             )
 
     def test_origin_visit_status_get_latest(self, swh_storage, sample_data):
         snapshot = sample_data.snapshots[2]
         origin1 = sample_data.origin
         swh_storage.origin_add([origin1])
 
         # to have some reference visits
 
         ov1, ov2 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
             ]
         )
         swh_storage.snapshot_add([snapshot])
 
         date_now = round_to_milliseconds(now())
         assert sample_data.date_visit1 < sample_data.date_visit2
         assert sample_data.date_visit2 < date_now
 
         ovs1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=sample_data.date_visit1,
             type=ov1.type,
             status="partial",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=sample_data.date_visit2,
             type=ov1.type,
             status="ongoing",
             snapshot=None,
         )
         ovs3 = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=sample_data.date_visit2
             + datetime.timedelta(minutes=1),  # to not be ignored
             type=ov2.type,
             status="ongoing",
             snapshot=None,
         )
         ovs4 = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=date_now,
             type=ov2.type,
             status="full",
             snapshot=snapshot.id,
             metadata={"something": "wicked"},
         )
 
         swh_storage.origin_visit_status_add([ovs1, ovs2, ovs3, ovs4])
 
         # unknown origin so no result
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             "unknown-origin", ov1.visit
         )
         assert actual_origin_visit is None
 
         # unknown visit so no result
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             ov1.origin, ov1.visit + 10
         )
         assert actual_origin_visit is None
 
         # Two visits, both with no snapshot, take the most recent
         actual_origin_visit2 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov1.visit
         )
         assert isinstance(actual_origin_visit2, OriginVisitStatus)
         assert actual_origin_visit2 == ovs2
         assert ovs2.origin == origin1.url
         assert ovs2.visit == ov1.visit
 
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov1.visit, require_snapshot=True
         )
         # there is no visit with snapshot yet for that visit
         assert actual_origin_visit is None
 
         actual_origin_visit2 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov1.visit, allowed_statuses=["partial", "ongoing"]
         )
         # visit status with partial status visit elected
         assert actual_origin_visit2 == ovs2
         assert actual_origin_visit2.status == "ongoing"
 
         actual_origin_visit4 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov2.visit, require_snapshot=True
         )
         assert actual_origin_visit4 == ovs4
         assert actual_origin_visit4.snapshot == snapshot.id
 
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov2.visit, require_snapshot=True, allowed_statuses=["ongoing"]
         )
         # nothing matches so nothing
         assert actual_origin_visit is None  # there is no visit with status full
 
         actual_origin_visit3 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov2.visit, allowed_statuses=["ongoing"]
         )
         assert actual_origin_visit3 == ovs3
 
     def test_person_fullname_unicity(self, swh_storage, sample_data):
         revision, rev2 = sample_data.revisions[0:2]
         # create a revision with same committer fullname but wo name and email
         revision2 = attr.evolve(
             rev2,
             committer=Person(
                 fullname=revision.committer.fullname, name=None, email=None
             ),
         )
 
         swh_storage.revision_add([revision, revision2])
 
         # when getting added revisions
         revisions = swh_storage.revision_get([revision.id, revision2.id])
 
         # then check committers are the same
         assert revisions[0].committer == revisions[1].committer
 
     def test_snapshot_add_get_empty(self, swh_storage, sample_data):
         empty_snapshot = sample_data.snapshots[1]
         empty_snapshot_dict = empty_snapshot.to_dict()
 
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 )
             ]
         )[0]
 
         actual_result = swh_storage.snapshot_add([empty_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         date_now = now()
 
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=ov1.origin,
                     visit=ov1.visit,
                     date=date_now,
                     type=ov1.type,
                     status="full",
                     snapshot=empty_snapshot.id,
                 )
             ]
         )
 
         by_id = swh_storage.snapshot_get(empty_snapshot.id)
         assert by_id == {**empty_snapshot_dict, "next_branch": None}
 
         ovs1 = OriginVisitStatus.from_dict(
             {
                 "origin": ov1.origin,
                 "date": sample_data.date_visit1,
                 "type": ov1.type,
                 "visit": ov1.visit,
                 "status": "created",
                 "snapshot": None,
                 "metadata": None,
             }
         )
         ovs2 = OriginVisitStatus.from_dict(
             {
                 "origin": ov1.origin,
                 "date": date_now,
                 "type": ov1.type,
                 "visit": ov1.visit,
                 "status": "full",
                 "metadata": None,
                 "snapshot": empty_snapshot.id,
             }
         )
         actual_objects = list(swh_storage.journal_writer.journal.objects)
 
         expected_objects = [
             ("origin", origin),
             ("origin_visit", ov1),
             ("origin_visit_status", ovs1,),
             ("snapshot", empty_snapshot),
             ("origin_visit_status", ovs2,),
         ]
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_snapshot_add_get_complete(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
         complete_snapshot_dict = complete_snapshot.to_dict()
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit1,
             type=sample_data.type_visit1,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=complete_snapshot.id,
                 )
             ]
         )
         assert actual_result == {"snapshot:add": 1}
 
         by_id = swh_storage.snapshot_get(complete_snapshot.id)
         assert by_id == {**complete_snapshot_dict, "next_branch": None}
 
     def test_snapshot_add_many(self, swh_storage, sample_data):
         snapshot, _, complete_snapshot = sample_data.snapshots[:3]
 
         actual_result = swh_storage.snapshot_add([snapshot, complete_snapshot])
         assert actual_result == {"snapshot:add": 2}
 
         assert swh_storage.snapshot_get(complete_snapshot.id) == {
             **complete_snapshot.to_dict(),
             "next_branch": None,
         }
 
         assert swh_storage.snapshot_get(snapshot.id) == {
             **snapshot.to_dict(),
             "next_branch": None,
         }
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["snapshot"] == 2
 
     def test_snapshot_add_many_incremental(self, swh_storage, sample_data):
         snapshot, _, complete_snapshot = sample_data.snapshots[:3]
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         actual_result2 = swh_storage.snapshot_add([snapshot, complete_snapshot])
         assert actual_result2 == {"snapshot:add": 1}
 
         assert swh_storage.snapshot_get(complete_snapshot.id) == {
             **complete_snapshot.to_dict(),
             "next_branch": None,
         }
 
         assert swh_storage.snapshot_get(snapshot.id) == {
             **snapshot.to_dict(),
             "next_branch": None,
         }
 
     def test_snapshot_add_twice(self, swh_storage, sample_data):
         snapshot, empty_snapshot = sample_data.snapshots[:2]
 
         actual_result = swh_storage.snapshot_add([empty_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("snapshot", empty_snapshot)
         ]
 
         actual_result = swh_storage.snapshot_add([snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("snapshot", empty_snapshot),
             ("snapshot", snapshot),
         ]
 
     def test_snapshot_add_count_branches(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         snp_size = swh_storage.snapshot_count_branches(complete_snapshot.id)
 
         expected_snp_size = {
             "alias": 1,
             "content": 1,
             "directory": 2,
             "release": 1,
             "revision": 1,
             "snapshot": 1,
             None: 1,
         }
         assert snp_size == expected_snp_size
 
     def test_snapshot_add_count_branches_with_filtering(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         snp_size = swh_storage.snapshot_count_branches(
             complete_snapshot.id, branch_name_exclude_prefix=b"release"
         )
 
         expected_snp_size = {
             "alias": 1,
             "content": 1,
             "directory": 2,
             "revision": 1,
             "snapshot": 1,
             None: 1,
         }
         assert snp_size == expected_snp_size
 
     def test_snapshot_add_count_branches_with_filtering_edge_cases(
         self, swh_storage, sample_data
     ):
         snapshot = Snapshot(
             branches={
                 b"\xaa\xff": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"\xaa\xff\x00": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"\xff\xff": SnapshotBranch(
                     target=sample_data.release.id, target_type=TargetType.RELEASE,
                 ),
                 b"\xff\xff\x00": SnapshotBranch(
                     target=sample_data.release.id, target_type=TargetType.RELEASE,
                 ),
                 b"dangling": None,
             },
         )
 
         swh_storage.snapshot_add([snapshot])
 
         assert swh_storage.snapshot_count_branches(
             snapshot.id, branch_name_exclude_prefix=b"\xaa\xff"
         ) == {None: 1, "release": 2}
 
         assert swh_storage.snapshot_count_branches(
             snapshot.id, branch_name_exclude_prefix=b"\xff\xff"
         ) == {None: 1, "revision": 2}
 
     def test_snapshot_add_get_paginated(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         swh_storage.snapshot_add([complete_snapshot])
 
         snp_id = complete_snapshot.id
         branches = complete_snapshot.branches
         branch_names = list(sorted(branches))
 
         # Test branch_from
         snapshot = swh_storage.snapshot_get_branches(snp_id, branches_from=b"release")
 
         rel_idx = branch_names.index(b"release")
         expected_snapshot = {
             "id": snp_id,
             "branches": {name: branches[name] for name in branch_names[rel_idx:]},
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
         # Test branches_count
         snapshot = swh_storage.snapshot_get_branches(snp_id, branches_count=1)
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {branch_names[0]: branches[branch_names[0]],},
             "next_branch": b"content",
         }
         assert snapshot == expected_snapshot
 
         # test branch_from + branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, branches_from=b"directory", branches_count=3
         )
 
         dir_idx = branch_names.index(b"directory")
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: branches[name] for name in branch_names[dir_idx : dir_idx + 3]
             },
             "next_branch": branch_names[dir_idx + 3],
         }
 
         assert snapshot == expected_snapshot
 
     def test_snapshot_add_get_filtered(self, swh_storage, sample_data):
         origin = sample_data.origin
         complete_snapshot = sample_data.snapshots[2]
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit1,
             type=sample_data.type_visit1,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([complete_snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=complete_snapshot.id,
                 )
             ]
         )
 
         snp_id = complete_snapshot.id
         branches = complete_snapshot.branches
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["release", "revision"]
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: tgt
                 for name, tgt in branches.items()
                 if tgt and tgt.target_type in [TargetType.RELEASE, TargetType.REVISION]
             },
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
         snapshot = swh_storage.snapshot_get_branches(snp_id, target_types=["alias"])
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: tgt
                 for name, tgt in branches.items()
                 if tgt and tgt.target_type == TargetType.ALIAS
             },
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
     def test_snapshot_add_get_filtered_and_paginated(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         swh_storage.snapshot_add([complete_snapshot])
 
         snp_id = complete_snapshot.id
         branches = complete_snapshot.branches
         branch_names = list(sorted(branches))
 
         # Test branch_from
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["directory", "release"], branches_from=b"directory2"
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {name: branches[name] for name in (b"directory2", b"release")},
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
         # Test branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["directory", "release"], branches_count=1
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {b"directory": branches[b"directory"]},
             "next_branch": b"directory2",
         }
         assert snapshot == expected_snapshot
 
         # Test branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["directory", "release"], branches_count=2
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: branches[name] for name in (b"directory", b"directory2")
             },
             "next_branch": b"release",
         }
         assert snapshot == expected_snapshot
 
         # test branch_from + branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id,
             target_types=["directory", "release"],
             branches_from=b"directory2",
             branches_count=1,
         )
 
         dir_idx = branch_names.index(b"directory2")
         expected_snapshot = {
             "id": snp_id,
             "branches": {branch_names[dir_idx]: branches[branch_names[dir_idx]],},
             "next_branch": b"release",
         }
 
         assert snapshot == expected_snapshot
 
     def test_snapshot_add_get_branch_by_type(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
         snapshot = complete_snapshot.to_dict()
 
         alias1 = b"alias1"
         alias2 = b"alias2"
         target1 = random.choice(list(snapshot["branches"].keys()))
         target2 = random.choice(list(snapshot["branches"].keys()))
 
         snapshot["branches"][alias2] = {
             "target": target2,
             "target_type": "alias",
         }
 
         snapshot["branches"][alias1] = {
             "target": target1,
             "target_type": "alias",
         }
 
         new_snapshot = Snapshot.from_dict(snapshot)
         swh_storage.snapshot_add([new_snapshot])
 
         branches = swh_storage.snapshot_get_branches(
             new_snapshot.id,
             target_types=["alias"],
             branches_from=alias1,
             branches_count=1,
         )["branches"]
 
         assert len(branches) == 1
         assert alias1 in branches
 
     def test_snapshot_add_get_by_branches_name_pattern(self, swh_storage, sample_data):
         snapshot = Snapshot(
             branches={
                 b"refs/heads/master": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"refs/heads/incoming": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"refs/pull/1": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"refs/pull/2": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"dangling": None,
                 b"\xaa\xff": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"\xaa\xff\x00": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"\xff\xff": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
                 b"\xff\xff\x00": SnapshotBranch(
                     target=sample_data.revision.id, target_type=TargetType.REVISION,
                 ),
             },
         )
         swh_storage.snapshot_add([snapshot])
 
         for include_pattern, exclude_prefix, nb_results in (
             (b"pull", None, 2),
             (b"incoming", None, 1),
             (b"dangling", None, 1),
             (None, b"refs/heads/", 7),
             (b"refs", b"refs/heads/master", 3),
             (b"refs", b"refs/heads/master", 3),
             (None, b"\xaa\xff", 7),
             (None, b"\xff\xff", 7),
         ):
             branches = swh_storage.snapshot_get_branches(
                 snapshot.id,
                 branch_name_include_substring=include_pattern,
                 branch_name_exclude_prefix=exclude_prefix,
             )["branches"]
 
             expected_branches = [
                 branch_name
                 for branch_name in snapshot.branches
                 if (include_pattern is None or include_pattern in branch_name)
                 and (
                     exclude_prefix is None or not branch_name.startswith(exclude_prefix)
                 )
             ]
             assert sorted(branches) == sorted(expected_branches)
             assert len(branches) == nb_results
 
     def test_snapshot_add_get_by_branches_name_pattern_filtered_paginated(
         self, swh_storage, sample_data
     ):
         pattern = b"foo"
         nb_branches_by_target_type = 10
         branches = {}
         for i in range(nb_branches_by_target_type):
             branches[f"branch/directory/bar{i}".encode()] = SnapshotBranch(
                 target=sample_data.directory.id, target_type=TargetType.DIRECTORY,
             )
             branches[f"branch/revision/bar{i}".encode()] = SnapshotBranch(
                 target=sample_data.revision.id, target_type=TargetType.REVISION,
             )
             branches[f"branch/directory/{pattern}{i}".encode()] = SnapshotBranch(
                 target=sample_data.directory.id, target_type=TargetType.DIRECTORY,
             )
             branches[f"branch/revision/{pattern}{i}".encode()] = SnapshotBranch(
                 target=sample_data.revision.id, target_type=TargetType.REVISION,
             )
 
         snapshot = Snapshot(branches=branches)
         swh_storage.snapshot_add([snapshot])
 
         branches_count = nb_branches_by_target_type // 2
 
         for target_type in (
             TargetType.DIRECTORY,
             TargetType.REVISION,
         ):
             target_type_str = target_type.value
             partial_branches = swh_storage.snapshot_get_branches(
                 snapshot.id,
                 branch_name_include_substring=pattern,
                 target_types=[target_type_str],
                 branches_count=branches_count,
             )
             branches = partial_branches["branches"]
 
             expected_branches = [
                 branch_name
                 for branch_name, branch_data in snapshot.branches.items()
                 if pattern in branch_name and branch_data.target_type == target_type
             ][:branches_count]
 
             assert sorted(branches) == sorted(expected_branches)
             assert (
                 partial_branches["next_branch"]
                 == f"branch/{target_type_str}/{pattern}{branches_count}".encode()
             )
 
             partial_branches = swh_storage.snapshot_get_branches(
                 snapshot.id,
                 branch_name_include_substring=pattern,
                 target_types=[target_type_str],
                 branches_from=partial_branches["next_branch"],
             )
             branches = partial_branches["branches"]
 
             expected_branches = [
                 branch_name
                 for branch_name, branch_data in snapshot.branches.items()
                 if pattern in branch_name and branch_data.target_type == target_type
             ][branches_count:]
 
             assert sorted(branches) == sorted(expected_branches)
             assert partial_branches["next_branch"] is None
 
     def test_snapshot_add_get(self, swh_storage, sample_data):
         snapshot = sample_data.snapshot
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit1,
             type=sample_data.type_visit1,
         )
         ov1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         expected_snapshot = {**snapshot.to_dict(), "next_branch": None}
 
         by_id = swh_storage.snapshot_get(snapshot.id)
         assert by_id == expected_snapshot
 
         actual_visit = swh_storage.origin_visit_get_by(origin.url, ov1.visit)
         assert actual_visit == ov1
 
         visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov1.visit, require_snapshot=True
         )
         assert visit_status.snapshot == snapshot.id
 
     def test_snapshot_get_random(self, swh_storage, sample_data):
         snapshot, empty_snapshot, complete_snapshot = sample_data.snapshots[:3]
         swh_storage.snapshot_add([snapshot, empty_snapshot, complete_snapshot])
 
         assert swh_storage.snapshot_get_random() in {
             snapshot.id,
             empty_snapshot.id,
             complete_snapshot.id,
         }
 
     def test_snapshot_missing(self, swh_storage, sample_data):
         snapshot, missing_snapshot = sample_data.snapshots[:2]
         snapshots = [snapshot.id, missing_snapshot.id]
         swh_storage.snapshot_add([snapshot])
 
         missing_snapshots = swh_storage.snapshot_missing(snapshots)
 
         assert list(missing_snapshots) == [missing_snapshot.id]
 
     def test_stat_counters(self, swh_storage, sample_data):
         origin = sample_data.origin
         snapshot = sample_data.snapshot
         revision = sample_data.revision
         release = sample_data.release
         directory = sample_data.directory
         content = sample_data.content
 
         expected_keys = ["content", "directory", "origin", "revision"]
 
         # Initially, all counters are 0
 
         swh_storage.refresh_stat_counters()
         counters = swh_storage.stat_counters()
         assert set(expected_keys) <= set(counters)
         for key in expected_keys:
             assert counters[key] == 0
 
         # Add a content. Only the content counter should increase.
 
         swh_storage.content_add([content])
 
         swh_storage.refresh_stat_counters()
         counters = swh_storage.stat_counters()
 
         assert set(expected_keys) <= set(counters)
         for key in expected_keys:
             if key != "content":
                 assert counters[key] == 0
         assert counters["content"] == 1
 
         # Add other objects. Check their counter increased as well.
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit2,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
         swh_storage.directory_add([directory])
         swh_storage.revision_add([revision])
         swh_storage.release_add([release])
 
         swh_storage.refresh_stat_counters()
         counters = swh_storage.stat_counters()
         assert counters["content"] == 1
         assert counters["directory"] == 1
         assert counters["snapshot"] == 1
         assert counters["origin"] == 1
         assert counters["origin_visit"] == 1
         assert counters["revision"] == 1
         assert counters["release"] == 1
         assert counters["snapshot"] == 1
         if "person" in counters:
             assert counters["person"] == 3
 
     def test_content_find_ctime(self, swh_storage, sample_data):
         origin_content = sample_data.content
         ctime = round_to_milliseconds(now())
         content = attr.evolve(origin_content, data=None, ctime=ctime)
         swh_storage.content_add_metadata([content])
 
         actually_present = swh_storage.content_find({"sha1": content.sha1})
         assert actually_present[0] == content
         assert actually_present[0].ctime is not None
         assert actually_present[0].ctime.tzinfo is not None
 
     def test_content_find_with_present_content(self, swh_storage, sample_data):
         content = sample_data.content
         expected_content = attr.evolve(content, data=None)
 
         # 1. with something to find
         swh_storage.content_add([content])
 
         actually_present = swh_storage.content_find({"sha1": content.sha1})
         assert 1 == len(actually_present)
         assert actually_present[0] == expected_content
 
         # 2. with something to find
         actually_present = swh_storage.content_find({"sha1_git": content.sha1_git})
         assert 1 == len(actually_present)
         assert actually_present[0] == expected_content
 
         # 3. with something to find
         actually_present = swh_storage.content_find({"sha256": content.sha256})
         assert 1 == len(actually_present)
         assert actually_present[0] == expected_content
 
         # 4. with something to find
         actually_present = swh_storage.content_find(content.hashes())
         assert 1 == len(actually_present)
         assert actually_present[0] == expected_content
 
     def test_content_find_with_non_present_content(self, swh_storage, sample_data):
         missing_content = sample_data.skipped_content
         # 1. with something that does not exist
         actually_present = swh_storage.content_find({"sha1": missing_content.sha1})
         assert actually_present == []
 
         # 2. with something that does not exist
         actually_present = swh_storage.content_find(
             {"sha1_git": missing_content.sha1_git}
         )
         assert actually_present == []
 
         # 3. with something that does not exist
         actually_present = swh_storage.content_find({"sha256": missing_content.sha256})
         assert actually_present == []
 
     def test_content_find_with_duplicate_input(self, swh_storage, sample_data):
         content = sample_data.content
 
         # Create fake data with colliding sha256 and blake2s256
         sha1_array = bytearray(content.sha1)
         sha1_array[0] += 1
         sha1git_array = bytearray(content.sha1_git)
         sha1git_array[0] += 1
         duplicated_content = attr.evolve(
             content, sha1=bytes(sha1_array), sha1_git=bytes(sha1git_array)
         )
 
         # Inject the data
         swh_storage.content_add([content, duplicated_content])
 
         actual_result = swh_storage.content_find(
             {
                 "blake2s256": duplicated_content.blake2s256,
                 "sha256": duplicated_content.sha256,
             }
         )
 
         expected_content = attr.evolve(content, data=None)
         expected_duplicated_content = attr.evolve(duplicated_content, data=None)
 
         for result in actual_result:
             assert result in [expected_content, expected_duplicated_content]
 
     def test_content_find_with_duplicate_sha256(self, swh_storage, sample_data):
         content = sample_data.content
 
         hashes = {}
         # Create fake data with colliding sha256
         for hashalgo in ("sha1", "sha1_git", "blake2s256"):
             value = bytearray(getattr(content, hashalgo))
             value[0] += 1
             hashes[hashalgo] = bytes(value)
 
         duplicated_content = attr.evolve(
             content,
             sha1=hashes["sha1"],
             sha1_git=hashes["sha1_git"],
             blake2s256=hashes["blake2s256"],
         )
         swh_storage.content_add([content, duplicated_content])
 
         actual_result = swh_storage.content_find({"sha256": duplicated_content.sha256})
         assert len(actual_result) == 2
 
         expected_content = attr.evolve(content, data=None)
         expected_duplicated_content = attr.evolve(duplicated_content, data=None)
 
         for result in actual_result:
             assert result in [expected_content, expected_duplicated_content]
 
         # Find with both sha256 and blake2s256
         actual_result = swh_storage.content_find(
             {
                 "sha256": duplicated_content.sha256,
                 "blake2s256": duplicated_content.blake2s256,
             }
         )
 
         assert len(actual_result) == 1
         assert actual_result == [expected_duplicated_content]
 
     def test_content_find_with_duplicate_blake2s256(self, swh_storage, sample_data):
         content = sample_data.content
 
         # Create fake data with colliding sha256 and blake2s256
         sha1_array = bytearray(content.sha1)
         sha1_array[0] += 1
         sha1git_array = bytearray(content.sha1_git)
         sha1git_array[0] += 1
         sha256_array = bytearray(content.sha256)
         sha256_array[0] += 1
 
         duplicated_content = attr.evolve(
             content,
             sha1=bytes(sha1_array),
             sha1_git=bytes(sha1git_array),
             sha256=bytes(sha256_array),
         )
 
         swh_storage.content_add([content, duplicated_content])
 
         actual_result = swh_storage.content_find(
             {"blake2s256": duplicated_content.blake2s256}
         )
 
         expected_content = attr.evolve(content, data=None)
         expected_duplicated_content = attr.evolve(duplicated_content, data=None)
 
         for result in actual_result:
             assert result in [expected_content, expected_duplicated_content]
 
         # Find with both sha256 and blake2s256
         actual_result = swh_storage.content_find(
             {
                 "sha256": duplicated_content.sha256,
                 "blake2s256": duplicated_content.blake2s256,
             }
         )
 
         assert actual_result == [expected_duplicated_content]
 
     def test_content_find_bad_input(self, swh_storage):
         # 1. with no hash to lookup
         with pytest.raises(StorageArgumentException):
             swh_storage.content_find({})  # need at least one hash
 
         # 2. with bad hash
         with pytest.raises(StorageArgumentException):
             swh_storage.content_find({"unknown-sha1": "something"})  # not the right key
 
     def test_object_find_by_sha1_git(self, swh_storage, sample_data):
         content = sample_data.content
         directory = sample_data.directory
         revision = sample_data.revision
         release = sample_data.release
 
         sha1_gits = [b"00000000000000000000"]
         expected = {
             b"00000000000000000000": [],
         }
 
         swh_storage.content_add([content])
         sha1_gits.append(content.sha1_git)
 
         expected[content.sha1_git] = [
             {"sha1_git": content.sha1_git, "type": "content",}
         ]
 
         swh_storage.directory_add([directory])
         sha1_gits.append(directory.id)
         expected[directory.id] = [{"sha1_git": directory.id, "type": "directory",}]
 
         swh_storage.revision_add([revision])
         sha1_gits.append(revision.id)
         expected[revision.id] = [{"sha1_git": revision.id, "type": "revision",}]
 
         swh_storage.release_add([release])
         sha1_gits.append(release.id)
         expected[release.id] = [{"sha1_git": release.id, "type": "release",}]
 
         ret = swh_storage.object_find_by_sha1_git(sha1_gits)
 
         assert expected == ret
 
     def test_metadata_fetcher_add_get(self, swh_storage, sample_data):
         fetcher = sample_data.metadata_fetcher
         actual_fetcher = swh_storage.metadata_fetcher_get(fetcher.name, fetcher.version)
         assert actual_fetcher is None  # does not exist
 
         swh_storage.metadata_fetcher_add([fetcher])
 
         res = swh_storage.metadata_fetcher_get(fetcher.name, fetcher.version)
         assert res == fetcher
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = [
             ("metadata_fetcher", fetcher),
         ]
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_metadata_fetcher_add_zero(self, swh_storage, sample_data):
         fetcher = sample_data.metadata_fetcher
         actual_fetcher = swh_storage.metadata_fetcher_get(fetcher.name, fetcher.version)
         assert actual_fetcher is None  # does not exist
 
         swh_storage.metadata_fetcher_add([])
 
     def test_metadata_authority_add_get(self, swh_storage, sample_data):
         authority = sample_data.metadata_authority
 
         actual_authority = swh_storage.metadata_authority_get(
             authority.type, authority.url
         )
         assert actual_authority is None  # does not exist
 
         swh_storage.metadata_authority_add([authority])
 
         res = swh_storage.metadata_authority_get(authority.type, authority.url)
         assert res == authority
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = [
             ("metadata_authority", authority),
         ]
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_metadata_authority_add_zero(self, swh_storage, sample_data):
         authority = sample_data.metadata_authority
 
         actual_authority = swh_storage.metadata_authority_get(
             authority.type, authority.url
         )
         assert actual_authority is None  # does not exist
 
         swh_storage.metadata_authority_add([])
 
     def test_content_metadata_add(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata = sample_data.content_metadata[:2]
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add(content_metadata)
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(), authority
         )
         assert result.next_page_token is None
         assert list(sorted(result.results, key=lambda x: x.discovery_date,)) == list(
             content_metadata
         )
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = [
             ("metadata_authority", authority),
             ("metadata_fetcher", fetcher),
         ] + [("raw_extrinsic_metadata", item) for item in content_metadata]
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_content_metadata_add_duplicate(self, swh_storage, sample_data):
         """Duplicates should be silently ignored."""
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
         swh_storage.raw_extrinsic_metadata_add([content_metadata2, content_metadata])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(), authority
         )
         assert result.next_page_token is None
 
         expected_results = (content_metadata, content_metadata2)
 
         assert (
             tuple(sorted(result.results, key=lambda x: x.discovery_date,))
             == expected_results
         )
 
     def test_content_metadata_get(self, swh_storage, sample_data):
         content, content2 = sample_data.contents[:2]
         fetcher, fetcher2 = sample_data.fetchers[:2]
         authority, authority2 = sample_data.authorities[:2]
         (
             content1_metadata1,
             content1_metadata2,
             content1_metadata3,
         ) = sample_data.content_metadata[:3]
 
         content2_metadata = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(content1_metadata2.to_dict(), ("id",)),  # recompute id
                 "target": str(content2.swhid()),
             }
         )
 
         swh_storage.metadata_authority_add([authority, authority2])
         swh_storage.metadata_fetcher_add([fetcher, fetcher2])
 
         swh_storage.raw_extrinsic_metadata_add(
             [
                 content1_metadata1,
                 content1_metadata2,
                 content1_metadata3,
                 content2_metadata,
             ]
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(), authority
         )
         assert result.next_page_token is None
         assert [content1_metadata1, content1_metadata2] == list(
             sorted(result.results, key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(), authority2
         )
         assert result.next_page_token is None
         assert [content1_metadata3] == list(
             sorted(result.results, key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content2.swhid().to_extended(), authority
         )
         assert result.next_page_token is None
         assert [content2_metadata] == list(result.results,)
 
     def test_content_metadata_get_after(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(),
             authority,
             after=content_metadata.discovery_date - timedelta(seconds=1),
         )
         assert result.next_page_token is None
         assert [content_metadata, content_metadata2] == list(
             sorted(result.results, key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(),
             authority,
             after=content_metadata.discovery_date,
         )
         assert result.next_page_token is None
         assert result.results == [content_metadata2]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(),
             authority,
             after=content_metadata2.discovery_date,
         )
         assert result.next_page_token is None
         assert result.results == []
 
     def test_content_metadata_get_paginate(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
         swh_storage.raw_extrinsic_metadata_get(content.swhid().to_extended(), authority)
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(), authority, limit=1
         )
         assert result.next_page_token is not None
         assert result.results == [content_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(),
             authority,
             limit=1,
             page_token=result.next_page_token,
         )
         assert result.next_page_token is None
         assert result.results == [content_metadata2]
 
     def test_content_metadata_get_paginate_same_date(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher1, fetcher2 = sample_data.fetchers[:2]
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         swh_storage.metadata_fetcher_add([fetcher1, fetcher2])
         swh_storage.metadata_authority_add([authority])
 
         new_content_metadata2 = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(content_metadata2.to_dict(), ("id",)),  # recompute id
                 "discovery_date": content_metadata2.discovery_date,
                 "fetcher": attr.evolve(fetcher2, metadata=None).to_dict(),
             }
         )
 
         swh_storage.raw_extrinsic_metadata_add(
             [content_metadata, new_content_metadata2]
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(), authority, limit=1
         )
         assert result.next_page_token is not None
         assert result.results == [content_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             content.swhid().to_extended(),
             authority,
             limit=1,
             page_token=result.next_page_token,
         )
         assert result.next_page_token is None
         assert result.results[0].to_dict() == new_content_metadata2.to_dict()
         assert result.results == [new_content_metadata2]
 
     def test_content_metadata_get_by_ids(self, swh_storage, sample_data):
         content, content2 = sample_data.contents[:2]
         fetcher, fetcher2 = sample_data.fetchers[:2]
         authority, authority2 = sample_data.authorities[:2]
         (
             content1_metadata1,
             content1_metadata2,
             content1_metadata3,
         ) = sample_data.content_metadata[:3]
 
         content2_metadata = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(content1_metadata2.to_dict(), ("id",)),  # recompute id
                 "target": str(content2.swhid()),
             }
         )
 
         swh_storage.metadata_authority_add([authority, authority2])
         swh_storage.metadata_fetcher_add([fetcher, fetcher2])
 
         swh_storage.raw_extrinsic_metadata_add(
             [
                 content1_metadata1,
                 content1_metadata2,
                 content1_metadata3,
                 content2_metadata,
             ]
         )
 
         assert set(
             swh_storage.raw_extrinsic_metadata_get_by_ids(
                 [content1_metadata1.id, b"\x00" * 20, content2_metadata.id]
             )
         ) == {content1_metadata1, content2_metadata}
 
     def test_content_metadata_get_authorities(self, swh_storage, sample_data):
         content1, content2, content3 = sample_data.contents[:3]
         fetcher, fetcher2 = sample_data.fetchers[:2]
         authority, authority2 = sample_data.authorities[:2]
         (
             content1_metadata1,
             content1_metadata2,
             content1_metadata3,
         ) = sample_data.content_metadata[:3]
 
         content2_metadata = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(content1_metadata2.to_dict(), ("id",)),  # recompute id
                 "target": str(content2.swhid()),
             }
         )
 
         content1_metadata2 = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(content1_metadata2.to_dict(), ("id",)),  # recompute id
                 "authority": authority2.to_dict(),
             }
         )
 
         swh_storage.metadata_authority_add([authority, authority2])
         swh_storage.metadata_fetcher_add([fetcher, fetcher2])
 
         swh_storage.raw_extrinsic_metadata_add(
             [
                 content1_metadata1,
                 content1_metadata2,
                 content1_metadata3,
                 content2_metadata,
             ]
         )
 
         assert swh_storage.raw_extrinsic_metadata_get_authorities(content1.swhid()) in (
             [authority, authority2],
             [authority2, authority],
         )
 
         assert swh_storage.raw_extrinsic_metadata_get_authorities(content2.swhid()) == [
             authority
         ]
 
         assert (
             swh_storage.raw_extrinsic_metadata_get_authorities(content3.swhid()) == []
         )
 
     def test_origin_metadata_add(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
 
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority
         )
         assert result.next_page_token is None
         assert list(sorted(result.results, key=lambda x: x.discovery_date)) == [
             origin_metadata,
             origin_metadata2,
         ]
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = [
             ("metadata_authority", authority),
             ("metadata_fetcher", fetcher),
             ("raw_extrinsic_metadata", origin_metadata),
             ("raw_extrinsic_metadata", origin_metadata2),
         ]
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_origin_metadata_add_duplicate(self, swh_storage, sample_data):
         """Duplicates should be silently updated."""
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
         swh_storage.raw_extrinsic_metadata_add([origin_metadata2, origin_metadata])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority
         )
         assert result.next_page_token is None
 
         # which of the two behavior happens is backend-specific.
         expected_results = (origin_metadata, origin_metadata2)
 
         assert (
             tuple(sorted(result.results, key=lambda x: x.discovery_date,))
             == expected_results
         )
 
     def test_origin_metadata_get(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         fetcher, fetcher2 = sample_data.fetchers[:2]
         authority, authority2 = sample_data.authorities[:2]
         (
             origin1_metadata1,
             origin1_metadata2,
             origin1_metadata3,
         ) = sample_data.origin_metadata[:3]
 
         assert swh_storage.origin_add([origin, origin2]) == {"origin:add": 2}
 
         origin2_metadata = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(origin1_metadata2.to_dict(), ("id",)),  # recompute id
                 "target": str(Origin(origin2.url).swhid()),
             }
         )
 
         swh_storage.metadata_authority_add([authority, authority2])
         swh_storage.metadata_fetcher_add([fetcher, fetcher2])
 
         swh_storage.raw_extrinsic_metadata_add(
             [origin1_metadata1, origin1_metadata2, origin1_metadata3, origin2_metadata]
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority
         )
         assert result.next_page_token is None
         assert [origin1_metadata1, origin1_metadata2] == list(
             sorted(result.results, key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority2
         )
         assert result.next_page_token is None
         assert [origin1_metadata3] == list(
             sorted(result.results, key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin2.url).swhid(), authority
         )
         assert result.next_page_token is None
         assert [origin2_metadata] == list(result.results,)
 
     def test_origin_metadata_get_after(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
 
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(),
             authority,
             after=origin_metadata.discovery_date - timedelta(seconds=1),
         )
         assert result.next_page_token is None
         assert list(sorted(result.results, key=lambda x: x.discovery_date,)) == [
             origin_metadata,
             origin_metadata2,
         ]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority, after=origin_metadata.discovery_date,
         )
         assert result.next_page_token is None
         assert result.results == [origin_metadata2]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(),
             authority,
             after=origin_metadata2.discovery_date,
         )
         assert result.next_page_token is None
         assert result.results == []
 
     def test_origin_metadata_get_paginate(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         swh_storage.raw_extrinsic_metadata_get(Origin(origin.url).swhid(), authority)
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority, limit=1
         )
         assert result.next_page_token is not None
         assert result.results == [origin_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(),
             authority,
             limit=1,
             page_token=result.next_page_token,
         )
         assert result.next_page_token is None
         assert result.results == [origin_metadata2]
 
     def test_origin_metadata_get_paginate_same_date(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher1, fetcher2 = sample_data.fetchers[:2]
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher1, fetcher2])
         swh_storage.metadata_authority_add([authority])
 
         new_origin_metadata2 = RawExtrinsicMetadata.from_dict(
             {
                 **remove_keys(origin_metadata2.to_dict(), ("id",)),  # recompute id
                 "discovery_date": origin_metadata2.discovery_date,
                 "fetcher": attr.evolve(fetcher2, metadata=None).to_dict(),
             }
         )
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, new_origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(), authority, limit=1
         )
         assert result.next_page_token is not None
         assert result.results == [origin_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             Origin(origin.url).swhid(),
             authority,
             limit=1,
             page_token=result.next_page_token,
         )
         assert result.next_page_token is None
         assert result.results == [new_origin_metadata2]
 
     def test_origin_metadata_add_missing_authority(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
 
         with pytest.raises(StorageArgumentException, match="authority"):
             swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
     def test_origin_metadata_add_missing_fetcher(self, swh_storage, sample_data):
         origin = sample_data.origin
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_authority_add([authority])
 
         with pytest.raises(StorageArgumentException, match="fetcher"):
             swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
 
 class TestStorageGeneratedData:
     def test_generate_content_get_data(self, swh_storage, swh_contents):
         contents_with_data = [c for c in swh_contents if c.status != "absent"]
 
         # retrieve contents
         for content in contents_with_data:
             actual_content_data = swh_storage.content_get_data(content.sha1)
             assert actual_content_data is not None
             assert actual_content_data == content.data
 
     def test_generate_content_get(self, swh_storage, swh_contents):
         expected_contents = [
             attr.evolve(c, data=None) for c in swh_contents if c.status != "absent"
         ]
 
         actual_contents = swh_storage.content_get([c.sha1 for c in expected_contents])
 
         assert len(actual_contents) == len(expected_contents)
         assert actual_contents == expected_contents
 
     @pytest.mark.parametrize("limit", [1, 7, 10, 100, 1000])
     def test_origin_list(self, swh_storage, swh_origins, limit):
         returned_origins = []
 
         page_token = None
         i = 0
         while True:
             actual_page = swh_storage.origin_list(page_token=page_token, limit=limit)
             assert len(actual_page.results) <= limit
 
             returned_origins.extend(actual_page.results)
 
             i += 1
             page_token = actual_page.next_page_token
 
             if page_token is None:
                 assert i * limit >= len(swh_origins)
                 break
             else:
                 assert len(actual_page.results) == limit
 
         assert sorted(returned_origins) == sorted(swh_origins)
 
     def test_origin_count(self, swh_storage, sample_data):
         swh_storage.origin_add(sample_data.origins)
 
         assert swh_storage.origin_count("github") == 3
         assert swh_storage.origin_count("gitlab") == 2
         assert swh_storage.origin_count(".*user.*", regexp=True) == 5
         assert swh_storage.origin_count(".*user.*", regexp=False) == 0
         assert swh_storage.origin_count(".*user1.*", regexp=True) == 2
         assert swh_storage.origin_count(".*user1.*", regexp=False) == 0
 
     def test_origin_count_with_visit_no_visits(self, swh_storage, sample_data):
         swh_storage.origin_add(sample_data.origins)
 
         # none of them have visits, so with_visit=True => 0
         assert swh_storage.origin_count("github", with_visit=True) == 0
         assert swh_storage.origin_count("gitlab", with_visit=True) == 0
         assert swh_storage.origin_count(".*user.*", regexp=True, with_visit=True) == 0
         assert swh_storage.origin_count(".*user.*", regexp=False, with_visit=True) == 0
         assert swh_storage.origin_count(".*user1.*", regexp=True, with_visit=True) == 0
         assert swh_storage.origin_count(".*user1.*", regexp=False, with_visit=True) == 0
 
     def test_origin_count_with_visit_with_visits_no_snapshot(
         self, swh_storage, sample_data
     ):
         swh_storage.origin_add(sample_data.origins)
 
         origin_url = "https://github.com/user1/repo1"
         visit = OriginVisit(origin=origin_url, date=now(), type="git",)
         swh_storage.origin_visit_add([visit])
 
         assert swh_storage.origin_count("github", with_visit=False) == 3
         # it has a visit, but no snapshot, so with_visit=True => 0
         assert swh_storage.origin_count("github", with_visit=True) == 0
 
         assert swh_storage.origin_count("gitlab", with_visit=False) == 2
         # these gitlab origins have no visit
         assert swh_storage.origin_count("gitlab", with_visit=True) == 0
 
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=False)
             == 1
         )
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=True) == 0
         )
         assert swh_storage.origin_count("github", regexp=True, with_visit=True) == 0
 
     def test_origin_count_with_visit_with_visits_and_snapshot(
         self, swh_storage, sample_data
     ):
         snapshot = sample_data.snapshot
         swh_storage.origin_add(sample_data.origins)
 
         swh_storage.snapshot_add([snapshot])
         origin_url = "https://github.com/user1/repo1"
         visit = OriginVisit(origin=origin_url, date=now(), type="git",)
         visit = swh_storage.origin_visit_add([visit])[0]
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin_url,
                     visit=visit.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         assert swh_storage.origin_count("github", with_visit=False) == 3
         # github/user1 has a visit and a snapshot, so with_visit=True => 1
         assert swh_storage.origin_count("github", with_visit=True) == 1
 
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=False)
             == 1
         )
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=True) == 1
         )
         assert swh_storage.origin_count("github", regexp=True, with_visit=True) == 1
 
     @settings(
         suppress_health_check=[HealthCheck.too_slow] + function_scoped_fixture_check,
     )
     @given(strategies.lists(objects(split_content=True), max_size=2))
     def test_add_arbitrary(self, swh_storage, objects):
         for (obj_type, obj) in objects:
             if obj.object_type == "origin_visit":
                 swh_storage.origin_add([Origin(url=obj.origin)])
                 visit = OriginVisit(origin=obj.origin, date=obj.date, type=obj.type,)
                 swh_storage.origin_visit_add([visit])
             elif obj.object_type == "raw_extrinsic_metadata":
                 swh_storage.metadata_authority_add([obj.authority])
                 swh_storage.metadata_fetcher_add([obj.fetcher])
                 swh_storage.raw_extrinsic_metadata_add([obj])
             else:
                 method = getattr(swh_storage, obj_type + "_add")
                 try:
                     method([obj])
                 except HashCollision:
                     pass