diff --git a/swh/storage/db.py b/swh/storage/db.py
index 086bded2..3d963bea 100644
--- a/swh/storage/db.py
+++ b/swh/storage/db.py
@@ -1,917 +1,917 @@
 # Copyright (C) 2015-2019  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 select
 
 from swh.core.db import BaseDb
 from swh.core.db.db_utils import stored_procedure, jsonize
 from swh.core.db.db_utils import execute_values_generator
 
 
 class Db(BaseDb):
     """Proxy to the SWH DB, with wrappers around stored procedures
 
     """
 
     def mktemp_dir_entry(self, entry_type, cur=None):
         self._cursor(cur).execute('SELECT swh_mktemp_dir_entry(%s)',
                                   (('directory_entry_%s' % entry_type),))
 
     @stored_procedure('swh_mktemp_revision')
     def mktemp_revision(self, cur=None): pass
 
     @stored_procedure('swh_mktemp_release')
     def mktemp_release(self, cur=None): pass
 
     @stored_procedure('swh_mktemp_snapshot_branch')
     def mktemp_snapshot_branch(self, cur=None): pass
 
     def register_listener(self, notify_queue, cur=None):
         """Register a listener for NOTIFY queue `notify_queue`"""
         self._cursor(cur).execute("LISTEN %s" % notify_queue)
 
     def listen_notifies(self, timeout):
         """Listen to notifications for `timeout` seconds"""
         if select.select([self.conn], [], [], timeout) == ([], [], []):
             return
         else:
             self.conn.poll()
             while self.conn.notifies:
                 yield self.conn.notifies.pop(0)
 
     @stored_procedure('swh_content_add')
     def content_add_from_temp(self, cur=None): pass
 
     @stored_procedure('swh_directory_add')
     def directory_add_from_temp(self, cur=None): pass
 
     @stored_procedure('swh_skipped_content_add')
     def skipped_content_add_from_temp(self, cur=None): pass
 
     @stored_procedure('swh_revision_add')
     def revision_add_from_temp(self, cur=None): pass
 
     @stored_procedure('swh_release_add')
     def release_add_from_temp(self, cur=None): pass
 
     def content_update_from_temp(self, keys_to_update, cur=None):
         cur = self._cursor(cur)
         cur.execute("""select swh_content_update(ARRAY[%s] :: text[])""" %
                     keys_to_update)
 
     content_get_metadata_keys = [
         'sha1', 'sha1_git', 'sha256', 'blake2s256', 'length', 'status']
 
     content_add_keys = content_get_metadata_keys + ['ctime']
 
     skipped_content_keys = [
         'sha1', 'sha1_git', 'sha256', 'blake2s256',
         'length', 'reason', 'status', 'origin']
 
     def content_get_metadata_from_sha1s(self, sha1s, cur=None):
         cur = self._cursor(cur)
 
         yield from execute_values_generator(
             cur, """
             select t.sha1, %s from (values %%s) as t (sha1)
             left join content using (sha1)
             """ % ', '.join(self.content_get_metadata_keys[1:]),
             ((sha1,) for sha1 in sha1s),
         )
 
     def content_get_range(self, start, end, limit=None, cur=None):
         """Retrieve contents within range [start, end].
 
         """
         cur = self._cursor(cur)
         query = """select %s from content
                    where %%s <= sha1 and sha1 <= %%s
                    order by sha1
                    limit %%s""" % ', '.join(self.content_get_metadata_keys)
         cur.execute(query, (start, end, limit))
         yield from cur
 
     content_hash_keys = ['sha1', 'sha1_git', 'sha256', 'blake2s256']
 
     def content_missing_from_list(self, contents, cur=None):
         cur = self._cursor(cur)
 
         keys = ', '.join(self.content_hash_keys)
         equality = ' AND '.join(
             ('t.%s = c.%s' % (key, key))
             for key in self.content_hash_keys
         )
 
         yield from execute_values_generator(
             cur, """
             SELECT %s
             FROM (VALUES %%s) as t(%s)
             WHERE NOT EXISTS (
                 SELECT 1 FROM content c
                 WHERE %s
             )
             """ % (keys, keys, equality),
             (tuple(c[key] for key in self.content_hash_keys) for c in contents)
         )
 
     def content_missing_per_sha1(self, sha1s, cur=None):
         cur = self._cursor(cur)
 
         yield from execute_values_generator(cur, """
         SELECT t.sha1 FROM (VALUES %s) AS t(sha1)
         WHERE NOT EXISTS (
             SELECT 1 FROM content c WHERE c.sha1 = t.sha1
         )""", ((sha1,) for sha1 in sha1s))
 
     def skipped_content_missing(self, contents, cur=None):
         if not contents:
             return []
         cur = self._cursor(cur)
 
         query = """SELECT * FROM (VALUES %s) AS t (%s)
                    WHERE not exists
                    (SELECT 1 FROM skipped_content s WHERE
                        s.sha1 is not distinct from t.sha1 and
                        s.sha1_git is not distinct from t.sha1_git and
                        s.sha256 is not distinct from t.sha256);""" % \
                 ((', '.join('%s' for _ in contents)),
                  ', '.join(self.content_hash_keys))
         cur.execute(query,
                     [tuple(cont[key] for key in self.content_hash_keys)
                      for cont in contents])
 
         yield from cur
 
     def snapshot_exists(self, snapshot_id, cur=None):
         """Check whether a snapshot with the given id exists"""
         cur = self._cursor(cur)
 
         cur.execute("""SELECT 1 FROM snapshot where id=%s""", (snapshot_id,))
 
         return bool(cur.fetchone())
 
     def snapshot_add(self, snapshot_id, cur=None):
         """Add a snapshot from the temporary table"""
         cur = self._cursor(cur)
 
         cur.execute("""SELECT swh_snapshot_add(%s)""", (snapshot_id,))
 
     snapshot_count_cols = ['target_type', 'count']
 
     def snapshot_count_branches(self, snapshot_id, cur=None):
         cur = self._cursor(cur)
         query = """\
            SELECT %s FROM swh_snapshot_count_branches(%%s)
         """ % ', '.join(self.snapshot_count_cols)
 
         cur.execute(query, (snapshot_id,))
 
         yield from cur
 
     snapshot_get_cols = ['snapshot_id', 'name', 'target', 'target_type']
 
     def snapshot_get_by_id(self, snapshot_id, branches_from=b'',
                            branches_count=None, target_types=None,
                            cur=None):
         cur = self._cursor(cur)
         query = """\
            SELECT %s
            FROM swh_snapshot_get_by_id(%%s, %%s, %%s, %%s :: snapshot_target[])
         """ % ', '.join(self.snapshot_get_cols)
 
         cur.execute(query, (snapshot_id, branches_from, branches_count,
                             target_types))
 
         yield from cur
 
     def snapshot_get_by_origin_visit(self, origin_url, visit_id, cur=None):
         cur = self._cursor(cur)
         query = """\
            SELECT snapshot FROM origin_visit
            INNER JOIN origin ON origin.id = origin_visit.origin
            WHERE origin.url=%s AND origin_visit.visit=%s;
         """
 
         cur.execute(query, (origin_url, visit_id))
         ret = cur.fetchone()
         if ret:
             return ret[0]
 
     content_find_cols = ['sha1', 'sha1_git', 'sha256', 'blake2s256', 'length',
                          'ctime', 'status']
 
     def content_find(self, sha1=None, sha1_git=None, sha256=None,
                      blake2s256=None, cur=None):
         """Find the content optionally on a combination of the following
         checksums sha1, sha1_git, sha256 or blake2s256.
 
         Args:
             sha1: sha1 content
             git_sha1: the sha1 computed `a la git` sha1 of the content
             sha256: sha256 content
             blake2s256: blake2s256 content
 
         Returns:
             The tuple (sha1, sha1_git, sha256, blake2s256) if found or None.
 
         """
         cur = self._cursor(cur)
 
         checksum_dict = {'sha1': sha1, 'sha1_git': sha1_git,
                          'sha256': sha256, 'blake2s256': blake2s256}
         where_parts = []
         args = []
         # Adds only those keys which have value other than None
         for algorithm in checksum_dict:
             if checksum_dict[algorithm] is not None:
                 args.append(checksum_dict[algorithm])
                 where_parts.append(algorithm + '= %s')
         query = ' AND '.join(where_parts)
         cur.execute("""SELECT %s
                        FROM content WHERE %s
                        """
                     % (','.join(self.content_find_cols),  query),
                     args)
         content = cur.fetchall()
         return content
 
     def directory_missing_from_list(self, directories, cur=None):
         cur = self._cursor(cur)
         yield from execute_values_generator(
             cur, """
             SELECT id FROM (VALUES %s) as t(id)
             WHERE NOT EXISTS (
                 SELECT 1 FROM directory d WHERE d.id = t.id
             )
             """, ((id,) for id in directories))
 
     directory_ls_cols = ['dir_id', 'type', 'target', 'name', 'perms',
                          'status', 'sha1', 'sha1_git', 'sha256', 'length']
 
     def directory_walk_one(self, directory, cur=None):
         cur = self._cursor(cur)
         cols = ', '.join(self.directory_ls_cols)
         query = 'SELECT %s FROM swh_directory_walk_one(%%s)' % cols
         cur.execute(query, (directory,))
         yield from cur
 
     def directory_walk(self, directory, cur=None):
         cur = self._cursor(cur)
         cols = ', '.join(self.directory_ls_cols)
         query = 'SELECT %s FROM swh_directory_walk(%%s)' % cols
         cur.execute(query, (directory,))
         yield from cur
 
     def directory_entry_get_by_path(self, directory, paths, cur=None):
         """Retrieve a directory entry by path.
 
         """
         cur = self._cursor(cur)
 
         cols = ', '.join(self.directory_ls_cols)
         query = (
             'SELECT %s FROM swh_find_directory_entry_by_path(%%s, %%s)' % cols)
         cur.execute(query, (directory, paths))
 
         data = cur.fetchone()
         if set(data) == {None}:
             return None
         return data
 
     def revision_missing_from_list(self, revisions, cur=None):
         cur = self._cursor(cur)
 
         yield from execute_values_generator(
             cur, """
             SELECT id FROM (VALUES %s) as t(id)
             WHERE NOT EXISTS (
                 SELECT 1 FROM revision r WHERE r.id = t.id
             )
             """, ((id,) for id in revisions))
 
     revision_add_cols = [
         'id', 'date', 'date_offset', 'date_neg_utc_offset', 'committer_date',
         'committer_date_offset', 'committer_date_neg_utc_offset', 'type',
         'directory', 'message', 'author_fullname', 'author_name',
         'author_email', 'committer_fullname', 'committer_name',
         'committer_email', 'metadata', 'synthetic',
     ]
 
     revision_get_cols = revision_add_cols + ['parents']
 
     def origin_visit_add(self, origin, ts, type, cur=None):
         """Add a new origin_visit for origin origin at timestamp ts with
         status 'ongoing'.
 
         Args:
             origin: origin concerned by the visit
             ts: the date of the visit
             type: type of loader for the visit
 
         Returns:
             The new visit index step for that origin
 
         """
         cur = self._cursor(cur)
         self._cursor(cur).execute('SELECT swh_origin_visit_add(%s, %s, %s)',
                                   (origin, ts, type))
         return cur.fetchone()[0]
 
     def origin_visit_update(self, origin_id, visit_id, updates, cur=None):
         """Update origin_visit's status."""
         cur = self._cursor(cur)
         update_cols = []
         values = []
         where = ['origin.id = origin_visit.origin',
                  'origin.url=%s',
                  'visit=%s']
         where_values = [origin_id, visit_id]
         if 'status' in updates:
             update_cols.append('status=%s')
             values.append(updates.pop('status'))
         if 'metadata' in updates:
             update_cols.append('metadata=%s')
             values.append(jsonize(updates.pop('metadata')))
         if 'snapshot' in updates:
             update_cols.append('snapshot=%s')
             values.append(updates.pop('snapshot'))
         assert not updates, 'Unknown fields: %r' % updates
         query = """UPDATE origin_visit
                    SET {update_cols}
                    FROM origin
                    WHERE {where}""".format(**{
             'update_cols': ', '.join(update_cols),
             'where': ' AND '.join(where)
         })
         cur.execute(query, (*values, *where_values))
 
     def origin_visit_upsert(self, origin, visit, date, type, status,
                             metadata, snapshot, cur=None):
         # doing an extra query like this is way simpler than trying to join
         # the origin id in the query below
         origin_id = next(self.origin_id_get_by_url([origin]))
 
         cur = self._cursor(cur)
         query = """INSERT INTO origin_visit ({cols}) VALUES ({values})
                    ON CONFLICT ON CONSTRAINT origin_visit_pkey DO
                    UPDATE SET {updates}""".format(
                 cols=', '.join(self.origin_visit_get_cols),
                 values=', '.join('%s' for col in self.origin_visit_get_cols),
                 updates=', '.join('{0}=excluded.{0}'.format(col)
                                   for col in self.origin_visit_get_cols))
         cur.execute(
             query, (origin_id, visit, date, type, status, metadata, snapshot))
 
     origin_visit_get_cols = [
         'origin', 'visit', 'date', 'type',
         'status', 'metadata', 'snapshot']
     origin_visit_select_cols = [
         'origin.url AS origin', 'visit', 'date', 'origin_visit.type AS type',
         'status', 'metadata', 'snapshot']
 
     def origin_visit_get_all(self, origin_id,
                              last_visit=None, limit=None, cur=None):
         """Retrieve all visits for origin with id origin_id.
 
         Args:
             origin_id: The occurrence's origin
 
         Yields:
             The occurrence's history visits
 
         """
         cur = self._cursor(cur)
 
         if last_visit:
             extra_condition = 'and visit > %s'
             args = (origin_id, last_visit, limit)
         else:
             extra_condition = ''
             args = (origin_id, limit)
 
         query = """\
         SELECT %s
         FROM origin_visit
         INNER JOIN origin ON origin.id = origin_visit.origin
         WHERE origin.url=%%s %s
         order by visit asc
         limit %%s""" % (
             ', '.join(self.origin_visit_select_cols), extra_condition
         )
 
         cur.execute(query, args)
 
         yield from cur
 
     def origin_visit_get(self, origin_id, visit_id, cur=None):
         """Retrieve information on visit visit_id of origin origin_id.
 
         Args:
             origin_id: the origin concerned
             visit_id: The visit step for that origin
 
         Returns:
             The origin_visit information
 
         """
         cur = self._cursor(cur)
 
         query = """\
             SELECT %s
             FROM origin_visit
             INNER JOIN origin ON origin.id = origin_visit.origin
             WHERE origin.url = %%s AND visit = %%s
             """ % (', '.join(self.origin_visit_select_cols))
 
         cur.execute(query, (origin_id, visit_id))
         r = cur.fetchall()
         if not r:
             return None
         return r[0]
 
     def origin_visit_find_by_date(self, origin, visit_date, cur=None):
         cur = self._cursor(cur)
         cur.execute(
             'SELECT * FROM swh_visit_find_by_date(%s, %s)',
             (origin, visit_date))
         r = cur.fetchall()
         if r:
             return r[0]
 
     def origin_visit_exists(self, origin_id, visit_id, cur=None):
         """Check whether an origin visit with the given ids exists"""
         cur = self._cursor(cur)
 
         query = "SELECT 1 FROM origin_visit where origin = %s AND visit = %s"
 
         cur.execute(query, (origin_id, visit_id))
 
         return bool(cur.fetchone())
 
     def origin_visit_get_latest(
             self, origin_id, allowed_statuses=None, require_snapshot=False,
             cur=None):
         """Retrieve the most recent origin_visit of the given origin,
         with optional filters.
 
         Args:
             origin_id: the origin concerned
             allowed_statuses: the visit statuses allowed for the returned visit
             require_snapshot (bool): If True, only a visit with a known
                 snapshot will be returned.
 
         Returns:
             The origin_visit information, or None if no visit matches.
         """
         cur = self._cursor(cur)
 
         query_parts = [
             'SELECT %s' % ', '.join(self.origin_visit_select_cols),
             'FROM origin_visit',
             'INNER JOIN origin ON origin.id = origin_visit.origin']
 
         query_parts.append('WHERE origin.url = %s')
 
         if require_snapshot:
             query_parts.append('AND snapshot is not null')
 
         if allowed_statuses:
             query_parts.append(
                 cur.mogrify('AND status IN %s',
                             (tuple(allowed_statuses),)).decode())
 
         query_parts.append('ORDER BY date DESC, visit DESC LIMIT 1')
 
         query = '\n'.join(query_parts)
 
         cur.execute(query, (origin_id,))
         r = cur.fetchone()
         if not r:
             return None
         return r
 
+    def origin_visit_get_random(self, type, cur=None):
+        """Randomly select one origin visit that was full and in the last 3
+           months
+
+        """
+        cur = self._cursor(cur)
+        columns = ','.join(self.origin_visit_select_cols)
+        query = f"""with visits as (
+                      select *
+                      from origin_visit
+                      where origin_visit.status='full' and
+                            origin_visit.type=%s and
+                            origin_visit.date > now() - '3 months'::interval
+                    )
+                    select {columns}
+                    from visits as origin_visit
+                    inner join origin
+                    on origin_visit.origin=origin.id
+                    where random() < 0.1
+                    limit 1
+                 """
+        cur.execute(query, (type, ))
+        return cur.fetchone()
+
     @staticmethod
     def mangle_query_key(key, main_table):
         if key == 'id':
             return 't.id'
         if key == 'parents':
             return '''
             ARRAY(
             SELECT rh.parent_id::bytea
             FROM revision_history rh
             WHERE rh.id = t.id
             ORDER BY rh.parent_rank
             )'''
         if '_' not in key:
             return '%s.%s' % (main_table, key)
 
         head, tail = key.split('_', 1)
         if (head in ('author', 'committer')
                 and tail in ('name', 'email', 'id', 'fullname')):
             return '%s.%s' % (head, tail)
 
         return '%s.%s' % (main_table, key)
 
     def revision_get_from_list(self, revisions, cur=None):
         cur = self._cursor(cur)
 
         query_keys = ', '.join(
             self.mangle_query_key(k, 'revision')
             for k in self.revision_get_cols
         )
 
         yield from execute_values_generator(
             cur, """
             SELECT %s FROM (VALUES %%s) as t(id)
             LEFT JOIN revision ON t.id = revision.id
             LEFT JOIN person author ON revision.author = author.id
             LEFT JOIN person committer ON revision.committer = committer.id
             """ % query_keys,
             ((id,) for id in revisions))
 
     def revision_log(self, root_revisions, limit=None, cur=None):
         cur = self._cursor(cur)
 
         query = """SELECT %s
                    FROM swh_revision_log(%%s, %%s)
                 """ % ', '.join(self.revision_get_cols)
 
         cur.execute(query, (root_revisions, limit))
         yield from cur
 
     revision_shortlog_cols = ['id', 'parents']
 
     def revision_shortlog(self, root_revisions, limit=None, cur=None):
         cur = self._cursor(cur)
 
         query = """SELECT %s
                    FROM swh_revision_list(%%s, %%s)
                 """ % ', '.join(self.revision_shortlog_cols)
 
         cur.execute(query, (root_revisions, limit))
         yield from cur
 
     def release_missing_from_list(self, releases, cur=None):
         cur = self._cursor(cur)
         yield from execute_values_generator(
             cur, """
             SELECT id FROM (VALUES %s) as t(id)
             WHERE NOT EXISTS (
                 SELECT 1 FROM release r WHERE r.id = t.id
             )
             """, ((id,) for id in releases))
 
     object_find_by_sha1_git_cols = ['sha1_git', 'type', 'id', 'object_id']
 
     def object_find_by_sha1_git(self, ids, cur=None):
         cur = self._cursor(cur)
 
         yield from execute_values_generator(
             cur, """
             WITH t (id) AS (VALUES %s),
             known_objects as ((
                 select
                   id as sha1_git,
                   'release'::object_type as type,
                   id,
                   object_id
                 from release r
                 where exists (select 1 from t where t.id = r.id)
             ) union all (
                 select
                   id as sha1_git,
                   'revision'::object_type as type,
                   id,
                   object_id
                 from revision r
                 where exists (select 1 from t where t.id = r.id)
             ) union all (
                 select
                   id as sha1_git,
                   'directory'::object_type as type,
                   id,
                   object_id
                 from directory d
                 where exists (select 1 from t where t.id = d.id)
             ) union all (
                 select
                   sha1_git as sha1_git,
                   'content'::object_type as type,
                   sha1 as id,
                   object_id
                 from content c
                 where exists (select 1 from t where t.id = c.sha1_git)
             ))
             select t.id as sha1_git, k.type, k.id, k.object_id
             from t
             left join known_objects k on t.id = k.sha1_git
             """,
             ((id,) for id in ids)
         )
 
     def stat_counters(self, cur=None):
         cur = self._cursor(cur)
         cur.execute('SELECT * FROM swh_stat_counters()')
         yield from cur
 
     def origin_add(self, url, cur=None):
         """Insert a new origin and return the new identifier."""
         insert = """INSERT INTO origin (url) values (%s)
                     RETURNING url"""
 
         cur.execute(insert, (url,))
         return cur.fetchone()[0]
 
     origin_cols = ['url']
 
     def origin_get_by_url(self, origins, cur=None):
         """Retrieve origin `(type, url)` from urls if found."""
         cur = self._cursor(cur)
 
         query = """SELECT %s FROM (VALUES %%s) as t(url)
                    LEFT JOIN origin ON t.url = origin.url
                 """ % ','.join('origin.' + col for col in self.origin_cols)
 
         yield from execute_values_generator(
             cur, query, ((url,) for url in origins))
 
     def origin_get_by_sha1(self, sha1s, cur=None):
         """Retrieve origin urls from sha1s if found."""
         cur = self._cursor(cur)
 
         query = """SELECT %s FROM (VALUES %%s) as t(sha1)
                    LEFT JOIN origin ON t.sha1 = digest(origin.url, 'sha1')
                 """ % ','.join('origin.' + col for col in self.origin_cols)
 
         yield from execute_values_generator(
             cur, query, ((sha1,) for sha1 in sha1s))
 
-    def origin_visit_get_random(self, type, cur=None):
-        """Randomly select one origin whose last visit was full in the last 3
-           months
-
-        """
-        cur = self._cursor(cur)
-        columns = ','.join(self.origin_visit_select_cols)
-        query = f"""with visits as (
-                      select *
-                      from origin_visit
-                      where origin_visit.status='full' and
-                            origin_visit.type=%s and
-                            origin_visit.date > now() - '3 months'::interval
-                    )
-                    select {columns}
-                    from visits as origin_visit
-                    inner join origin
-                    on origin_visit.origin=origin.id
-                    where random() < 0.1
-                    limit 1
-                 """
-        cur.execute(query, (type, ))
-        return cur.fetchone()
-
     def origin_id_get_by_url(self, origins, cur=None):
         """Retrieve origin `(type, url)` from urls if found."""
         cur = self._cursor(cur)
 
         query = """SELECT id FROM (VALUES %s) as t(url)
                    LEFT JOIN origin ON t.url = origin.url
                 """
 
         for row in execute_values_generator(
                 cur, query, ((url,) for url in origins)):
             yield row[0]
 
     origin_get_range_cols = ['id', 'url']
 
     def origin_get_range(self, origin_from=1, origin_count=100, cur=None):
         """Retrieve ``origin_count`` origins whose ids are greater
         or equal than ``origin_from``.
 
         Origins are sorted by id before retrieving them.
 
         Args:
             origin_from (int): the minimum id of origins to retrieve
             origin_count (int): the maximum number of origins to retrieve
         """
         cur = self._cursor(cur)
 
         query = """SELECT %s
                    FROM origin WHERE id >= %%s
                    ORDER BY id LIMIT %%s
                 """ % ','.join(self.origin_get_range_cols)
 
         cur.execute(query, (origin_from, origin_count))
         yield from cur
 
     def _origin_query(self, url_pattern, count=False, offset=0, limit=50,
                       regexp=False, with_visit=False, cur=None):
         """
         Method factorizing query creation for searching and counting origins.
         """
         cur = self._cursor(cur)
 
         if count:
             origin_cols = 'COUNT(*)'
         else:
             origin_cols = ','.join(self.origin_cols)
 
         query = """SELECT %s
                    FROM origin
                    WHERE """
         if with_visit:
             query += """
                    EXISTS (
                      SELECT 1
                      FROM origin_visit
                        INNER JOIN snapshot ON snapshot=snapshot.id
                      WHERE origin=origin.id
                      )
                    AND """
         query += 'url %s %%s '
         if not count:
             query += 'ORDER BY id OFFSET %%s LIMIT %%s'
 
         if not regexp:
             query = query % (origin_cols, 'ILIKE')
             query_params = ('%'+url_pattern+'%', offset, limit)
         else:
             query = query % (origin_cols, '~*')
             query_params = (url_pattern, offset, limit)
 
         if count:
             query_params = (query_params[0],)
 
         cur.execute(query, query_params)
 
     def origin_search(self, url_pattern, offset=0, limit=50,
                       regexp=False, with_visit=False, cur=None):
         """Search for origins whose urls contain a provided string pattern
         or match a provided regular expression.
         The search is performed in a case insensitive way.
 
         Args:
             url_pattern (str): the string pattern to search for in origin urls
             offset (int): number of found origins to skip before returning
                 results
             limit (int): the maximum number of found origins to return
             regexp (bool): if True, consider the provided pattern as a regular
                 expression and returns origins whose urls match it
             with_visit (bool): if True, filter out origins with no visit
 
         """
         self._origin_query(url_pattern, offset=offset, limit=limit,
                            regexp=regexp, with_visit=with_visit, cur=cur)
         yield from cur
 
     def origin_count(self, url_pattern, regexp=False,
                      with_visit=False, cur=None):
         """Count origins whose urls contain a provided string pattern
         or match a provided regular expression.
         The pattern search in origin urls is performed in a case insensitive
         way.
 
         Args:
             url_pattern (str): the string pattern to search for in origin urls
             regexp (bool): if True, consider the provided pattern as a regular
                 expression and returns origins whose urls match it
             with_visit (bool): if True, filter out origins with no visit
         """
         self._origin_query(url_pattern, count=True,
                            regexp=regexp, with_visit=with_visit, cur=cur)
         return cur.fetchone()[0]
 
     release_add_cols = [
         'id', 'target', 'target_type', 'date', 'date_offset',
         'date_neg_utc_offset', 'name', 'comment', 'synthetic',
         'author_fullname', 'author_name', 'author_email',
     ]
     release_get_cols = release_add_cols
 
     def release_get_from_list(self, releases, cur=None):
         cur = self._cursor(cur)
         query_keys = ', '.join(
             self.mangle_query_key(k, 'release')
             for k in self.release_get_cols
         )
 
         yield from execute_values_generator(
             cur, """
             SELECT %s FROM (VALUES %%s) as t(id)
             LEFT JOIN release ON t.id = release.id
             LEFT JOIN person author ON release.author = author.id
             """ % query_keys,
             ((id,) for id in releases))
 
     def origin_metadata_add(self, origin, ts, provider, tool,
                             metadata, cur=None):
         """ Add an origin_metadata for the origin at ts with provider, tool and
         metadata.
 
         Args:
             origin (int): the origin's id for which the metadata is added
             ts (datetime): time when the metadata was found
             provider (int): the metadata provider identifier
             tool (int): the tool's identifier used to extract metadata
             metadata (jsonb): the metadata retrieved at the time and location
 
         Returns:
             id (int): the origin_metadata unique id
 
         """
         cur = self._cursor(cur)
         insert = """INSERT INTO origin_metadata (origin_id, discovery_date,
                     provider_id, tool_id, metadata)
                     SELECT id, %s, %s, %s, %s FROM origin WHERE url = %s"""
         cur.execute(insert, (ts, provider, tool, jsonize(metadata), origin))
 
     origin_metadata_get_cols = ['origin_url', 'discovery_date',
                                 'tool_id', 'metadata', 'provider_id',
                                 'provider_name', 'provider_type',
                                 'provider_url']
 
     def origin_metadata_get_by(self, origin_url, provider_type=None, cur=None):
         """Retrieve all origin_metadata entries for one origin_url
 
         """
         cur = self._cursor(cur)
         if not provider_type:
             query = '''SELECT %s
                        FROM swh_origin_metadata_get_by_origin(
                             %%s)''' % (','.join(
                 self.origin_metadata_get_cols))
 
             cur.execute(query, (origin_url, ))
 
         else:
             query = '''SELECT %s
                        FROM swh_origin_metadata_get_by_provider_type(
                             %%s, %%s)''' % (','.join(
                 self.origin_metadata_get_cols))
 
             cur.execute(query, (origin_url, provider_type))
 
         yield from cur
 
     tool_cols = ['id', 'name', 'version', 'configuration']
 
     @stored_procedure('swh_mktemp_tool')
     def mktemp_tool(self, cur=None):
         pass
 
     def tool_add_from_temp(self, cur=None):
         cur = self._cursor(cur)
         cur.execute("SELECT %s from swh_tool_add()" % (
             ','.join(self.tool_cols), ))
         yield from cur
 
     def tool_get(self, name, version, configuration, cur=None):
         cur = self._cursor(cur)
         cur.execute('''select %s
                        from tool
                        where name=%%s and
                              version=%%s and
                              configuration=%%s''' % (
             ','.join(self.tool_cols)),
             (name, version, configuration))
 
         return cur.fetchone()
 
     metadata_provider_cols = ['id', 'provider_name', 'provider_type',
                               'provider_url', 'metadata']
 
     def metadata_provider_add(self, provider_name, provider_type,
                               provider_url, metadata, cur=None):
         """Insert a new provider and return the new identifier."""
         cur = self._cursor(cur)
         insert = """INSERT INTO metadata_provider (provider_name, provider_type,
                     provider_url, metadata) values (%s, %s, %s, %s)
                     RETURNING id"""
 
         cur.execute(insert, (provider_name, provider_type, provider_url,
                              jsonize(metadata)))
         return cur.fetchone()[0]
 
     def metadata_provider_get(self, provider_id, cur=None):
         cur = self._cursor(cur)
         cur.execute('''select %s
                        from metadata_provider
                        where id=%%s ''' % (
             ','.join(self.metadata_provider_cols)),
             (provider_id, ))
 
         return cur.fetchone()
 
     def metadata_provider_get_by(self, provider_name, provider_url,
                                  cur=None):
         cur = self._cursor(cur)
         cur.execute('''select %s
                        from metadata_provider
                        where provider_name=%%s and
                              provider_url=%%s''' % (
             ','.join(self.metadata_provider_cols)),
             (provider_name, provider_url))
 
         return cur.fetchone()
diff --git a/swh/storage/in_memory.py b/swh/storage/in_memory.py
index f5d47182..34d2b66e 100644
--- a/swh/storage/in_memory.py
+++ b/swh/storage/in_memory.py
@@ -1,1741 +1,1742 @@
 # Copyright (C) 2015-2019  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 re
 import bisect
 import dateutil
 import collections
 import copy
 import datetime
 import itertools
 import random
 
 from collections import defaultdict
 from datetime import timedelta
 from typing import Any, Dict, Mapping
 
 import attr
 
 from swh.model.model import \
     Content, Directory, Revision, Release, Snapshot, OriginVisit, Origin
 from swh.model.hashutil import DEFAULT_ALGORITHMS
 from swh.objstorage import get_objstorage
 from swh.objstorage.exc import ObjNotFoundError
 
 from .storage import get_journal_writer
 from .converters import origin_url_to_sha1
 
 # Max block size of contents to return
 BULK_BLOCK_CONTENT_LEN_MAX = 10000
 
 
 def now():
     return datetime.datetime.now(tz=datetime.timezone.utc)
 
 
 class Storage:
     def __init__(self, journal_writer=None):
         self._contents = {}
         self._content_indexes = defaultdict(lambda: defaultdict(set))
         self._skipped_contents = {}
         self._skipped_content_indexes = defaultdict(lambda: defaultdict(set))
 
         self.reset()
 
         if journal_writer:
             self.journal_writer = get_journal_writer(**journal_writer)
         else:
             self.journal_writer = None
 
     def reset(self):
         self._directories = {}
         self._revisions = {}
         self._releases = {}
         self._snapshots = {}
         self._origins = {}
         self._origins_by_id = []
         self._origins_by_sha1 = {}
         self._origin_visits = {}
         self._persons = []
         self._origin_metadata = defaultdict(list)
         self._tools = {}
         self._metadata_providers = {}
         self._objects = defaultdict(list)
 
         # ideally we would want a skip list for both fast inserts and searches
         self._sorted_sha1s = []
 
         self.objstorage = get_objstorage('memory', {})
 
     def check_config(self, *, check_write):
         """Check that the storage is configured and ready to go."""
         return True
 
     def _content_add(self, contents, with_data):
         content_with_data = []
         content_without_data = []
         for content in contents:
             if content.status is None:
                 content.status = 'visible'
             if content.length is None:
                 content.length = -1
             if content.status != 'absent':
                 if self._content_key(content) not in self._contents:
                     content_with_data.append(content)
             else:
                 if self._content_key(content) not in self._skipped_contents:
                     content_without_data.append(content)
 
         if self.journal_writer:
             for content in content_with_data:
                 content = attr.evolve(content, data=None)
                 self.journal_writer.write_addition('content', content)
             for content in content_without_data:
                 self.journal_writer.write_addition('content', content)
 
         count_content_added, count_content_bytes_added = \
             self._content_add_present(content_with_data, with_data)
 
         count_skipped_content_added = self._content_add_absent(
             content_without_data
         )
 
         summary = {
             'content:add': count_content_added,
             'skipped_content:add': count_skipped_content_added,
         }
 
         if with_data:
             summary['content:add:bytes'] = count_content_bytes_added
 
         return summary
 
     def _content_add_present(self, contents, with_data):
         count_content_added = 0
         count_content_bytes_added = 0
         for content in contents:
             key = self._content_key(content)
             if key in self._contents:
                 continue
             for algorithm in DEFAULT_ALGORITHMS:
                 hash_ = content.get_hash(algorithm)
                 if hash_ in self._content_indexes[algorithm]\
                    and (algorithm not in {'blake2s256', 'sha256'}):
                     from . import HashCollision
                     raise HashCollision(algorithm, hash_, key)
             for algorithm in DEFAULT_ALGORITHMS:
                 hash_ = content.get_hash(algorithm)
                 self._content_indexes[algorithm][hash_].add(key)
             self._objects[content.sha1_git].append(
                 ('content', content.sha1))
             self._contents[key] = content
             bisect.insort(self._sorted_sha1s, content.sha1)
             count_content_added += 1
             if with_data:
                 content_data = self._contents[key].data
                 self._contents[key] = attr.evolve(
                     self._contents[key],
                     data=None)
                 count_content_bytes_added += len(content_data)
                 self.objstorage.add(content_data, content.sha1)
 
         return (count_content_added, count_content_bytes_added)
 
     def _content_add_absent(self, contents):
         count = 0
         skipped_content_missing = self.skipped_content_missing(contents)
         for content in skipped_content_missing:
             key = self._content_key(content)
             for algo in DEFAULT_ALGORITHMS:
                 self._skipped_content_indexes[algo][content.get_hash(algo)] \
                     .add(key)
             self._skipped_contents[key] = content
             count += 1
 
         return count
 
     def _content_to_model(self, contents):
         """Takes a list of content dicts, optionally with an extra 'origin'
         key, and yields tuples (model.Content, origin)."""
         for content in contents:
             content = content.copy()
             content.pop('origin', None)
             yield Content.from_dict(content)
 
     def content_add(self, content):
         """Add content blobs to the storage
 
         Args:
             content (iterable): iterable of dictionaries representing
                 individual pieces of content to add. Each dictionary has the
                 following keys:
 
                 - data (bytes): the actual content
                 - length (int): content length (default: -1)
                 - one key for each checksum algorithm in
                   :data:`swh.model.hashutil.DEFAULT_ALGORITHMS`, mapped to the
                   corresponding checksum
                 - status (str): one of visible, hidden, absent
                 - reason (str): if status = absent, the reason why
                 - origin (int): if status = absent, the origin we saw the
                   content in
 
         Raises:
             HashCollision in case of collision
 
         Returns:
             Summary dict with the following key and associated values:
 
                 content:add: New contents added
                 content_bytes:add: Sum of the contents' length data
                 skipped_content:add: New skipped contents (no data) added
 
         """
         now = datetime.datetime.now(tz=datetime.timezone.utc)
         content = [attr.evolve(c, ctime=now)
                    for c in self._content_to_model(content)]
         return self._content_add(content, with_data=True)
 
     def content_add_metadata(self, content):
         """Add content metadata to the storage (like `content_add`, but
         without inserting to the objstorage).
 
         Args:
             content (iterable): iterable of dictionaries representing
                 individual pieces of content to add. Each dictionary has the
                 following keys:
 
                 - length (int): content length (default: -1)
                 - one key for each checksum algorithm in
                   :data:`swh.model.hashutil.DEFAULT_ALGORITHMS`, mapped to the
                   corresponding checksum
                 - status (str): one of visible, hidden, absent
                 - reason (str): if status = absent, the reason why
                 - origin (int): if status = absent, the origin we saw the
                   content in
                 - ctime (datetime): time of insertion in the archive
 
         Raises:
             HashCollision in case of collision
 
         Returns:
             Summary dict with the following key and associated values:
 
                 content:add: New contents added
                 skipped_content:add: New skipped contents (no data) added
 
         """
         content = list(self._content_to_model(content))
         return self._content_add(content, with_data=False)
 
     def content_get(self, content):
         """Retrieve in bulk contents and their data.
 
         This function may yield more blobs than provided sha1 identifiers,
         in case they collide.
 
         Args:
             content: iterables of sha1
 
         Yields:
             Dict[str, bytes]: Generates streams of contents as dict with their
                 raw data:
 
                 - sha1 (bytes): content id
                 - data (bytes): content's raw data
 
         Raises:
             ValueError in case of too much contents are required.
             cf. BULK_BLOCK_CONTENT_LEN_MAX
 
         """
         # FIXME: Make this method support slicing the `data`.
         if len(content) > BULK_BLOCK_CONTENT_LEN_MAX:
             raise ValueError(
                 "Sending at most %s contents." % BULK_BLOCK_CONTENT_LEN_MAX)
         for obj_id in content:
             try:
                 data = self.objstorage.get(obj_id)
             except ObjNotFoundError:
                 yield None
                 continue
 
             yield {'sha1': obj_id, 'data': data}
 
     def content_get_range(self, start, end, limit=1000, db=None, cur=None):
         """Retrieve contents within range [start, end] bound by limit.
 
         Note that this function may return more than one blob per hash. The
         limit is enforced with multiplicity (ie. two blobs with the same hash
         will count twice toward the limit).
 
         Args:
             **start** (bytes): Starting identifier range (expected smaller
                            than end)
             **end** (bytes): Ending identifier range (expected larger
                              than start)
             **limit** (int): Limit result (default to 1000)
 
         Returns:
             a dict with keys:
             - contents [dict]: iterable of contents in between the range.
             - next (bytes): There remains content in the range
               starting from this next sha1
 
         """
         if limit is None:
             raise ValueError('Development error: limit should not be None')
         from_index = bisect.bisect_left(self._sorted_sha1s, start)
         sha1s = itertools.islice(self._sorted_sha1s, from_index, None)
         sha1s = ((sha1, content_key)
                  for sha1 in sha1s
                  for content_key in self._content_indexes['sha1'][sha1])
         matched = []
         next_content = None
         for sha1, key in sha1s:
             if sha1 > end:
                 break
             if len(matched) >= limit:
                 next_content = sha1
                 break
             matched.append(self._contents[key].to_dict())
         return {
             'contents': matched,
             'next': next_content,
         }
 
     def content_get_metadata(self, content):
         """Retrieve content metadata in bulk
 
         Args:
             content: iterable of content identifiers (sha1)
 
         Returns:
             an iterable with content metadata corresponding to the given ids
         """
         # FIXME: the return value should be a mapping from search key to found
         # content*s*
         for sha1 in content:
             if sha1 in self._content_indexes['sha1']:
                 objs = self._content_indexes['sha1'][sha1]
                 # FIXME: rather than selecting one of the objects with that
                 # hash, we should return all of them. See:
                 # https://forge.softwareheritage.org/D645?id=1994#inline-3389
                 key = random.sample(objs, 1)[0]
                 d = self._contents[key].to_dict()
                 del d['ctime']
                 yield d
             else:
                 # FIXME: should really be None
                 yield {
                     'sha1': sha1,
                     'sha1_git': None,
                     'sha256': None,
                     'blake2s256': None,
                     'length': None,
                     'status': None,
                 }
 
     def content_find(self, content):
         if not set(content).intersection(DEFAULT_ALGORITHMS):
             raise ValueError('content keys must contain at least one of: '
                              '%s' % ', '.join(sorted(DEFAULT_ALGORITHMS)))
         found = []
         for algo in DEFAULT_ALGORITHMS:
             hash = content.get(algo)
             if hash and hash in self._content_indexes[algo]:
                 found.append(self._content_indexes[algo][hash])
 
         if not found:
             return []
 
         keys = list(set.intersection(*found))
         return [self._contents[key].to_dict() for key in keys]
 
     def content_missing(self, content, key_hash='sha1'):
         """List content missing from storage
 
         Args:
             contents ([dict]): iterable of dictionaries whose keys are
                                either 'length' or an item of
                                :data:`swh.model.hashutil.ALGORITHMS`;
                                mapped to the corresponding checksum
                                (or length).
 
             key_hash (str): name of the column to use as hash id
                             result (default: 'sha1')
 
         Returns:
             iterable ([bytes]): missing content ids (as per the
             key_hash column)
         """
         for cont in content:
             for (algo, hash_) in cont.items():
                 if algo not in DEFAULT_ALGORITHMS:
                     continue
                 if hash_ not in self._content_indexes.get(algo, []):
                     yield cont[key_hash]
                     break
             else:
                 for result in self.content_find(cont):
                     if result['status'] == 'missing':
                         yield cont[key_hash]
 
     def content_missing_per_sha1(self, contents):
         """List content missing from storage based only on sha1.
 
         Args:
             contents: Iterable of sha1 to check for absence.
 
         Returns:
             iterable: missing ids
 
         Raises:
             TODO: an exception when we get a hash collision.
 
         """
         for content in contents:
             if content not in self._content_indexes['sha1']:
                 yield content
 
     def skipped_content_missing(self, contents):
         """List all skipped_content missing from storage
 
         Args:
             contents: Iterable of sha1 to check for skipped content entry
 
         Returns:
             iterable: dict of skipped content entry
         """
 
         for content in contents:
             for (key, algorithm) in self._content_key_algorithm(content):
                 if algorithm == 'blake2s256':
                     continue
                 if key not in self._skipped_content_indexes[algorithm]:
                     # index must contain hashes of algos except blake2s256
                     # else the content is considered skipped
                     yield content
                     break
 
     def directory_add(self, directories):
         """Add directories to the storage
 
         Args:
             directories (iterable): iterable of dictionaries representing the
                 individual directories to add. Each dict has the following
                 keys:
 
                 - id (sha1_git): the id of the directory to add
                 - entries (list): list of dicts for each entry in the
                       directory.  Each dict has the following keys:
 
                       - name (bytes)
                       - type (one of 'file', 'dir', 'rev'): type of the
                         directory entry (file, directory, revision)
                       - target (sha1_git): id of the object pointed at by the
                         directory entry
                       - perms (int): entry permissions
         Returns:
             Summary dict of keys with associated count as values:
 
                 directory:add: Number of directories actually added
 
         """
         if self.journal_writer:
             self.journal_writer.write_additions(
                 'directory',
                 (dir_ for dir_ in directories
                  if dir_['id'] not in self._directories))
 
         directories = [Directory.from_dict(d) for d in directories]
 
         count = 0
         for directory in directories:
             if directory.id not in self._directories:
                 count += 1
                 self._directories[directory.id] = directory
                 self._objects[directory.id].append(
                     ('directory', directory.id))
 
         return {'directory:add': count}
 
     def directory_missing(self, directories):
         """List directories missing from storage
 
         Args:
             directories (iterable): an iterable of directory ids
 
         Yields:
             missing directory ids
 
         """
         for id in directories:
             if id not in self._directories:
                 yield id
 
     def _join_dentry_to_content(self, dentry):
         keys = (
             'status',
             'sha1',
             'sha1_git',
             'sha256',
             'length',
         )
         ret = dict.fromkeys(keys)
         ret.update(dentry)
         if ret['type'] == 'file':
             # TODO: Make it able to handle more than one content
             content = self.content_find({'sha1_git': ret['target']})
             if content:
                 content = content[0]
                 for key in keys:
                     ret[key] = content[key]
         return ret
 
     def _directory_ls(self, directory_id, recursive, prefix=b''):
         if directory_id in self._directories:
             for entry in self._directories[directory_id].entries:
                 ret = self._join_dentry_to_content(entry.to_dict())
                 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_ls(self, directory, recursive=False):
         """Get entries for one directory.
 
         Args:
             - directory: the directory to list entries from.
             - recursive: if flag on, this list recursively from this directory.
 
         Returns:
             List of entries for such directory.
 
         If `recursive=True`, names in the path of a dir/file not at the
         root are concatenated with a slash (`/`).
         """
         yield from self._directory_ls(directory, recursive)
 
     def directory_entry_get_by_path(self, directory, paths):
         """Get the directory entry (either file or dir) from directory with path.
 
         Args:
             - directory: sha1 of the top level directory
             - paths: path to lookup from the top level directory. From left
               (top) to right (bottom).
 
         Returns:
             The corresponding directory entry if found, None otherwise.
 
         """
         return self._directory_entry_get_by_path(directory, paths, b'')
 
     def _directory_entry_get_by_path(self, directory, paths, prefix):
         if not paths:
             return
 
         contents = list(self.directory_ls(directory))
 
         if not contents:
             return
 
         def _get_entry(entries, name):
             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
 
         return self._directory_entry_get_by_path(
                 first_item['target'], paths[1:], prefix + paths[0] + b'/')
 
     def revision_add(self, revisions):
         """Add revisions to the storage
 
         Args:
             revisions (Iterable[dict]): iterable of dictionaries representing
                 the individual revisions to add. Each dict has the following
                 keys:
 
                 - **id** (:class:`sha1_git`): id of the revision to add
                 - **date** (:class:`dict`): date the revision was written
                 - **committer_date** (:class:`dict`): date the revision got
                   added to the origin
                 - **type** (one of 'git', 'tar'): type of the
                   revision added
                 - **directory** (:class:`sha1_git`): the directory the
                   revision points at
                 - **message** (:class:`bytes`): the message associated with
                   the revision
                 - **author** (:class:`Dict[str, bytes]`): dictionary with
                   keys: name, fullname, email
                 - **committer** (:class:`Dict[str, bytes]`): dictionary with
                   keys: name, fullname, email
                 - **metadata** (:class:`jsonb`): extra information as
                   dictionary
                 - **synthetic** (:class:`bool`): revision's nature (tarball,
                   directory creates synthetic revision`)
                 - **parents** (:class:`list[sha1_git]`): the parents of
                   this revision
 
         date dictionaries have the form defined in :mod:`swh.model`.
 
         Returns:
             Summary dict of keys with associated count as values
 
                 revision_added: New objects actually stored in db
 
         """
         if self.journal_writer:
             self.journal_writer.write_additions(
                 'revision',
                 (rev for rev in revisions
                  if rev['id'] not in self._revisions))
 
         revisions = [Revision.from_dict(rev) for rev in revisions]
 
         count = 0
         for revision in revisions:
             if revision.id not in self._revisions:
                 revision = attr.evolve(
                     revision,
                     committer=self._person_add(revision.committer),
                     author=self._person_add(revision.author))
                 self._revisions[revision.id] = revision
                 self._objects[revision.id].append(
                     ('revision', revision.id))
                 count += 1
 
         return {'revision:add': count}
 
     def revision_missing(self, revisions):
         """List revisions missing from storage
 
         Args:
             revisions (iterable): revision ids
 
         Yields:
             missing revision ids
 
         """
         for id in revisions:
             if id not in self._revisions:
                 yield id
 
     def revision_get(self, revisions):
         for id in revisions:
             if id in self._revisions:
                 yield self._revisions.get(id).to_dict()
             else:
                 yield None
 
     def _get_parent_revs(self, rev_id, seen, limit):
         if limit and len(seen) >= limit:
             return
         if rev_id in seen or rev_id not in self._revisions:
             return
         seen.add(rev_id)
         yield self._revisions[rev_id].to_dict()
         for parent in self._revisions[rev_id].parents:
             yield from self._get_parent_revs(parent, seen, limit)
 
     def revision_log(self, revisions, limit=None):
         """Fetch revision entry from the given root revisions.
 
         Args:
             revisions: array of root revision to lookup
             limit: limitation on the output result. Default to None.
 
         Yields:
             List of revision log from such revisions root.
 
         """
         seen = set()
         for rev_id in revisions:
             yield from self._get_parent_revs(rev_id, seen, limit)
 
     def revision_shortlog(self, revisions, limit=None):
         """Fetch the shortlog for the given revisions
 
         Args:
             revisions: list of root revisions to lookup
             limit: depth limitation for the output
 
         Yields:
             a list of (id, parents) tuples.
 
         """
         yield from ((rev['id'], rev['parents'])
                     for rev in self.revision_log(revisions, limit))
 
     def release_add(self, releases):
         """Add releases to the storage
 
         Args:
             releases (Iterable[dict]): iterable of dictionaries representing
                 the individual releases to add. Each dict has the following
                 keys:
 
                 - **id** (:class:`sha1_git`): id of the release to add
                 - **revision** (:class:`sha1_git`): id of the revision the
                   release points to
                 - **date** (:class:`dict`): the date the release was made
                 - **name** (:class:`bytes`): the name of the release
                 - **comment** (:class:`bytes`): the comment associated with
                   the release
                 - **author** (:class:`Dict[str, bytes]`): dictionary with
                   keys: name, fullname, email
 
         the date dictionary has the form defined in :mod:`swh.model`.
 
         Returns:
             Summary dict of keys with associated count as values
 
                 release:add: New objects contents actually stored in db
 
         """
         if self.journal_writer:
             self.journal_writer.write_additions(
                 'release',
                 (rel for rel in releases
                  if rel['id'] not in self._releases))
 
         releases = [Release.from_dict(rel) for rel in releases]
 
         count = 0
         for rel in releases:
             if rel.id not in self._releases:
                 if rel.author:
                     self._person_add(rel.author)
                 self._objects[rel.id].append(
                     ('release', rel.id))
                 self._releases[rel.id] = rel
                 count += 1
 
         return {'release:add': count}
 
     def release_missing(self, releases):
         """List releases missing from storage
 
         Args:
             releases: an iterable of release ids
 
         Returns:
             a list of missing release ids
 
         """
         yield from (rel for rel in releases if rel not in self._releases)
 
     def release_get(self, releases):
         """Given a list of sha1, return the releases's information
 
         Args:
             releases: list of sha1s
 
         Yields:
             dicts with the same keys as those given to `release_add`
             (or ``None`` if a release does not exist)
 
         """
         for rel_id in releases:
             if rel_id in self._releases:
                 yield self._releases[rel_id].to_dict()
             else:
                 yield None
 
     def snapshot_add(self, snapshots):
         """Add a snapshot to the storage
 
         Args:
             snapshot ([dict]): the snapshots to add, containing the
               following keys:
 
               - **id** (:class:`bytes`): id of the snapshot
               - **branches** (:class:`dict`): branches the snapshot contains,
                 mapping the branch name (:class:`bytes`) to the branch target,
                 itself a :class:`dict` (or ``None`` if the branch points to an
                 unknown object)
 
                 - **target_type** (:class:`str`): one of ``content``,
                   ``directory``, ``revision``, ``release``,
                   ``snapshot``, ``alias``
                 - **target** (:class:`bytes`): identifier of the target
                   (currently a ``sha1_git`` for all object kinds, or the name
                   of the target branch for aliases)
 
         Raises:
             ValueError: if the origin's or visit's identifier does not exist.
 
         Returns:
             Summary dict of keys with associated count as values
 
                 snapshot_added: Count of object actually stored in db
 
         """
         count = 0
         snapshots = (Snapshot.from_dict(d) for d in snapshots)
         snapshots = (snap for snap in snapshots
                      if snap.id not in self._snapshots)
         for snapshot in snapshots:
             if self.journal_writer:
                 self.journal_writer.write_addition('snapshot', snapshot)
 
             sorted_branch_names = sorted(snapshot.branches)
             self._snapshots[snapshot.id] = (snapshot, sorted_branch_names)
             self._objects[snapshot.id].append(('snapshot', snapshot.id))
             count += 1
 
         return {'snapshot:add': count}
 
     def snapshot_get(self, snapshot_id):
         """Get the content, possibly partial, of a snapshot with the given id
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         .. warning:: At most 1000 branches contained in the snapshot will be
             returned for performance reasons. In order to browse the whole
             set of branches, the method :meth:`snapshot_get_branches`
             should be used instead.
 
         Args:
             snapshot_id (bytes): identifier of the snapshot
         Returns:
             dict: a dict with three keys:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than 1000
                   branches.
         """
         return self.snapshot_get_branches(snapshot_id)
 
     def snapshot_get_by_origin_visit(self, origin, visit):
         """Get the content, possibly partial, of a snapshot for the given origin visit
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         .. warning:: At most 1000 branches contained in the snapshot will be
             returned for performance reasons. In order to browse the whole
             set of branches, the method :meth:`snapshot_get_branches`
             should be used instead.
 
         Args:
             origin (int): the origin's identifier
             visit (int): the visit's identifier
         Returns:
             dict: None if the snapshot does not exist;
               a dict with three keys otherwise:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than 1000
                   branches.
 
         """
         origin_url = self._get_origin_url(origin)
         if not origin_url:
             return
 
         if origin_url not in self._origins or \
            visit > len(self._origin_visits[origin_url]):
             return None
         snapshot_id = self._origin_visits[origin_url][visit-1].snapshot
         if snapshot_id:
             return self.snapshot_get(snapshot_id)
         else:
             return None
 
     def snapshot_get_latest(self, origin, allowed_statuses=None):
         """Get the content, possibly partial, of the latest snapshot for the
         given origin, optionally only from visits that have one of the given
         allowed_statuses
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         .. warning:: At most 1000 branches contained in the snapshot will be
             returned for performance reasons. In order to browse the whole
             set of branches, the methods :meth:`origin_visit_get_latest`
             and :meth:`snapshot_get_branches` should be used instead.
 
         Args:
             origin (str): the origin's URL
             allowed_statuses (list of str): list of visit statuses considered
                 to find the latest snapshot for the origin. For instance,
                 ``allowed_statuses=['full']`` will only consider visits that
                 have successfully run to completion.
         Returns:
             dict: a dict with three keys:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than 1000
                   branches.
         """
         origin_url = self._get_origin_url(origin)
         if not origin_url:
             return
 
         visit = self.origin_visit_get_latest(
             origin_url,
             allowed_statuses=allowed_statuses,
             require_snapshot=True)
         if visit and visit['snapshot']:
             snapshot = self.snapshot_get(visit['snapshot'])
             if not snapshot:
                 raise ValueError(
                     'last origin visit references an unknown snapshot')
             return snapshot
 
     def snapshot_count_branches(self, snapshot_id, db=None, cur=None):
         """Count the number of branches in the snapshot with the given id
 
         Args:
             snapshot_id (bytes): identifier of the snapshot
 
         Returns:
             dict: A dict whose keys are the target types of branches and
             values their corresponding amount
         """
         (snapshot, _) = self._snapshots[snapshot_id]
         return collections.Counter(branch.target_type.value if branch else None
                                    for branch in snapshot.branches.values())
 
     def snapshot_get_branches(self, snapshot_id, branches_from=b'',
                               branches_count=1000, target_types=None):
         """Get the content, possibly partial, of a snapshot with the given id
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         Args:
             snapshot_id (bytes): identifier of the snapshot
             branches_from (bytes): optional parameter used to skip branches
                 whose name is lesser than it before returning them
             branches_count (int): optional parameter used to restrain
                 the amount of returned branches
             target_types (list): optional parameter used to filter the
                 target types of branch to return (possible values that can be
                 contained in that list are `'content', 'directory',
                 'revision', 'release', 'snapshot', 'alias'`)
         Returns:
             dict: None if the snapshot does not exist;
               a dict with three keys otherwise:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than
                   `branches_count` branches after `branches_from` included.
         """
         res = self._snapshots.get(snapshot_id)
         if res is None:
             return None
         (snapshot, sorted_branch_names) = res
         from_index = bisect.bisect_left(
                 sorted_branch_names, branches_from)
         if target_types:
             next_branch = None
             branches = {}
             for branch_name in sorted_branch_names[from_index:]:
                 branch = snapshot.branches[branch_name]
                 if branch and branch.target_type.value in target_types:
                     if len(branches) < branches_count:
                         branches[branch_name] = branch
                     else:
                         next_branch = branch_name
                         break
         else:
             # As there is no 'target_types', we can do that much faster
             to_index = from_index + branches_count
             returned_branch_names = sorted_branch_names[from_index:to_index]
             branches = {branch_name: snapshot.branches[branch_name]
                         for branch_name in returned_branch_names}
             if to_index >= len(sorted_branch_names):
                 next_branch = None
             else:
                 next_branch = sorted_branch_names[to_index]
 
         branches = {name: branch.to_dict() if branch else None
                     for (name, branch) in branches.items()}
 
         return {
                 'id': snapshot_id,
                 'branches': branches,
                 'next_branch': next_branch,
                 }
 
     def object_find_by_sha1_git(self, ids, db=None, cur=None):
         """Return the objects found with the given ids.
 
         Args:
             ids: a generator of sha1_gits
 
         Returns:
             dict: a mapping from id to the list of objects found. Each object
             found is itself a dict with keys:
 
             - sha1_git: the input id
             - type: the type of object found
             - id: the id of the object found
             - object_id: the numeric id of the object found.
 
         """
         ret = {}
         for id_ in ids:
             objs = self._objects.get(id_, [])
             ret[id_] = [{
                     'sha1_git': id_,
                     'type': obj[0],
                     'id': obj[1],
                     'object_id': id_,
                     } for obj in objs]
         return ret
 
     def _convert_origin(self, t):
         if t is None:
             return None
 
         return t.to_dict()
 
     def origin_get(self, origins):
         """Return origins, either all identified by their ids or all
         identified by urls.
 
         Args:
             origin: a list of dictionaries representing the individual
                 origins to find.
                 These dicts have either the key url (and optionally type):
 
                 - url (bytes): the url the origin points to
 
                 or the id:
 
                 - id (int): the origin's identifier
 
         Returns:
             dict: the origin dictionary with the keys:
 
             - id: origin's id
             - url: origin's url
 
         Raises:
             ValueError: if the keys does not match (url and type) nor id.
 
         """
         if isinstance(origins, dict):
             # Old API
             return_single = True
             origins = [origins]
         else:
             return_single = False
 
         # Sanity check to be error-compatible with the pgsql backend
         if any('id' in origin for origin in origins) \
                 and not all('id' in origin for origin in origins):
             raise ValueError(
                 'Either all origins or none at all should have an "id".')
         if any('url' in origin for origin in origins) \
                 and not all('url' in origin for origin in origins):
             raise ValueError(
                 'Either all origins or none at all should have '
                 'an "url" key.')
 
         results = []
         for origin in origins:
             result = None
             if 'url' in origin:
                 if origin['url'] in self._origins:
                     result = self._origins[origin['url']]
             else:
                 raise ValueError(
                     'Origin must have an url.')
             results.append(self._convert_origin(result))
 
         if return_single:
             assert len(results) == 1
             return results[0]
         else:
             return results
 
     def origin_get_by_sha1(self, sha1s):
         """Return origins, identified by the sha1 of their URLs.
 
         Args:
             sha1s (list[bytes]): a list of sha1s
 
         Yields:
             dicts containing origin information as returned
             by :meth:`swh.storage.in_memory.Storage.origin_get`, or None if an
             origin matching the sha1 is not found.
         """
         return [
             self._convert_origin(self._origins_by_sha1.get(sha1))
             for sha1 in sha1s
         ]
 
-    def _select_random_origin_by_type(self, type: str) -> str:
-        """Select randomly an origin visit """
-        while True:
-            url = random.choice(list(self._origin_visits.keys()))
-            random_origin_visits = self._origin_visits[url]
-            if random_origin_visits[0].type == type:
-                return url
-
-    def origin_visit_get_random(self, type: str) -> Mapping[str, Any]:
-        """Randomly select one origin with <type> whose visit was successful
-        in the last 3 months.
-
-        Returns:
-            origin dict selected randomly on the dataset
-
-        """
-        random_visit: Dict[str, Any] = {}
-        if not self._origin_visits:  # empty dataset
-            return random_visit
-        url = self._select_random_origin_by_type(type)
-        random_origin_visits = copy.deepcopy(self._origin_visits[url])
-        random_origin_visits.reverse()
-        back_in_the_day = now() - timedelta(weeks=12)  # 3 months back
-        # This should be enough for tests
-        for visit in random_origin_visits:
-            if visit.date > back_in_the_day and visit.status == 'full':
-                random_visit = visit.to_dict()
-                break
-        return random_visit
-
     def origin_get_range(self, origin_from=1, origin_count=100):
         """Retrieve ``origin_count`` origins whose ids are greater
         or equal than ``origin_from``.
 
         Origins are sorted by id before retrieving them.
 
         Args:
             origin_from (int): the minimum id of origins to retrieve
             origin_count (int): the maximum number of origins to retrieve
 
         Yields:
             dicts containing origin information as returned
             by :meth:`swh.storage.in_memory.Storage.origin_get`, plus
             an 'id' key.
         """
         origin_from = max(origin_from, 1)
         if origin_from <= len(self._origins_by_id):
             max_idx = origin_from + origin_count - 1
             if max_idx > len(self._origins_by_id):
                 max_idx = len(self._origins_by_id)
             for idx in range(origin_from-1, max_idx):
                 origin = self._convert_origin(
                     self._origins[self._origins_by_id[idx]])
                 yield {'id': idx+1, **origin}
 
     def origin_search(self, url_pattern, offset=0, limit=50,
                       regexp=False, with_visit=False, db=None, cur=None):
         """Search for origins whose urls contain a provided string pattern
         or match a provided regular expression.
         The search is performed in a case insensitive way.
 
         Args:
             url_pattern (str): the string pattern to search for in origin urls
             offset (int): number of found origins to skip before returning
                 results
             limit (int): the maximum number of found origins to return
             regexp (bool): if True, consider the provided pattern as a regular
                 expression and return origins whose urls match it
             with_visit (bool): if True, filter out origins with no visit
 
         Returns:
             An iterable of dict containing origin information as returned
             by :meth:`swh.storage.storage.Storage.origin_get`.
         """
         origins = map(self._convert_origin, self._origins.values())
         if regexp:
             pat = re.compile(url_pattern)
             origins = [orig for orig in origins if pat.search(orig['url'])]
         else:
             origins = [orig for orig in origins if url_pattern in orig['url']]
         if with_visit:
             origins = [
                 orig for orig in origins
                 if len(self._origin_visits[orig['url']]) > 0 and
                 set(ov.snapshot
                     for ov in self._origin_visits[orig['url']]
                     if ov.snapshot) &
                 set(self._snapshots)]
 
         return origins[offset:offset+limit]
 
     def origin_count(self, url_pattern, regexp=False, with_visit=False,
                      db=None, cur=None):
         """Count origins whose urls contain a provided string pattern
         or match a provided regular expression.
         The pattern search in origin urls is performed in a case insensitive
         way.
 
         Args:
             url_pattern (str): the string pattern to search for in origin urls
             regexp (bool): if True, consider the provided pattern as a regular
                 expression and return origins whose urls match it
             with_visit (bool): if True, filter out origins with no visit
 
         Returns:
             int: The number of origins matching the search criterion.
         """
         return len(self.origin_search(url_pattern, regexp=regexp,
                                       with_visit=with_visit,
                                       limit=len(self._origins)))
 
     def origin_add(self, origins):
         """Add origins to the storage
 
         Args:
             origins: list of dictionaries representing the individual origins,
                 with the following keys:
 
                 - url (bytes): the url the origin points to
 
         Returns:
             list: given origins as dict updated with their id
 
         """
         origins = copy.deepcopy(origins)
         for origin in origins:
             self.origin_add_one(origin)
         return origins
 
     def origin_add_one(self, origin):
         """Add origin to the storage
 
         Args:
             origin: dictionary representing the individual origin to add. This
                 dict has the following keys:
 
                 - url (bytes): the url the origin points to
 
         Returns:
             the id of the added origin, or of the identical one that already
             exists.
 
         """
         origin = Origin.from_dict(origin)
         if origin.url not in self._origins:
             if self.journal_writer:
                 self.journal_writer.write_addition('origin', origin)
 
             # generate an origin_id because it is needed by origin_get_range.
             # TODO: remove this when we remove origin_get_range
             origin_id = len(self._origins) + 1
             self._origins_by_id.append(origin.url)
             assert len(self._origins_by_id) == origin_id
 
             self._origins[origin.url] = origin
             self._origins_by_sha1[origin_url_to_sha1(origin.url)] = origin
             self._origin_visits[origin.url] = []
             self._objects[origin.url].append(('origin', origin.url))
 
         return origin.url
 
     def origin_visit_add(self, origin, date, type):
         """Add an origin_visit for the origin at date with status 'ongoing'.
 
         Args:
             origin (str): visited origin's identifier or URL
             date (Union[str,datetime]): timestamp of such visit
             type (str): the type of loader used for the visit (hg, git, ...)
 
         Returns:
             dict: dictionary with keys origin and visit where:
 
             - origin: origin's identifier
             - visit: the visit's identifier for the new visit occurrence
 
         """
         origin_url = origin
         if origin_url is None:
             raise ValueError('Unknown origin.')
 
         if isinstance(date, str):
             # FIXME: Converge on iso8601 at some point
             date = dateutil.parser.parse(date)
         elif not isinstance(date, datetime.datetime):
             raise TypeError('date must be a datetime or a string.')
 
         visit_ret = None
         if origin_url in self._origins:
             origin = self._origins[origin_url]
             # visit ids are in the range [1, +inf[
             visit_id = len(self._origin_visits[origin_url]) + 1
             status = 'ongoing'
             visit = OriginVisit(
                 origin=origin.url,
                 date=date,
                 type=type,
                 status=status,
                 snapshot=None,
                 metadata=None,
                 visit=visit_id,
             )
             self._origin_visits[origin_url].append(visit)
             visit_ret = {
                 'origin': origin.url,
                 'visit': visit_id,
             }
 
             self._objects[(origin_url, visit_id)].append(
                 ('origin_visit', None))
 
             if self.journal_writer:
                 self.journal_writer.write_addition('origin_visit', visit)
 
         return visit_ret
 
     def origin_visit_update(self, origin, visit_id, status=None,
                             metadata=None, snapshot=None):
         """Update an origin_visit's status.
 
         Args:
             origin (str): visited origin's URL
             visit_id (int): visit's identifier
             status: visit's new status
             metadata: data associated to the visit
             snapshot (sha1_git): identifier of the snapshot to add to
                 the visit
 
         Returns:
             None
 
         """
         if not isinstance(origin, str):
             raise TypeError('origin must be a string, not %r' % (origin,))
         origin_url = self._get_origin_url(origin)
         if origin_url is None:
             raise ValueError('Unknown origin.')
 
         try:
             visit = self._origin_visits[origin_url][visit_id-1]
         except IndexError:
             raise ValueError('Unknown visit_id for this origin') \
                 from None
 
         updates = {}
         if status:
             updates['status'] = status
         if metadata:
             updates['metadata'] = metadata
         if snapshot:
             updates['snapshot'] = snapshot
 
         visit = attr.evolve(visit, **updates)
 
         if self.journal_writer:
             self.journal_writer.write_update('origin_visit', visit)
 
         self._origin_visits[origin_url][visit_id-1] = visit
 
     def origin_visit_upsert(self, visits):
         """Add a origin_visits with a specific id and with all its data.
         If there is already an origin_visit with the same
         `(origin_url, visit_id)`, updates it instead of inserting a new one.
 
         Args:
             visits: iterable of dicts with keys:
 
                 origin: origin url
                 visit: origin visit id
                 type: type of loader used for the visit
                 date: timestamp of such visit
                 status: Visit's new status
                 metadata: Data associated to the visit
                 snapshot (sha1_git): identifier of the snapshot to add to
                     the visit
         """
         for visit in visits:
             if not isinstance(visit['origin'], str):
                 raise TypeError("visit['origin'] must be a string, not %r"
                                 % (visit['origin'],))
         visits = [OriginVisit.from_dict(d) for d in visits]
 
         if self.journal_writer:
             for visit in visits:
                 self.journal_writer.write_addition('origin_visit', visit)
 
         for visit in visits:
             visit_id = visit.visit
             origin_url = visit.origin
 
             visit = attr.evolve(visit, origin=origin_url)
 
             self._objects[(origin_url, visit_id)].append(
                 ('origin_visit', None))
 
             while len(self._origin_visits[origin_url]) <= visit_id:
                 self._origin_visits[origin_url].append(None)
 
             self._origin_visits[origin_url][visit_id-1] = visit
 
     def _convert_visit(self, visit):
         if visit is None:
             return
 
         visit = visit.to_dict()
 
         return visit
 
     def origin_visit_get(self, origin, last_visit=None, limit=None):
         """Retrieve all the origin's visit's information.
 
         Args:
             origin (int): the origin's identifier
             last_visit (int): visit's id from which listing the next ones,
                 default to None
             limit (int): maximum number of results to return,
                 default to None
 
         Yields:
             List of visits.
 
         """
         origin_url = self._get_origin_url(origin)
         if origin_url in self._origin_visits:
             visits = self._origin_visits[origin_url]
             if last_visit is not None:
                 visits = visits[last_visit:]
             if limit is not None:
                 visits = visits[:limit]
             for visit in visits:
                 if not visit:
                     continue
                 visit_id = visit.visit
 
                 yield self._convert_visit(
                     self._origin_visits[origin_url][visit_id-1])
 
     def origin_visit_find_by_date(self, origin, visit_date):
         """Retrieves the origin visit whose date is closest to the provided
         timestamp.
         In case of a tie, the visit with largest id is selected.
 
         Args:
             origin (str): The occurrence's origin (URL).
             target (datetime): target timestamp
 
         Returns:
             A visit.
 
         """
         origin_url = self._get_origin_url(origin)
         if origin_url in self._origin_visits:
             visits = self._origin_visits[origin_url]
             visit = min(
                 visits,
                 key=lambda v: (abs(v.date - visit_date), -v.visit))
             return self._convert_visit(visit)
 
     def origin_visit_get_by(self, origin, visit):
         """Retrieve origin visit's information.
 
         Args:
             origin (int): the origin's identifier
 
         Returns:
             The information on that particular (origin, visit) or None if
             it does not exist
 
         """
         origin_url = self._get_origin_url(origin)
         if origin_url in self._origin_visits and \
            visit <= len(self._origin_visits[origin_url]):
             return self._convert_visit(
                 self._origin_visits[origin_url][visit-1])
 
     def origin_visit_get_latest(
             self, origin, allowed_statuses=None, require_snapshot=False):
         """Get the latest origin visit for the given origin, optionally
         looking only for those with one of the given allowed_statuses
         or for those with a known snapshot.
 
         Args:
             origin (str): the origin's URL
             allowed_statuses (list of str): list of visit statuses considered
                 to find the latest visit. For instance,
                 ``allowed_statuses=['full']`` will only consider visits that
                 have successfully run to completion.
             require_snapshot (bool): If True, only a visit with a snapshot
                 will be returned.
         Returns:
             dict: a dict with the following keys:
 
                 origin: the URL of the origin
                 visit: origin visit id
                 type: type of loader used for the visit
                 date: timestamp of such visit
                 status: Visit's new status
                 metadata: Data associated to the visit
                 snapshot (Optional[sha1_git]): identifier of the snapshot
                     associated to the visit
         """
         origin = self._origins.get(origin)
         if not origin:
             return
         visits = self._origin_visits[origin.url]
         if allowed_statuses is not None:
             visits = [visit for visit in visits
                       if visit.status in allowed_statuses]
         if require_snapshot:
             visits = [visit for visit in visits
                       if visit.snapshot]
 
         visit = max(
             visits, key=lambda v: (v.date, v.visit), default=None)
         return self._convert_visit(visit)
 
+    def _select_random_origin_visit_by_type(self, type: str) -> str:
+        """Select randomly an origin visit """
+        while True:
+            url = random.choice(list(self._origin_visits.keys()))
+            random_origin_visits = self._origin_visits[url]
+            if random_origin_visits[0].type == type:
+                return url
+
+    def origin_visit_get_random(self, type: str) -> Mapping[str, Any]:
+        """Randomly select one successful origin visit with <type>
+        made in the last 3 months.
+
+        Returns:
+            dict representing an origin visit, in the same format as
+            `origin_visit_get`.
+
+        """
+        random_visit: Dict[str, Any] = {}
+        if not self._origin_visits:  # empty dataset
+            return random_visit
+        url = self._select_random_origin_visit_by_type(type)
+        random_origin_visits = copy.deepcopy(self._origin_visits[url])
+        random_origin_visits.reverse()
+        back_in_the_day = now() - timedelta(weeks=12)  # 3 months back
+        # This should be enough for tests
+        for visit in random_origin_visits:
+            if visit.date > back_in_the_day and visit.status == 'full':
+                random_visit = visit.to_dict()
+                break
+        return random_visit
+
     def stat_counters(self):
         """compute statistics about the number of tuples in various tables
 
         Returns:
             dict: a dictionary mapping textual labels (e.g., content) to
             integer values (e.g., the number of tuples in table content)
 
         """
         keys = (
             'content',
             'directory',
             'origin',
             'origin_visit',
             'person',
             'release',
             'revision',
             'skipped_content',
             'snapshot'
             )
         stats = {key: 0 for key in keys}
         stats.update(collections.Counter(
             obj_type
             for (obj_type, obj_id)
             in itertools.chain(*self._objects.values())))
         return stats
 
     def refresh_stat_counters(self):
         """Recomputes the statistics for `stat_counters`."""
         pass
 
     def origin_metadata_add(self, origin_url, ts, provider, tool, metadata,
                             db=None, cur=None):
         """ Add an origin_metadata for the origin at ts with provenance and
         metadata.
 
         Args:
             origin_url (str): the origin url for which the metadata is added
             ts (datetime): timestamp of the found metadata
             provider: id of the provider of metadata (ex:'hal')
             tool: id of the tool used to extract metadata
             metadata (jsonb): the metadata retrieved at the time and location
         """
         if not isinstance(origin_url, str):
             raise TypeError('origin_id must be str, not %r' % (origin_url,))
 
         if isinstance(ts, str):
             ts = dateutil.parser.parse(ts)
 
         origin_metadata = {
                 'origin_url': origin_url,
                 'discovery_date': ts,
                 'tool_id': tool,
                 'metadata': metadata,
                 'provider_id': provider,
                 }
         self._origin_metadata[origin_url].append(origin_metadata)
         return None
 
     def origin_metadata_get_by(self, origin_url, provider_type=None, db=None,
                                cur=None):
         """Retrieve list of all origin_metadata entries for the origin_url
 
         Args:
             origin_url (str): the origin's url
             provider_type (str): (optional) type of provider
 
         Returns:
             list of dicts: the origin_metadata dictionary with the keys:
 
             - origin_url (int): origin's URL
             - discovery_date (datetime): timestamp of discovery
             - tool_id (int): metadata's extracting tool
             - metadata (jsonb)
             - provider_id (int): metadata's provider
             - provider_name (str)
             - provider_type (str)
             - provider_url (str)
 
         """
         if not isinstance(origin_url, str):
             raise TypeError('origin_url must be str, not %r' % (origin_url,))
         metadata = []
         for item in self._origin_metadata[origin_url]:
             item = copy.deepcopy(item)
             provider = self.metadata_provider_get(item['provider_id'])
             for attr_name in ('name', 'type', 'url'):
                 item['provider_' + attr_name] = \
                     provider['provider_' + attr_name]
             metadata.append(item)
         return metadata
 
     def tool_add(self, tools):
         """Add new tools to the storage.
 
         Args:
             tools (iterable of :class:`dict`): Tool information to add to
               storage. Each tool is a :class:`dict` with the following keys:
 
               - name (:class:`str`): name of the tool
               - version (:class:`str`): version of the tool
               - configuration (:class:`dict`): configuration of the tool,
                 must be json-encodable
 
         Returns:
             :class:`dict`: All the tools inserted in storage
             (including the internal ``id``). The order of the list is not
             guaranteed to match the order of the initial list.
 
         """
         inserted = []
         for tool in tools:
             key = self._tool_key(tool)
             assert 'id' not in tool
             record = copy.deepcopy(tool)
             record['id'] = key  # TODO: remove this
             if key not in self._tools:
                 self._tools[key] = record
             inserted.append(copy.deepcopy(self._tools[key]))
 
         return inserted
 
     def tool_get(self, tool):
         """Retrieve tool information.
 
         Args:
             tool (dict): Tool information we want to retrieve from storage.
               The dicts have the same keys as those used in :func:`tool_add`.
 
         Returns:
             dict: The full tool information if it exists (``id`` included),
             None otherwise.
 
         """
         return self._tools.get(self._tool_key(tool))
 
     def metadata_provider_add(self, provider_name, provider_type, provider_url,
                               metadata):
         """Add a metadata provider.
 
         Args:
             provider_name (str): Its name
             provider_type (str): Its type
             provider_url (str): Its URL
             metadata: JSON-encodable object
 
         Returns:
             an identifier of the provider
         """
         provider = {
                 'provider_name': provider_name,
                 'provider_type': provider_type,
                 'provider_url': provider_url,
                 'metadata': metadata,
                 }
         key = self._metadata_provider_key(provider)
         provider['id'] = key
         self._metadata_providers[key] = provider
         return key
 
     def metadata_provider_get(self, provider_id, db=None, cur=None):
         """Get a metadata provider
 
         Args:
             provider_id: Its identifier, as given by `metadata_provider_add`.
 
         Returns:
             dict: same as `metadata_provider_add`;
                   or None if it does not exist.
         """
         return self._metadata_providers.get(provider_id)
 
     def metadata_provider_get_by(self, provider, db=None, cur=None):
         """Get a metadata provider
 
         Args:
             provider_name: Its name
             provider_url: Its URL
 
         Returns:
             dict: same as `metadata_provider_add`;
                   or None if it does not exist.
         """
         key = self._metadata_provider_key(provider)
         return self._metadata_providers.get(key)
 
     def _get_origin_url(self, origin):
         if isinstance(origin, str):
             return origin
         else:
             raise TypeError('origin must be a string.')
 
     def _person_add(self, person):
         """Add a person in storage.
 
         Note: Private method, do not use outside of this class.
 
         Args:
             person: dictionary with keys fullname, name and email.
 
         """
         key = ('person', person.fullname)
         if key not in self._objects:
             person_id = len(self._persons) + 1
             self._persons.append(person)
             self._objects[key].append(('person', person_id))
         else:
             person_id = self._objects[key][0][1]
             person = self._persons[person_id-1]
         return person
 
     @staticmethod
     def _content_key(content):
         """A stable key for a content"""
         return tuple(getattr(content, key)
                      for key in sorted(DEFAULT_ALGORITHMS))
 
     @staticmethod
     def _content_key_algorithm(content):
         """ A stable key and the algorithm for a content"""
         if isinstance(content, Content):
             content = content.to_dict()
         return tuple((content.get(key), key)
                      for key in sorted(DEFAULT_ALGORITHMS))
 
     @staticmethod
     def _tool_key(tool):
         return '%r %r %r' % (tool['name'], tool['version'],
                              tuple(sorted(tool['configuration'].items())))
 
     @staticmethod
     def _metadata_provider_key(provider):
         return '%r %r' % (provider['provider_name'], provider['provider_url'])
diff --git a/swh/storage/storage.py b/swh/storage/storage.py
index f61f3bdd..db1aef94 100644
--- a/swh/storage/storage.py
+++ b/swh/storage/storage.py
@@ -1,1990 +1,1992 @@
 # Copyright (C) 2015-2019  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 copy
 import datetime
 import itertools
 import json
 
 from collections import defaultdict
 from concurrent.futures import ThreadPoolExecutor
 from contextlib import contextmanager
 from typing import Any, Dict, Mapping
 
 import dateutil.parser
 import psycopg2
 import psycopg2.pool
 
 from swh.core.api import remote_api_endpoint
 from swh.model.hashutil import ALGORITHMS, hash_to_bytes
 from swh.objstorage import get_objstorage
 from swh.objstorage.exc import ObjNotFoundError
 try:
     from swh.journal.writer import get_journal_writer
 except ImportError:
     get_journal_writer = None  # type: ignore
     # mypy limitation, see https://github.com/python/mypy/issues/1153
 
 from . import converters
 from .common import db_transaction_generator, db_transaction
 from .db import Db
 from .exc import StorageDBError
 from .algos import diff
 from .metrics import timed, send_metric, process_metrics
 
 
 # Max block size of contents to return
 BULK_BLOCK_CONTENT_LEN_MAX = 10000
 
 EMPTY_SNAPSHOT_ID = hash_to_bytes('1a8893e6a86f444e8be8e7bda6cb34fb1735a00e')
 """Identifier for the empty snapshot"""
 
 
 class Storage():
     """SWH storage proxy, encompassing DB and object storage
 
     """
 
     def __init__(self, db, objstorage, min_pool_conns=1, max_pool_conns=10,
                  journal_writer=None):
         """
         Args:
             db_conn: either a libpq connection string, or a psycopg2 connection
             obj_root: path to the root of the object storage
 
         """
         try:
             if isinstance(db, psycopg2.extensions.connection):
                 self._pool = None
                 self._db = Db(db)
             else:
                 self._pool = psycopg2.pool.ThreadedConnectionPool(
                     min_pool_conns, max_pool_conns, db
                 )
                 self._db = None
         except psycopg2.OperationalError as e:
             raise StorageDBError(e)
 
         self.objstorage = get_objstorage(**objstorage)
         if journal_writer:
             if get_journal_writer is None:
                 raise EnvironmentError(
                     'You need the swh.journal package to use the '
                     'journal_writer feature')
             self.journal_writer = get_journal_writer(**journal_writer)
         else:
             self.journal_writer = None
 
     def get_db(self):
         if self._db:
             return self._db
         else:
             return Db.from_pool(self._pool)
 
     def put_db(self, db):
         if db is not self._db:
             db.put_conn()
 
     @contextmanager
     def db(self):
         db = None
         try:
             db = self.get_db()
             yield db
         finally:
             if db:
                 self.put_db(db)
 
     @remote_api_endpoint('check_config')
     @timed
     @db_transaction()
     def check_config(self, *, check_write, db=None, cur=None):
         """Check that the storage is configured and ready to go."""
 
         if not self.objstorage.check_config(check_write=check_write):
             return False
 
         # Check permissions on one of the tables
         if check_write:
             check = 'INSERT'
         else:
             check = 'SELECT'
 
         cur.execute(
             "select has_table_privilege(current_user, 'content', %s)",
             (check,)
         )
         return cur.fetchone()[0]
 
     def _content_unique_key(self, hash, db):
         """Given a hash (tuple or dict), return a unique key from the
            aggregation of keys.
 
         """
         keys = db.content_hash_keys
         if isinstance(hash, tuple):
             return hash
         return tuple([hash[k] for k in keys])
 
     @staticmethod
     def _normalize_content(d):
         d = d.copy()
 
         if 'status' not in d:
             d['status'] = 'visible'
 
         if 'length' not in d:
             d['length'] = -1
 
         return d
 
     @staticmethod
     def _validate_content(d):
         """Sanity checks on status / reason / length, that postgresql
         doesn't enforce."""
         if d['status'] not in ('visible', 'absent', 'hidden'):
             raise ValueError('Invalid content status: {}'.format(d['status']))
 
         if d['status'] != 'absent' and d.get('reason') is not None:
             raise ValueError(
                 'Must not provide a reason if content is not absent.')
 
         if d['length'] < -1:
             raise ValueError('Content length must be positive or -1.')
 
     def _filter_new_content(self, content, db=None, cur=None):
         """Sort contents into buckets 'with data' and 'without data',
         and filter out those already in the database."""
         content_by_status = defaultdict(list)
         for d in content:
             content_by_status[d['status']].append(d)
 
         content_with_data = content_by_status['visible'] \
             + content_by_status['hidden']
         content_without_data = content_by_status['absent']
 
         missing_content = set(self.content_missing(content_with_data,
                                                    db=db, cur=cur))
         missing_skipped = set(self._content_unique_key(hashes, db)
                               for hashes in self.skipped_content_missing(
                                   content_without_data, db=db, cur=cur))
 
         content_with_data = [
             cont for cont in content_with_data
             if cont['sha1'] in missing_content]
         content_without_data = [
             cont for cont in content_without_data
             if self._content_unique_key(cont, db) in missing_skipped]
 
         summary = {
             'content:add': len(missing_content),
             'skipped_content:add': len(missing_skipped),
         }
 
         return (content_with_data, content_without_data, summary)
 
     def _content_add_metadata(self, db, cur,
                               content_with_data, content_without_data):
         """Add content to the postgresql database but not the object storage.
         """
         if content_with_data:
             # create temporary table for metadata injection
             db.mktemp('content', cur)
 
             db.copy_to(content_with_data, 'tmp_content',
                        db.content_add_keys, cur)
 
             # move metadata in place
             try:
                 db.content_add_from_temp(cur)
             except psycopg2.IntegrityError as e:
                 from . import HashCollision
                 if e.diag.sqlstate == '23505' and \
                         e.diag.table_name == 'content':
                     constraint_to_hash_name = {
                         'content_pkey': 'sha1',
                         'content_sha1_git_idx': 'sha1_git',
                         'content_sha256_idx': 'sha256',
                         }
                     colliding_hash_name = constraint_to_hash_name \
                         .get(e.diag.constraint_name)
                     raise HashCollision(colliding_hash_name) from None
                 else:
                     raise
 
         if content_without_data:
             content_without_data = \
                 [cont.copy() for cont in content_without_data]
             origin_ids = db.origin_id_get_by_url(
                 [cont.get('origin') for cont in content_without_data],
                 cur=cur)
             for (cont, origin_id) in zip(content_without_data, origin_ids):
                 if 'origin' in cont:
                     cont['origin'] = origin_id
             db.mktemp('skipped_content', cur)
             db.copy_to(content_without_data, 'tmp_skipped_content',
                        db.skipped_content_keys, cur)
 
             # move metadata in place
             db.skipped_content_add_from_temp(cur)
 
     @remote_api_endpoint('content/add')
     @timed
     @process_metrics
     @db_transaction()
     def content_add(self, content, db=None, cur=None):
         """Add content blobs to the storage
 
         Note: in case of DB errors, objects might have already been added to
         the object storage and will not be removed. Since addition to the
         object storage is idempotent, that should not be a problem.
 
         Args:
             contents (iterable): iterable of dictionaries representing
                 individual pieces of content to add. Each dictionary has the
                 following keys:
 
                 - data (bytes): the actual content
                 - length (int): content length (default: -1)
                 - one key for each checksum algorithm in
                   :data:`swh.model.hashutil.ALGORITHMS`, mapped to the
                   corresponding checksum
                 - status (str): one of visible, hidden, absent
                 - reason (str): if status = absent, the reason why
                 - origin (int): if status = absent, the origin we saw the
                   content in
 
         Raises:
 
             In case of errors, nothing is stored in the db (in the
             objstorage, it could though). The following exceptions can
             occur:
 
             - HashCollision in case of collision
             - Any other exceptions raise by the db
 
         Returns:
             Summary dict with the following key and associated values:
 
                 content:add: New contents added
                 content:add:bytes: Sum of the contents' length data
                 skipped_content:add: New skipped contents (no data) added
         """
         content = [dict(c.items()) for c in content]  # semi-shallow copy
         now = datetime.datetime.now(tz=datetime.timezone.utc)
         for item in content:
             item['ctime'] = now
 
         content = [self._normalize_content(c) for c in content]
         for c in content:
             self._validate_content(c)
 
         (content_with_data, content_without_data, summary) = \
             self._filter_new_content(content, db, cur)
 
         if self.journal_writer:
             for item in content_with_data:
                 if 'data' in item:
                     item = item.copy()
                     del item['data']
                 self.journal_writer.write_addition('content', item)
             for item in content_without_data:
                 self.journal_writer.write_addition('content', item)
 
         def add_to_objstorage():
             """Add to objstorage the new missing_content
 
             Returns:
                 Sum of all the content's data length pushed to the
                 objstorage. Content present twice is only sent once.
 
             """
             content_bytes_added = 0
             data = {}
             for cont in content_with_data:
                 if cont['sha1'] not in data:
                     data[cont['sha1']] = cont['data']
                     content_bytes_added += max(0, cont['length'])
 
             # FIXME: Since we do the filtering anyway now, we might as
             # well make the objstorage's add_batch call return what we
             # want here (real bytes added)... that'd simplify this...
             self.objstorage.add_batch(data)
             return content_bytes_added
 
         with ThreadPoolExecutor(max_workers=1) as executor:
             added_to_objstorage = executor.submit(add_to_objstorage)
 
             self._content_add_metadata(
                 db, cur, content_with_data, content_without_data)
 
             # Wait for objstorage addition before returning from the
             # transaction, bubbling up any exception
             content_bytes_added = added_to_objstorage.result()
 
         summary['content:add:bytes'] = content_bytes_added
         return summary
 
     @remote_api_endpoint('content/update')
     @timed
     @db_transaction()
     def content_update(self, content, keys=[], db=None, cur=None):
         """Update content blobs to the storage. Does nothing for unknown
         contents or skipped ones.
 
         Args:
             content (iterable): iterable of dictionaries representing
                 individual pieces of content to update. Each dictionary has the
                 following keys:
 
                 - data (bytes): the actual content
                 - length (int): content length (default: -1)
                 - one key for each checksum algorithm in
                   :data:`swh.model.hashutil.ALGORITHMS`, mapped to the
                   corresponding checksum
                 - status (str): one of visible, hidden, absent
 
             keys (list): List of keys (str) whose values needs an update, e.g.,
                 new hash column
 
         """
         # TODO: Add a check on input keys. How to properly implement
         # this? We don't know yet the new columns.
 
         if self.journal_writer:
             raise NotImplementedError(
                 'content_update is not yet support with a journal_writer.')
 
         db.mktemp('content', cur)
         select_keys = list(set(db.content_get_metadata_keys).union(set(keys)))
         db.copy_to(content, 'tmp_content', select_keys, cur)
         db.content_update_from_temp(keys_to_update=keys,
                                     cur=cur)
 
     @remote_api_endpoint('content/add_metadata')
     @timed
     @process_metrics
     @db_transaction()
     def content_add_metadata(self, content, db=None, cur=None):
         """Add content metadata to the storage (like `content_add`, but
         without inserting to the objstorage).
 
         Args:
             content (iterable): iterable of dictionaries representing
                 individual pieces of content to add. Each dictionary has the
                 following keys:
 
                 - length (int): content length (default: -1)
                 - one key for each checksum algorithm in
                   :data:`swh.model.hashutil.ALGORITHMS`, mapped to the
                   corresponding checksum
                 - status (str): one of visible, hidden, absent
                 - reason (str): if status = absent, the reason why
                 - origin (int): if status = absent, the origin we saw the
                   content in
                 - ctime (datetime): time of insertion in the archive
 
         Returns:
             Summary dict with the following key and associated values:
 
                 content:add: New contents added
                 skipped_content:add: New skipped contents (no data) added
         """
 
         content = [self._normalize_content(c) for c in content]
         for c in content:
             self._validate_content(c)
 
         (content_with_data, content_without_data, summary) = \
             self._filter_new_content(content, db, cur)
 
         if self.journal_writer:
             for item in itertools.chain(content_with_data,
                                         content_without_data):
                 assert 'data' not in content
                 self.journal_writer.write_addition('content', item)
 
         self._content_add_metadata(
             db, cur, content_with_data, content_without_data)
 
         return summary
 
     @remote_api_endpoint('content/data')
     @timed
     def content_get(self, content):
         """Retrieve in bulk contents and their data.
 
         This generator yields exactly as many items than provided sha1
         identifiers, but callers should not assume this will always be true.
 
         It may also yield `None` values in case an object was not found.
 
         Args:
             content: iterables of sha1
 
         Yields:
             Dict[str, bytes]: Generates streams of contents as dict with their
                 raw data:
 
                 - sha1 (bytes): content id
                 - data (bytes): content's raw data
 
         Raises:
             ValueError in case of too much contents are required.
             cf. BULK_BLOCK_CONTENT_LEN_MAX
 
         """
         # FIXME: Make this method support slicing the `data`.
         if len(content) > BULK_BLOCK_CONTENT_LEN_MAX:
             raise ValueError(
                 "Send at maximum %s contents." % BULK_BLOCK_CONTENT_LEN_MAX)
 
         for obj_id in content:
             try:
                 data = self.objstorage.get(obj_id)
             except ObjNotFoundError:
                 yield None
                 continue
 
             yield {'sha1': obj_id, 'data': data}
 
     @remote_api_endpoint('content/range')
     @timed
     @db_transaction()
     def content_get_range(self, start, end, limit=1000, db=None, cur=None):
         """Retrieve contents within range [start, end] bound by limit.
 
         Note that this function may return more than one blob per hash. The
         limit is enforced with multiplicity (ie. two blobs with the same hash
         will count twice toward the limit).
 
         Args:
             **start** (bytes): Starting identifier range (expected smaller
                            than end)
             **end** (bytes): Ending identifier range (expected larger
                              than start)
             **limit** (int): Limit result (default to 1000)
 
         Returns:
             a dict with keys:
             - contents [dict]: iterable of contents in between the range.
             - next (bytes): There remains content in the range
               starting from this next sha1
 
         """
         if limit is None:
             raise ValueError('Development error: limit should not be None')
         contents = []
         next_content = None
         for counter, content_row in enumerate(
                 db.content_get_range(start, end, limit+1, cur)):
             content = dict(zip(db.content_get_metadata_keys, content_row))
             if counter >= limit:
                 # take the last commit for the next page starting from this
                 next_content = content['sha1']
                 break
             contents.append(content)
         return {
             'contents': contents,
             'next': next_content,
         }
 
     @remote_api_endpoint('content/metadata')
     @timed
     @db_transaction_generator(statement_timeout=500)
     def content_get_metadata(self, content, db=None, cur=None):
         """Retrieve content metadata in bulk
 
         Args:
             content: iterable of content identifiers (sha1)
 
         Returns:
             an iterable with content metadata corresponding to the given ids
         """
         for metadata in db.content_get_metadata_from_sha1s(content, cur):
             yield dict(zip(db.content_get_metadata_keys, metadata))
 
     @remote_api_endpoint('content/missing')
     @timed
     @db_transaction_generator()
     def content_missing(self, content, key_hash='sha1', db=None, cur=None):
         """List content missing from storage
 
         Args:
             content ([dict]): iterable of dictionaries whose keys are
                               either 'length' or an item of
                               :data:`swh.model.hashutil.ALGORITHMS`;
                               mapped to the corresponding checksum
                               (or length).
 
             key_hash (str): name of the column to use as hash id
                             result (default: 'sha1')
 
         Returns:
             iterable ([bytes]): missing content ids (as per the
             key_hash column)
 
         Raises:
             TODO: an exception when we get a hash collision.
 
         """
         keys = db.content_hash_keys
 
         if key_hash not in keys:
             raise ValueError("key_hash should be one of %s" % keys)
 
         key_hash_idx = keys.index(key_hash)
 
         if not content:
             return
 
         for obj in db.content_missing_from_list(content, cur):
             yield obj[key_hash_idx]
 
     @remote_api_endpoint('content/missing/sha1')
     @timed
     @db_transaction_generator()
     def content_missing_per_sha1(self, contents, db=None, cur=None):
         """List content missing from storage based only on sha1.
 
         Args:
             contents: Iterable of sha1 to check for absence.
 
         Returns:
             iterable: missing ids
 
         Raises:
             TODO: an exception when we get a hash collision.
 
         """
         for obj in db.content_missing_per_sha1(contents, cur):
             yield obj[0]
 
     @remote_api_endpoint('content/skipped/missing')
     @timed
     @db_transaction_generator()
     def skipped_content_missing(self, contents, db=None, cur=None):
         """List skipped_content missing from storage
 
         Args:
             content: iterable of dictionaries containing the data for each
                 checksum algorithm.
 
         Returns:
             iterable: missing signatures
 
         """
         for content in db.skipped_content_missing(contents, cur):
             yield dict(zip(db.content_hash_keys, content))
 
     @remote_api_endpoint('content/present')
     @timed
     @db_transaction()
     def content_find(self, content, db=None, cur=None):
         """Find a content hash in db.
 
         Args:
             content: a dictionary representing one content hash, mapping
                 checksum algorithm names (see swh.model.hashutil.ALGORITHMS) to
                 checksum values
 
         Returns:
             a triplet (sha1, sha1_git, sha256) if the content exist
             or None otherwise.
 
         Raises:
             ValueError: in case the key of the dictionary is not sha1, sha1_git
                 nor sha256.
 
         """
         if not set(content).intersection(ALGORITHMS):
             raise ValueError('content keys must contain at least one of: '
                              'sha1, sha1_git, sha256, blake2s256')
 
         contents = db.content_find(sha1=content.get('sha1'),
                                    sha1_git=content.get('sha1_git'),
                                    sha256=content.get('sha256'),
                                    blake2s256=content.get('blake2s256'),
                                    cur=cur)
         return [dict(zip(db.content_find_cols, content))
                 for content in contents]
 
     @remote_api_endpoint('directory/add')
     @timed
     @process_metrics
     @db_transaction()
     def directory_add(self, directories, db=None, cur=None):
         """Add directories to the storage
 
         Args:
             directories (iterable): iterable of dictionaries representing the
                 individual directories to add. Each dict has the following
                 keys:
 
                 - id (sha1_git): the id of the directory to add
                 - entries (list): list of dicts for each entry in the
                       directory.  Each dict has the following keys:
 
                       - name (bytes)
                       - type (one of 'file', 'dir', 'rev'): type of the
                         directory entry (file, directory, revision)
                       - target (sha1_git): id of the object pointed at by the
                         directory entry
                       - perms (int): entry permissions
 
         Returns:
             Summary dict of keys with associated count as values:
 
                 directory:add: Number of directories actually added
 
         """
         summary = {'directory:add': 0}
 
         dirs = set()
         dir_entries = {
             'file': defaultdict(list),
             'dir': defaultdict(list),
             'rev': defaultdict(list),
         }
 
         for cur_dir in directories:
             dir_id = cur_dir['id']
             dirs.add(dir_id)
             for src_entry in cur_dir['entries']:
                 entry = src_entry.copy()
                 entry['dir_id'] = dir_id
                 if entry['type'] not in ('file', 'dir', 'rev'):
                     raise ValueError(
                         'Entry type must be file, dir, or rev; not %s'
                         % entry['type'])
                 dir_entries[entry['type']][dir_id].append(entry)
 
         dirs_missing = set(self.directory_missing(dirs, db=db, cur=cur))
         if not dirs_missing:
             return summary
 
         if self.journal_writer:
             self.journal_writer.write_additions(
                 'directory',
                 (dir_ for dir_ in directories
                  if dir_['id'] in dirs_missing))
 
         # Copy directory ids
         dirs_missing_dict = ({'id': dir} for dir in dirs_missing)
         db.mktemp('directory', cur)
         db.copy_to(dirs_missing_dict, 'tmp_directory', ['id'], cur)
 
         # Copy entries
         for entry_type, entry_list in dir_entries.items():
             entries = itertools.chain.from_iterable(
                 entries_for_dir
                 for dir_id, entries_for_dir
                 in entry_list.items()
                 if dir_id in dirs_missing)
 
             db.mktemp_dir_entry(entry_type)
 
             db.copy_to(
                 entries,
                 'tmp_directory_entry_%s' % entry_type,
                 ['target', 'name', 'perms', 'dir_id'],
                 cur,
             )
 
         # Do the final copy
         db.directory_add_from_temp(cur)
         summary['directory:add'] = len(dirs_missing)
 
         return summary
 
     @remote_api_endpoint('directory/missing')
     @timed
     @db_transaction_generator()
     def directory_missing(self, directories, db=None, cur=None):
         """List directories missing from storage
 
         Args:
             directories (iterable): an iterable of directory ids
 
         Yields:
             missing directory ids
 
         """
         for obj in db.directory_missing_from_list(directories, cur):
             yield obj[0]
 
     @remote_api_endpoint('directory/ls')
     @timed
     @db_transaction_generator(statement_timeout=20000)
     def directory_ls(self, directory, recursive=False, db=None, cur=None):
         """Get entries for one directory.
 
         Args:
             - directory: the directory to list entries from.
             - recursive: if flag on, this list recursively from this directory.
 
         Returns:
             List of entries for such directory.
 
         If `recursive=True`, names in the path of a dir/file not at the
         root are concatenated with a slash (`/`).
 
         """
         if recursive:
             res_gen = db.directory_walk(directory, cur=cur)
         else:
             res_gen = db.directory_walk_one(directory, cur=cur)
 
         for line in res_gen:
             yield dict(zip(db.directory_ls_cols, line))
 
     @remote_api_endpoint('directory/path')
     @timed
     @db_transaction(statement_timeout=2000)
     def directory_entry_get_by_path(self, directory, paths, db=None, cur=None):
         """Get the directory entry (either file or dir) from directory with path.
 
         Args:
             - directory: sha1 of the top level directory
             - paths: path to lookup from the top level directory. From left
               (top) to right (bottom).
 
         Returns:
             The corresponding directory entry if found, None otherwise.
 
         """
         res = db.directory_entry_get_by_path(directory, paths, cur)
         if res:
             return dict(zip(db.directory_ls_cols, res))
 
     @remote_api_endpoint('revision/add')
     @timed
     @process_metrics
     @db_transaction()
     def revision_add(self, revisions, db=None, cur=None):
         """Add revisions to the storage
 
         Args:
             revisions (Iterable[dict]): iterable of dictionaries representing
                 the individual revisions to add. Each dict has the following
                 keys:
 
                 - **id** (:class:`sha1_git`): id of the revision to add
                 - **date** (:class:`dict`): date the revision was written
                 - **committer_date** (:class:`dict`): date the revision got
                   added to the origin
                 - **type** (one of 'git', 'tar'): type of the
                   revision added
                 - **directory** (:class:`sha1_git`): the directory the
                   revision points at
                 - **message** (:class:`bytes`): the message associated with
                   the revision
                 - **author** (:class:`Dict[str, bytes]`): dictionary with
                   keys: name, fullname, email
                 - **committer** (:class:`Dict[str, bytes]`): dictionary with
                   keys: name, fullname, email
                 - **metadata** (:class:`jsonb`): extra information as
                   dictionary
                 - **synthetic** (:class:`bool`): revision's nature (tarball,
                   directory creates synthetic revision`)
                 - **parents** (:class:`list[sha1_git]`): the parents of
                   this revision
 
         date dictionaries have the form defined in :mod:`swh.model`.
 
         Returns:
             Summary dict of keys with associated count as values
 
                 revision:add: New objects actually stored in db
 
         """
         summary = {'revision:add': 0}
 
         revisions_missing = set(self.revision_missing(
             set(revision['id'] for revision in revisions),
             db=db, cur=cur))
 
         if not revisions_missing:
             return summary
 
         db.mktemp_revision(cur)
 
         revisions_filtered = [
             revision for revision in revisions
             if revision['id'] in revisions_missing]
 
         if self.journal_writer:
             self.journal_writer.write_additions('revision', revisions_filtered)
 
         revisions_filtered = map(converters.revision_to_db, revisions_filtered)
 
         parents_filtered = []
 
         db.copy_to(
             revisions_filtered, 'tmp_revision', db.revision_add_cols,
             cur,
             lambda rev: parents_filtered.extend(rev['parents']))
 
         db.revision_add_from_temp(cur)
 
         db.copy_to(parents_filtered, 'revision_history',
                    ['id', 'parent_id', 'parent_rank'], cur)
 
         return {'revision:add': len(revisions_missing)}
 
     @remote_api_endpoint('revision/missing')
     @timed
     @db_transaction_generator()
     def revision_missing(self, revisions, db=None, cur=None):
         """List revisions missing from storage
 
         Args:
             revisions (iterable): revision ids
 
         Yields:
             missing revision ids
 
         """
         if not revisions:
             return
 
         for obj in db.revision_missing_from_list(revisions, cur):
             yield obj[0]
 
     @remote_api_endpoint('revision')
     @timed
     @db_transaction_generator(statement_timeout=1000)
     def revision_get(self, revisions, db=None, cur=None):
         """Get all revisions from storage
 
         Args:
             revisions: an iterable of revision ids
 
         Returns:
             iterable: an iterable of revisions as dictionaries (or None if the
                 revision doesn't exist)
 
         """
         for line in db.revision_get_from_list(revisions, cur):
             data = converters.db_to_revision(
                 dict(zip(db.revision_get_cols, line))
             )
             if not data['type']:
                 yield None
                 continue
             yield data
 
     @remote_api_endpoint('revision/log')
     @timed
     @db_transaction_generator(statement_timeout=2000)
     def revision_log(self, revisions, limit=None, db=None, cur=None):
         """Fetch revision entry from the given root revisions.
 
         Args:
             revisions: array of root revision to lookup
             limit: limitation on the output result. Default to None.
 
         Yields:
             List of revision log from such revisions root.
 
         """
         for line in db.revision_log(revisions, limit, cur):
             data = converters.db_to_revision(
                 dict(zip(db.revision_get_cols, line))
             )
             if not data['type']:
                 yield None
                 continue
             yield data
 
     @remote_api_endpoint('revision/shortlog')
     @timed
     @db_transaction_generator(statement_timeout=2000)
     def revision_shortlog(self, revisions, limit=None, db=None, cur=None):
         """Fetch the shortlog for the given revisions
 
         Args:
             revisions: list of root revisions to lookup
             limit: depth limitation for the output
 
         Yields:
             a list of (id, parents) tuples.
 
         """
 
         yield from db.revision_shortlog(revisions, limit, cur)
 
     @remote_api_endpoint('release/add')
     @timed
     @process_metrics
     @db_transaction()
     def release_add(self, releases, db=None, cur=None):
         """Add releases to the storage
 
         Args:
             releases (Iterable[dict]): iterable of dictionaries representing
                 the individual releases to add. Each dict has the following
                 keys:
 
                 - **id** (:class:`sha1_git`): id of the release to add
                 - **revision** (:class:`sha1_git`): id of the revision the
                   release points to
                 - **date** (:class:`dict`): the date the release was made
                 - **name** (:class:`bytes`): the name of the release
                 - **comment** (:class:`bytes`): the comment associated with
                   the release
                 - **author** (:class:`Dict[str, bytes]`): dictionary with
                   keys: name, fullname, email
 
         the date dictionary has the form defined in :mod:`swh.model`.
 
         Returns:
             Summary dict of keys with associated count as values
 
                 release:add: New objects contents actually stored in db
 
         """
         summary = {'release:add': 0}
 
         release_ids = set(release['id'] for release in releases)
         releases_missing = set(self.release_missing(release_ids,
                                                     db=db, cur=cur))
 
         if not releases_missing:
             return summary
 
         db.mktemp_release(cur)
 
         releases_missing = list(releases_missing)
 
         releases_filtered = [
             release for release in releases
             if release['id'] in releases_missing
         ]
 
         if self.journal_writer:
             self.journal_writer.write_additions('release', releases_filtered)
 
         releases_filtered = map(converters.release_to_db, releases_filtered)
 
         db.copy_to(releases_filtered, 'tmp_release', db.release_add_cols,
                    cur)
 
         db.release_add_from_temp(cur)
 
         return {'release:add': len(releases_missing)}
 
     @remote_api_endpoint('release/missing')
     @timed
     @db_transaction_generator()
     def release_missing(self, releases, db=None, cur=None):
         """List releases missing from storage
 
         Args:
             releases: an iterable of release ids
 
         Returns:
             a list of missing release ids
 
         """
         if not releases:
             return
 
         for obj in db.release_missing_from_list(releases, cur):
             yield obj[0]
 
     @remote_api_endpoint('release')
     @timed
     @db_transaction_generator(statement_timeout=500)
     def release_get(self, releases, db=None, cur=None):
         """Given a list of sha1, return the releases's information
 
         Args:
             releases: list of sha1s
 
         Yields:
             dicts with the same keys as those given to `release_add`
             (or ``None`` if a release does not exist)
 
         """
         for release in db.release_get_from_list(releases, cur):
             data = converters.db_to_release(
                 dict(zip(db.release_get_cols, release))
             )
             yield data if data['target_type'] else None
 
     @remote_api_endpoint('snapshot/add')
     @timed
     @process_metrics
     @db_transaction()
     def snapshot_add(self, snapshots, db=None, cur=None):
         """Add snapshots to the storage.
 
         Args:
             snapshot ([dict]): the snapshots to add, containing the
               following keys:
 
               - **id** (:class:`bytes`): id of the snapshot
               - **branches** (:class:`dict`): branches the snapshot contains,
                 mapping the branch name (:class:`bytes`) to the branch target,
                 itself a :class:`dict` (or ``None`` if the branch points to an
                 unknown object)
 
                 - **target_type** (:class:`str`): one of ``content``,
                   ``directory``, ``revision``, ``release``,
                   ``snapshot``, ``alias``
                 - **target** (:class:`bytes`): identifier of the target
                   (currently a ``sha1_git`` for all object kinds, or the name
                   of the target branch for aliases)
 
         Raises:
             ValueError: if the origin or visit id does not exist.
 
         Returns:
 
             Summary dict of keys with associated count as values
 
                 snapshot:add: Count of object actually stored in db
 
         """
         created_temp_table = False
 
         count = 0
         for snapshot in snapshots:
             if not db.snapshot_exists(snapshot['id'], cur):
                 if not created_temp_table:
                     db.mktemp_snapshot_branch(cur)
                     created_temp_table = True
 
                 db.copy_to(
                     (
                         {
                             'name': name,
                             'target': info['target'] if info else None,
                             'target_type': (info['target_type']
                                             if info else None),
                         }
                         for name, info in snapshot['branches'].items()
                     ),
                     'tmp_snapshot_branch',
                     ['name', 'target', 'target_type'],
                     cur,
                 )
 
                 if self.journal_writer:
                     self.journal_writer.write_addition('snapshot', snapshot)
 
                 db.snapshot_add(snapshot['id'], cur)
                 count += 1
 
         return {'snapshot:add': count}
 
     @remote_api_endpoint('snapshot')
     @timed
     @db_transaction(statement_timeout=2000)
     def snapshot_get(self, snapshot_id, db=None, cur=None):
         """Get the content, possibly partial, of a snapshot with the given id
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         .. warning:: At most 1000 branches contained in the snapshot will be
             returned for performance reasons. In order to browse the whole
             set of branches, the method :meth:`snapshot_get_branches`
             should be used instead.
 
         Args:
             snapshot_id (bytes): identifier of the snapshot
         Returns:
             dict: a dict with three keys:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than 1000
                   branches.
         """
 
         return self.snapshot_get_branches(snapshot_id, db=db, cur=cur)
 
     @remote_api_endpoint('snapshot/by_origin_visit')
     @timed
     @db_transaction(statement_timeout=2000)
     def snapshot_get_by_origin_visit(self, origin, visit, db=None, cur=None):
         """Get the content, possibly partial, of a snapshot for the given origin visit
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         .. warning:: At most 1000 branches contained in the snapshot will be
             returned for performance reasons. In order to browse the whole
             set of branches, the method :meth:`snapshot_get_branches`
             should be used instead.
 
         Args:
             origin (int): the origin identifier
             visit (int): the visit identifier
         Returns:
             dict: None if the snapshot does not exist;
               a dict with three keys otherwise:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than 1000
                   branches.
 
         """
         snapshot_id = db.snapshot_get_by_origin_visit(origin, visit, cur)
 
         if snapshot_id:
             return self.snapshot_get(snapshot_id, db=db, cur=cur)
 
         return None
 
     @remote_api_endpoint('snapshot/latest')
     @timed
     @db_transaction(statement_timeout=4000)
     def snapshot_get_latest(self, origin, allowed_statuses=None, db=None,
                             cur=None):
         """Get the content, possibly partial, of the latest snapshot for the
         given origin, optionally only from visits that have one of the given
         allowed_statuses
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         .. warning:: At most 1000 branches contained in the snapshot will be
             returned for performance reasons. In order to browse the whole
             set of branches, the method :meth:`snapshot_get_branches`
             should be used instead.
 
         Args:
             origin (str): the origin's URL
             allowed_statuses (list of str): list of visit statuses considered
                 to find the latest snapshot for the visit. For instance,
                 ``allowed_statuses=['full']`` will only consider visits that
                 have successfully run to completion.
         Returns:
             dict: a dict with three keys:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than 1000
                   branches.
         """
         if isinstance(origin, int):
             origin = self.origin_get({'id': origin}, db=db, cur=cur)
             if not origin:
                 return
             origin = origin['url']
 
         origin_visit = self.origin_visit_get_latest(
             origin, allowed_statuses=allowed_statuses, require_snapshot=True,
             db=db, cur=cur)
         if origin_visit and origin_visit['snapshot']:
             snapshot = self.snapshot_get(
                     origin_visit['snapshot'], db=db, cur=cur)
             if not snapshot:
                 raise ValueError(
                     'last origin visit references an unknown snapshot')
             return snapshot
 
     @remote_api_endpoint('snapshot/count_branches')
     @timed
     @db_transaction(statement_timeout=2000)
     def snapshot_count_branches(self, snapshot_id, db=None, cur=None):
         """Count the number of branches in the snapshot with the given id
 
         Args:
             snapshot_id (bytes): identifier of the snapshot
 
         Returns:
             dict: A dict whose keys are the target types of branches and
             values their corresponding amount
         """
         return dict([bc for bc in
                      db.snapshot_count_branches(snapshot_id, cur)])
 
     @remote_api_endpoint('snapshot/get_branches')
     @timed
     @db_transaction(statement_timeout=2000)
     def snapshot_get_branches(self, snapshot_id, branches_from=b'',
                               branches_count=1000, target_types=None,
                               db=None, cur=None):
         """Get the content, possibly partial, of a snapshot with the given id
 
         The branches of the snapshot are iterated in the lexicographical
         order of their names.
 
         Args:
             snapshot_id (bytes): identifier of the snapshot
             branches_from (bytes): optional parameter used to skip branches
                 whose name is lesser than it before returning them
             branches_count (int): optional parameter used to restrain
                 the amount of returned branches
             target_types (list): optional parameter used to filter the
                 target types of branch to return (possible values that can be
                 contained in that list are `'content', 'directory',
                 'revision', 'release', 'snapshot', 'alias'`)
         Returns:
             dict: None if the snapshot does not exist;
               a dict with three keys otherwise:
                 * **id**: identifier of the snapshot
                 * **branches**: a dict of branches contained in the snapshot
                   whose keys are the branches' names.
                 * **next_branch**: the name of the first branch not returned
                   or :const:`None` if the snapshot has less than
                   `branches_count` branches after `branches_from` included.
         """
         if snapshot_id == EMPTY_SNAPSHOT_ID:
             return {
                 'id': snapshot_id,
                 'branches': {},
                 'next_branch': None,
             }
 
         branches = {}
         next_branch = None
 
         fetched_branches = list(db.snapshot_get_by_id(
             snapshot_id, branches_from=branches_from,
             branches_count=branches_count+1, target_types=target_types,
             cur=cur,
         ))
         for branch in fetched_branches[:branches_count]:
             branch = dict(zip(db.snapshot_get_cols, branch))
             del branch['snapshot_id']
             name = branch.pop('name')
             if branch == {'target': None, 'target_type': None}:
                 branch = None
             branches[name] = branch
 
         if len(fetched_branches) > branches_count:
             branch = dict(zip(db.snapshot_get_cols, fetched_branches[-1]))
             next_branch = branch['name']
 
         if branches:
             return {
                 'id': snapshot_id,
                 'branches': branches,
                 'next_branch': next_branch,
             }
 
         return None
 
     @remote_api_endpoint('origin/visit/add')
     @timed
     @db_transaction()
     def origin_visit_add(self, origin, date, type,
                          db=None, cur=None):
         """Add an origin_visit for the origin at ts with status 'ongoing'.
 
         Args:
             origin (str): visited origin's identifier or URL
             date (Union[str,datetime]): timestamp of such visit
             type (str): the type of loader used for the visit (hg, git, ...)
 
         Returns:
             dict: dictionary with keys origin and visit where:
 
             - origin: origin identifier
             - visit: the visit identifier for the new visit occurrence
 
         """
         origin_url = origin
 
         if isinstance(date, str):
             # FIXME: Converge on iso8601 at some point
             date = dateutil.parser.parse(date)
 
         visit_id = db.origin_visit_add(origin_url, date, type, cur)
 
         if self.journal_writer:
             # We can write to the journal only after inserting to the
             # DB, because we want the id of the visit
             self.journal_writer.write_addition('origin_visit', {
                 'origin': origin_url, 'date': date, 'type': type,
                 'visit': visit_id,
                 'status': 'ongoing', 'metadata': None, 'snapshot': None})
 
         send_metric('origin_visit:add', count=1, method_name='origin_visit')
         return {
             'origin': origin_url,
             'visit': visit_id,
         }
 
     @remote_api_endpoint('origin/visit/update')
     @timed
     @db_transaction()
     def origin_visit_update(self, origin, visit_id, status=None,
                             metadata=None, snapshot=None,
                             db=None, cur=None):
         """Update an origin_visit's status.
 
         Args:
             origin (str): visited origin's URL
             visit_id: Visit's id
             status: Visit's new status
             metadata: Data associated to the visit
             snapshot (sha1_git): identifier of the snapshot to add to
                 the visit
 
         Returns:
             None
 
         """
         if not isinstance(origin, str):
             raise TypeError('origin must be a string, not %r' % (origin,))
         origin_url = origin
         visit = db.origin_visit_get(origin_url, visit_id, cur=cur)
 
         if not visit:
             raise ValueError('Invalid visit_id for this origin.')
 
         visit = dict(zip(db.origin_visit_get_cols, visit))
 
         updates = {}
         if status and status != visit['status']:
             updates['status'] = status
         if metadata and metadata != visit['metadata']:
             updates['metadata'] = metadata
         if snapshot and snapshot != visit['snapshot']:
             updates['snapshot'] = snapshot
 
         if updates:
             if self.journal_writer:
                 self.journal_writer.write_update('origin_visit', {
                     **visit, **updates})
 
             db.origin_visit_update(origin_url, visit_id, updates, cur)
 
     @remote_api_endpoint('origin/visit/upsert')
     @timed
     @db_transaction()
     def origin_visit_upsert(self, visits, db=None, cur=None):
         """Add a origin_visits with a specific id and with all its data.
         If there is already an origin_visit with the same
         `(origin_id, visit_id)`, overwrites it.
 
         Args:
             visits: iterable of dicts with keys:
 
                 origin: dict with keys either `id` or `url`
                 visit: origin visit id
                 date: timestamp of such visit
                 status: Visit's new status
                 metadata: Data associated to the visit
                 snapshot (sha1_git): identifier of the snapshot to add to
                     the visit
         """
         visits = copy.deepcopy(visits)
         for visit in visits:
             if isinstance(visit['date'], str):
                 visit['date'] = dateutil.parser.parse(visit['date'])
             if not isinstance(visit['origin'], str):
                 raise TypeError("visit['origin'] must be a string, not %r"
                                 % (visit['origin'],))
 
         if self.journal_writer:
             for visit in visits:
                 self.journal_writer.write_addition('origin_visit', visit)
 
         for visit in visits:
             # TODO: upsert them all in a single query
             db.origin_visit_upsert(**visit, cur=cur)
 
     @remote_api_endpoint('origin/visit/get')
     @timed
     @db_transaction_generator(statement_timeout=500)
     def origin_visit_get(self, origin, last_visit=None, limit=None, db=None,
                          cur=None):
         """Retrieve all the origin's visit's information.
 
         Args:
             origin (str): The visited origin
             last_visit: Starting point from which listing the next visits
                 Default to None
             limit (int): Number of results to return from the last visit.
                 Default to None
 
         Yields:
             List of visits.
 
         """
         for line in db.origin_visit_get_all(
                 origin, last_visit=last_visit, limit=limit, cur=cur):
             data = dict(zip(db.origin_visit_get_cols, line))
             yield data
 
     @remote_api_endpoint('origin/visit/find_by_date')
     @timed
     @db_transaction(statement_timeout=500)
     def origin_visit_find_by_date(self, origin, visit_date, db=None, cur=None):
         """Retrieves the origin visit whose date is closest to the provided
         timestamp.
         In case of a tie, the visit with largest id is selected.
 
         Args:
             origin (str): The occurrence's origin (URL).
             target (datetime): target timestamp
 
         Returns:
             A visit.
 
         """
         line = db.origin_visit_find_by_date(origin, visit_date, cur=cur)
         if line:
             return dict(zip(db.origin_visit_get_cols, line))
 
     @remote_api_endpoint('origin/visit/getby')
     @timed
     @db_transaction(statement_timeout=500)
     def origin_visit_get_by(self, origin, visit, db=None, cur=None):
         """Retrieve origin visit's information.
 
         Args:
             origin: The occurrence's origin (identifier).
 
         Returns:
             The information on that particular (origin, visit) or None if
             it does not exist
 
         """
         ori_visit = db.origin_visit_get(origin, visit, cur)
         if not ori_visit:
             return None
 
         return dict(zip(db.origin_visit_get_cols, ori_visit))
 
     @remote_api_endpoint('origin/visit/get_latest')
     @timed
     @db_transaction(statement_timeout=4000)
     def origin_visit_get_latest(
             self, origin, allowed_statuses=None, require_snapshot=False,
             db=None, cur=None):
         """Get the latest origin visit for the given origin, optionally
         looking only for those with one of the given allowed_statuses
         or for those with a known snapshot.
 
         Args:
             origin (str): the origin's URL
             allowed_statuses (list of str): list of visit statuses considered
                 to find the latest visit. For instance,
                 ``allowed_statuses=['full']`` will only consider visits that
                 have successfully run to completion.
             require_snapshot (bool): If True, only a visit with a snapshot
                 will be returned.
         Returns:
             dict: a dict with the following keys:
 
                 origin: the URL of the origin
                 visit: origin visit id
                 type: type of loader used for the visit
                 date: timestamp of such visit
                 status: Visit's new status
                 metadata: Data associated to the visit
                 snapshot (Optional[sha1_git]): identifier of the snapshot
                     associated to the visit
         """
         origin_visit = db.origin_visit_get_latest(
             origin, allowed_statuses=allowed_statuses,
             require_snapshot=require_snapshot, cur=cur)
         if origin_visit:
             return dict(zip(db.origin_visit_get_cols, origin_visit))
 
+    @remote_api_endpoint('origin/visit/get_random')
+    @timed
+    @db_transaction()
+    def origin_visit_get_random(
+            self, type: str, db=None, cur=None) -> Mapping[str, Any]:
+        """Randomly select one successful origin visit with <type>
+        made in the last 3 months.
+
+        Returns:
+            dict representing an origin visit, in the same format as
+            :py:meth:`origin_visit_get`.
+
+        """
+        data: Dict[str, Any] = {}
+        result = db.origin_visit_get_random(type, cur)
+        if result:
+            data = dict(zip(db.origin_visit_get_cols, result))
+        return data
+
     @remote_api_endpoint('object/find_by_sha1_git')
     @timed
     @db_transaction(statement_timeout=2000)
     def object_find_by_sha1_git(self, ids, db=None, cur=None):
         """Return the objects found with the given ids.
 
         Args:
             ids: a generator of sha1_gits
 
         Returns:
             dict: a mapping from id to the list of objects found. Each object
             found is itself a dict with keys:
 
             - sha1_git: the input id
             - type: the type of object found
             - id: the id of the object found
             - object_id: the numeric id of the object found.
 
         """
         ret = {id: [] for id in ids}
 
         for retval in db.object_find_by_sha1_git(ids, cur=cur):
             if retval[1]:
                 ret[retval[0]].append(dict(zip(db.object_find_by_sha1_git_cols,
                                                retval)))
 
         return ret
 
     @remote_api_endpoint('origin/get')
     @timed
     @db_transaction(statement_timeout=500)
     def origin_get(self, origins, db=None, cur=None):
         """Return origins, either all identified by their ids or all
         identified by tuples (type, url).
 
         If the url is given and the type is omitted, one of the origins with
         that url is returned.
 
         Args:
             origin: a list of dictionaries representing the individual
                 origins to find.
                 These dicts have the key url:
 
                 - url (bytes): the url the origin points to
 
         Returns:
             dict: the origin dictionary with the keys:
 
             - id: origin's id
             - url: origin's url
 
         Raises:
             ValueError: if the url or the id don't exist.
 
         """
         if isinstance(origins, dict):
             # Old API
             return_single = True
             origins = [origins]
         elif len(origins) == 0:
             return []
         else:
             return_single = False
 
         origin_urls = [origin['url'] for origin in origins]
         results = db.origin_get_by_url(origin_urls, cur)
 
         results = [dict(zip(db.origin_cols, result))
                    for result in results]
         if return_single:
             assert len(results) == 1
             if results[0]['url'] is not None:
                 return results[0]
             else:
                 return None
         else:
             return [None if res['url'] is None else res for res in results]
 
     @remote_api_endpoint('origin/get_sha1')
     @timed
     @db_transaction_generator(statement_timeout=500)
     def origin_get_by_sha1(self, sha1s, db=None, cur=None):
         """Return origins, identified by the sha1 of their URLs.
 
         Args:
             sha1s (list[bytes]): a list of sha1s
 
         Yields:
             dicts containing origin information as returned
             by :meth:`swh.storage.storage.Storage.origin_get`, or None if an
             origin matching the sha1 is not found.
 
         """
         for line in db.origin_get_by_sha1(sha1s, cur):
             if line[0] is not None:
                 yield dict(zip(db.origin_cols, line))
             else:
                 yield None
 
-    @remote_api_endpoint('origin/visit/get_random')
-    @timed
-    @db_transaction()
-    def origin_visit_get_random(
-            self, type, db=None, cur=None) -> Mapping[str, Any]:
-        """Randomly select one origin from the archive
-
-        Returns:
-            origin dict selected randomly on the dataset if found
-
-        """
-        data: Dict[str, Any] = {}
-        result = db.origin_visit_get_random(type, cur)
-        if result:
-            data = dict(zip(db.origin_visit_get_cols, result))
-        return data
-
     @remote_api_endpoint('origin/get_range')
     @timed
     @db_transaction_generator()
     def origin_get_range(self, origin_from=1, origin_count=100,
                          db=None, cur=None):
         """Retrieve ``origin_count`` origins whose ids are greater
         or equal than ``origin_from``.
 
         Origins are sorted by id before retrieving them.
 
         Args:
             origin_from (int): the minimum id of origins to retrieve
             origin_count (int): the maximum number of origins to retrieve
 
         Yields:
             dicts containing origin information as returned
             by :meth:`swh.storage.storage.Storage.origin_get`.
         """
         for origin in db.origin_get_range(origin_from, origin_count, cur):
             yield dict(zip(db.origin_get_range_cols, origin))
 
     @remote_api_endpoint('origin/search')
     @timed
     @db_transaction_generator()
     def origin_search(self, url_pattern, offset=0, limit=50,
                       regexp=False, with_visit=False, db=None, cur=None):
         """Search for origins whose urls contain a provided string pattern
         or match a provided regular expression.
         The search is performed in a case insensitive way.
 
         Args:
             url_pattern (str): the string pattern to search for in origin urls
             offset (int): number of found origins to skip before returning
                 results
             limit (int): the maximum number of found origins to return
             regexp (bool): if True, consider the provided pattern as a regular
                 expression and return origins whose urls match it
             with_visit (bool): if True, filter out origins with no visit
 
         Yields:
             dicts containing origin information as returned
             by :meth:`swh.storage.storage.Storage.origin_get`.
         """
         for origin in db.origin_search(url_pattern, offset, limit,
                                        regexp, with_visit, cur):
             yield dict(zip(db.origin_cols, origin))
 
     @remote_api_endpoint('origin/count')
     @timed
     @db_transaction()
     def origin_count(self, url_pattern, regexp=False,
                      with_visit=False, db=None, cur=None):
         """Count origins whose urls contain a provided string pattern
         or match a provided regular expression.
         The pattern search in origin urls is performed in a case insensitive
         way.
 
         Args:
             url_pattern (str): the string pattern to search for in origin urls
             regexp (bool): if True, consider the provided pattern as a regular
                 expression and return origins whose urls match it
             with_visit (bool): if True, filter out origins with no visit
 
         Returns:
             int: The number of origins matching the search criterion.
         """
         return db.origin_count(url_pattern, regexp, with_visit, cur)
 
     @remote_api_endpoint('origin/add_multi')
     @timed
     @db_transaction()
     def origin_add(self, origins, db=None, cur=None):
         """Add origins to the storage
 
         Args:
             origins: list of dictionaries representing the individual origins,
                 with the following keys:
 
                 - type: the origin type ('git', 'svn', 'deb', ...)
                 - url (bytes): the url the origin points to
 
         Returns:
             list: given origins as dict updated with their id
 
         """
         origins = copy.deepcopy(origins)
         for origin in origins:
             self.origin_add_one(origin, db=db, cur=cur)
 
         send_metric('origin:add', count=len(origins), method_name='origin_add')
         return origins
 
     @remote_api_endpoint('origin/add')
     @timed
     @db_transaction()
     def origin_add_one(self, origin, db=None, cur=None):
         """Add origin to the storage
 
         Args:
             origin: dictionary representing the individual origin to add. This
                 dict has the following keys:
 
                 - type (FIXME: enum TBD): the origin type ('git', 'wget', ...)
                 - url (bytes): the url the origin points to
 
         Returns:
             the id of the added origin, or of the identical one that already
             exists.
 
         """
         origin_row = list(db.origin_get_by_url([origin['url']], cur))[0]
         origin_url = dict(zip(db.origin_cols, origin_row))['url']
         if origin_url:
             return origin_url
 
         if self.journal_writer:
             self.journal_writer.write_addition('origin', origin)
 
         origins = db.origin_add(origin['url'], cur)
         send_metric('origin:add', count=len(origins), method_name='origin_add')
         return origins
 
     @db_transaction(statement_timeout=500)
     def stat_counters(self, db=None, cur=None):
         """compute statistics about the number of tuples in various tables
 
         Returns:
             dict: a dictionary mapping textual labels (e.g., content) to
             integer values (e.g., the number of tuples in table content)
 
         """
         return {k: v for (k, v) in db.stat_counters()}
 
     @db_transaction()
     def refresh_stat_counters(self, db=None, cur=None):
         """Recomputes the statistics for `stat_counters`."""
         keys = [
             'content',
             'directory',
             'directory_entry_dir',
             'directory_entry_file',
             'directory_entry_rev',
             'origin',
             'origin_visit',
             'person',
             'release',
             'revision',
             'revision_history',
             'skipped_content',
             'snapshot']
 
         for key in keys:
             cur.execute('select * from swh_update_counter(%s)', (key,))
 
     @remote_api_endpoint('origin/metadata/add')
     @timed
     @db_transaction()
     def origin_metadata_add(self, origin_url, ts, provider, tool, metadata,
                             db=None, cur=None):
         """ Add an origin_metadata for the origin at ts with provenance and
         metadata.
 
         Args:
             origin_url (str): the origin url for which the metadata is added
             ts (datetime): timestamp of the found metadata
             provider (int): the provider of metadata (ex:'hal')
             tool (int): tool used to extract metadata
             metadata (jsonb): the metadata retrieved at the time and location
         """
         if isinstance(ts, str):
             ts = dateutil.parser.parse(ts)
 
         db.origin_metadata_add(origin_url, ts, provider, tool,
                                metadata, cur)
         send_metric(
             'origin_metadata:add', count=1, method_name='origin_metadata_add')
 
     @remote_api_endpoint('origin/metadata/get')
     @timed
     @db_transaction_generator(statement_timeout=500)
     def origin_metadata_get_by(self, origin_url, provider_type=None, db=None,
                                cur=None):
         """Retrieve list of all origin_metadata entries for the origin_id
 
         Args:
             origin_url (str): the origin's URL
             provider_type (str): (optional) type of provider
 
         Returns:
             list of dicts: the origin_metadata dictionary with the keys:
 
             - origin_id (int): origin's id
             - discovery_date (datetime): timestamp of discovery
             - tool_id (int): metadata's extracting tool
             - metadata (jsonb)
             - provider_id (int): metadata's provider
             - provider_name (str)
             - provider_type (str)
             - provider_url (str)
 
         """
         for line in db.origin_metadata_get_by(origin_url, provider_type, cur):
             yield dict(zip(db.origin_metadata_get_cols, line))
 
     @remote_api_endpoint('tool/add')
     @timed
     @db_transaction()
     def tool_add(self, tools, db=None, cur=None):
         """Add new tools to the storage.
 
         Args:
             tools (iterable of :class:`dict`): Tool information to add to
               storage. Each tool is a :class:`dict` with the following keys:
 
               - name (:class:`str`): name of the tool
               - version (:class:`str`): version of the tool
               - configuration (:class:`dict`): configuration of the tool,
                 must be json-encodable
 
         Returns:
             :class:`dict`: All the tools inserted in storage
             (including the internal ``id``). The order of the list is not
             guaranteed to match the order of the initial list.
 
         """
         db.mktemp_tool(cur)
         db.copy_to(tools, 'tmp_tool',
                    ['name', 'version', 'configuration'],
                    cur)
 
         tools = db.tool_add_from_temp(cur)
         results = [dict(zip(db.tool_cols, line)) for line in tools]
         send_metric('tool:add', count=len(results), method_name='tool_add')
         return results
 
     @remote_api_endpoint('tool/data')
     @timed
     @db_transaction(statement_timeout=500)
     def tool_get(self, tool, db=None, cur=None):
         """Retrieve tool information.
 
         Args:
             tool (dict): Tool information we want to retrieve from storage.
               The dicts have the same keys as those used in :func:`tool_add`.
 
         Returns:
             dict: The full tool information if it exists (``id`` included),
             None otherwise.
 
         """
         tool_conf = tool['configuration']
         if isinstance(tool_conf, dict):
             tool_conf = json.dumps(tool_conf)
 
         idx = db.tool_get(tool['name'],
                           tool['version'],
                           tool_conf)
         if not idx:
             return None
         return dict(zip(db.tool_cols, idx))
 
     @remote_api_endpoint('provider/add')
     @timed
     @db_transaction()
     def metadata_provider_add(self, provider_name, provider_type, provider_url,
                               metadata, db=None, cur=None):
         """Add a metadata provider.
 
         Args:
             provider_name (str): Its name
             provider_type (str): Its type (eg. `'deposit-client'`)
             provider_url (str): Its URL
             metadata: JSON-encodable object
 
         Returns:
             int: an identifier of the provider
         """
         result = db.metadata_provider_add(provider_name, provider_type,
                                           provider_url, metadata, cur)
         send_metric(
             'metadata_provider:add', count=1, method_name='metadata_provider')
         return result
 
     @remote_api_endpoint('provider/get')
     @timed
     @db_transaction()
     def metadata_provider_get(self, provider_id, db=None, cur=None):
         """Get a metadata provider
 
         Args:
             provider_id: Its identifier, as given by `metadata_provider_add`.
 
         Returns:
             dict: same as `metadata_provider_add`;
                   or None if it does not exist.
         """
         result = db.metadata_provider_get(provider_id)
         if not result:
             return None
         return dict(zip(db.metadata_provider_cols, result))
 
     @remote_api_endpoint('provider/getby')
     @timed
     @db_transaction()
     def metadata_provider_get_by(self, provider, db=None, cur=None):
         """Get a metadata provider
 
         Args:
             provider (dict): A dictionary with keys:
                 * provider_name: Its name
                 * provider_url: Its URL
 
         Returns:
             dict: same as `metadata_provider_add`;
                   or None if it does not exist.
         """
         result = db.metadata_provider_get_by(provider['provider_name'],
                                              provider['provider_url'])
         if not result:
             return None
         return dict(zip(db.metadata_provider_cols, result))
 
     @remote_api_endpoint('algos/diff_directories')
     @timed
     def diff_directories(self, from_dir, to_dir, track_renaming=False):
         """Compute the list of file changes introduced between two arbitrary
         directories (insertion / deletion / modification / renaming of files).
 
         Args:
             from_dir (bytes): identifier of the directory to compare from
             to_dir (bytes): identifier of the directory to compare to
             track_renaming (bool): whether or not to track files renaming
 
         Returns:
             A list of dict describing the introduced file changes
             (see :func:`swh.storage.algos.diff.diff_directories`
             for more details).
         """
         return diff.diff_directories(self, from_dir, to_dir, track_renaming)
 
     @remote_api_endpoint('algos/diff_revisions')
     @timed
     def diff_revisions(self, from_rev, to_rev, track_renaming=False):
         """Compute the list of file changes introduced between two arbitrary
         revisions (insertion / deletion / modification / renaming of files).
 
         Args:
             from_rev (bytes): identifier of the revision to compare from
             to_rev (bytes): identifier of the revision to compare to
             track_renaming (bool): whether or not to track files renaming
 
         Returns:
             A list of dict describing the introduced file changes
             (see :func:`swh.storage.algos.diff.diff_directories`
             for more details).
         """
         return diff.diff_revisions(self, from_rev, to_rev, track_renaming)
 
     @remote_api_endpoint('algos/diff_revision')
     @timed
     def diff_revision(self, revision, track_renaming=False):
         """Compute the list of file changes introduced by a specific revision
         (insertion / deletion / modification / renaming of files) by comparing
         it against its first parent.
 
         Args:
             revision (bytes): identifier of the revision from which to
                 compute the list of files changes
             track_renaming (bool): whether or not to track files renaming
 
         Returns:
             A list of dict describing the introduced file changes
             (see :func:`swh.storage.algos.diff.diff_directories`
             for more details).
         """
         return diff.diff_revision(self, revision, track_renaming)