diff --git a/swh/indexer/indexer.py b/swh/indexer/indexer.py index 1bc69fc..bbd7b20 100644 --- a/swh/indexer/indexer.py +++ b/swh/indexer/indexer.py @@ -1,621 +1,644 @@ -# Copyright (C) 2016-2018 The Software Heritage developers +# Copyright (C) 2016-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import os import logging import shutil import tempfile import datetime from copy import deepcopy from contextlib import contextmanager -from typing import Any, Dict, Tuple +from typing import Any, Dict, Tuple, Generator, Union, List, Optional +from typing import Set from swh.scheduler import get_scheduler from swh.scheduler import CONFIG as SWH_CONFIG from swh.storage import get_storage from swh.core.config import SWHConfig from swh.objstorage import get_objstorage from swh.objstorage.exc import ObjNotFoundError from swh.indexer.storage import get_indexer_storage, INDEXER_CFG_KEY from swh.model import hashutil from swh.core import utils @contextmanager -def write_to_temp(filename, data, working_directory): +def write_to_temp( + filename: str, data: bytes, working_directory: str +) -> Generator[str, None, None]: """Write the sha1's content in a temporary file. Args: - filename (str): one of sha1's many filenames - data (bytes): the sha1's content to write in temporary + filename: one of sha1's many filenames + data: the sha1's content to write in temporary file + working_directory: the directory into which the + file is written Returns: The path to the temporary file created. That file is filled in with the raw content's data. """ os.makedirs(working_directory, exist_ok=True) temp_dir = tempfile.mkdtemp(dir=working_directory) content_path = os.path.join(temp_dir, filename) with open(content_path, 'wb') as f: f.write(data) yield content_path shutil.rmtree(temp_dir) class BaseIndexer(SWHConfig, metaclass=abc.ABCMeta): """Base class for indexers to inherit from. The main entry point is the :func:`run` function which is in charge of triggering the computations on the batch dict/ids received. Indexers can: - filter out ids whose data has already been indexed. - retrieve ids data from storage or objstorage - index this data depending on the object and store the result in storage. To implement a new object type indexer, inherit from the BaseIndexer and implement indexing: :meth:`~BaseIndexer.run`: object_ids are different depending on object. For example: sha1 for content, sha1_git for revision, directory, release, and id for origin To implement a new concrete indexer, inherit from the object level classes: :class:`ContentIndexer`, :class:`RevisionIndexer`, :class:`OriginIndexer`. Then you need to implement the following functions: :meth:`~BaseIndexer.filter`: filter out data already indexed (in storage). :meth:`~BaseIndexer.index_object`: compute index on id with data (retrieved from the storage or the objstorage by the id key) and return the resulting index computation. :meth:`~BaseIndexer.persist_index_computations`: persist the results of multiple index computations in the storage. The new indexer implementation can also override the following functions: :meth:`~BaseIndexer.prepare`: Configuration preparation for the indexer. When overriding, this must call the `super().prepare()` instruction. :meth:`~BaseIndexer.check`: Configuration check for the indexer. When overriding, this must call the `super().check()` instruction. :meth:`~BaseIndexer.register_tools`: This should return a dict of the tool(s) to use when indexing or filtering. """ + results: List[Dict] + CONFIG = 'indexer/base' DEFAULT_CONFIG = { INDEXER_CFG_KEY: ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5007/' } }), 'storage': ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5002/', } }), 'objstorage': ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5003/', } }) } ADDITIONAL_CONFIG = {} # type: Dict[str, Tuple[str, Any]] USE_TOOLS = True catch_exceptions = True """Prevents exceptions in `index()` from raising too high. Set to False in tests to properly catch all exceptions.""" - def __init__(self, config=None, **kw): + scheduler: Any + + def __init__(self, config=None, **kw) -> None: """Prepare and check that the indexer is ready to run. """ super().__init__() if config is not None: self.config = config elif SWH_CONFIG: self.config = SWH_CONFIG.copy() else: config_keys = ('base_filename', 'config_filename', 'additional_configs', 'global_config') config_args = {k: v for k, v in kw.items() if k in config_keys} if self.ADDITIONAL_CONFIG: config_args.setdefault('additional_configs', []).append( self.ADDITIONAL_CONFIG) self.config = self.parse_config_file(**config_args) self.prepare() self.check() self.log.debug('%s: config=%s', self, self.config) - def prepare(self): + def prepare(self) -> None: """Prepare the indexer's needed runtime configuration. Without this step, the indexer cannot possibly run. """ config_storage = self.config.get('storage') if config_storage: self.storage = get_storage(**config_storage) objstorage = self.config['objstorage'] self.objstorage = get_objstorage(objstorage['cls'], objstorage['args']) idx_storage = self.config[INDEXER_CFG_KEY] self.idx_storage = get_indexer_storage(**idx_storage) _log = logging.getLogger('requests.packages.urllib3.connectionpool') _log.setLevel(logging.WARN) self.log = logging.getLogger('swh.indexer') if self.USE_TOOLS: self.tools = list(self.register_tools( self.config.get('tools', []))) self.results = [] @property - def tool(self): + def tool(self) -> Dict: return self.tools[0] - def check(self): + def check(self) -> None: """Check the indexer's configuration is ok before proceeding. If ok, does nothing. If not raise error. """ if self.USE_TOOLS and not self.tools: raise ValueError('Tools %s is unknown, cannot continue' % self.tools) - def _prepare_tool(self, tool): + def _prepare_tool(self, tool: Dict[str, Any]) -> Dict[str, Any]: """Prepare the tool dict to be compliant with the storage api. """ return {'tool_%s' % key: value for key, value in tool.items()} - def register_tools(self, tools): + def register_tools( + self, tools: Union[Dict[str, Any], List[Dict[str, Any]]] + ) -> List[Dict[str, Any]]: """Permit to register tools to the storage. Add a sensible default which can be overridden if not sufficient. (For now, all indexers use only one tool) Expects the self.config['tools'] property to be set with one or more tools. Args: - tools (dict/[dict]): Either a dict or a list of dict. + tools: Either a dict or a list of dict. Returns: list: List of dicts with additional id key. Raises: ValueError: if not a list nor a dict. """ if isinstance(tools, list): tools = list(map(self._prepare_tool, tools)) elif isinstance(tools, dict): tools = [self._prepare_tool(tools)] else: raise ValueError('Configuration tool(s) must be a dict or list!') if tools: return self.idx_storage.indexer_configuration_add(tools) else: return [] - def index(self, id, data): + def index( + self, id: bytes, data: bytes + ) -> Dict[str, Any]: """Index computation for the id and associated raw data. Args: - id (bytes): identifier - data (bytes): id's data from storage or objstorage depending on + id: identifier + data: id's data from storage or objstorage depending on object type Returns: dict: a dict that makes sense for the :meth:`.persist_index_computations` method. """ raise NotImplementedError() - def filter(self, ids): + def filter(self, ids: List[bytes]) -> Generator[bytes, None, None]: """Filter missing ids for that particular indexer. Args: - ids ([bytes]): list of ids + ids: list of ids Yields: iterator of missing ids """ yield from ids @abc.abstractmethod def persist_index_computations(self, results, policy_update): """Persist the computation resulting from the index. Args: results ([result]): List of results. One result is the result of the index function. policy_update ([str]): either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them Returns: None """ pass - def next_step(self, results, task): + def next_step( + self, results: List[Dict], task: Optional[Dict[str, Any]] + ) -> None: """Do something else with computations results (e.g. send to another queue, ...). (This is not an abstractmethod since it is optional). Args: - results ([result]): List of results (dict) as returned + results: List of results (dict) as returned by index function. - task (dict): a dict in the form expected by + task: a dict in the form expected by `scheduler.backend.SchedulerBackend.create_tasks` without `next_run`, plus an optional `result_name` key. Returns: None """ if task: if getattr(self, 'scheduler', None): scheduler = self.scheduler else: scheduler = get_scheduler(**self.config['scheduler']) task = deepcopy(task) result_name = task.pop('result_name', None) task['next_run'] = datetime.datetime.now() if result_name: task['arguments']['kwargs'][result_name] = self.results scheduler.create_tasks([task]) @abc.abstractmethod def run(self, ids, policy_update, next_step=None, **kwargs): """Given a list of ids: - retrieves the data from the storage - executes the indexing computations - stores the results (according to policy_update) Args: ids ([bytes]): id's identifier list policy_update (str): either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them next_step (dict): a dict in the form expected by `scheduler.backend.SchedulerBackend.create_tasks` without `next_run`, plus a `result_name` key. **kwargs: passed to the `index` method """ pass class ContentIndexer(BaseIndexer): """A content indexer working on a list of ids directly. To work on indexer range, use the :class:`ContentRangeIndexer` instead. Note: :class:`ContentIndexer` is not an instantiable object. To use it, one should inherit from this class and override the methods mentioned in the :class:`BaseIndexer` class. """ def run(self, ids, policy_update, next_step=None, **kwargs): """Given a list of ids: - retrieve the content from the storage - execute the indexing computations - store the results (according to policy_update) Args: ids (Iterable[Union[bytes, str]]): sha1's identifier list policy_update (str): either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them next_step (dict): a dict in the form expected by `scheduler.backend.SchedulerBackend.create_tasks` without `next_run`, plus an optional `result_name` key. **kwargs: passed to the `index` method """ ids = [hashutil.hash_to_bytes(id_) if isinstance(id_, str) else id_ for id_ in ids] results = [] try: for sha1 in ids: try: raw_content = self.objstorage.get(sha1) except ObjNotFoundError: self.log.warning('Content %s not found in objstorage' % hashutil.hash_to_hex(sha1)) continue res = self.index(sha1, raw_content, **kwargs) if res: # If no results, skip it results.append(res) self.persist_index_computations(results, policy_update) self.results = results return self.next_step(results, task=next_step) except Exception: if not self.catch_exceptions: raise self.log.exception( 'Problem when reading contents metadata.') class ContentRangeIndexer(BaseIndexer): """A content range indexer. This expects as input a range of ids to index. To work on a list of ids, use the :class:`ContentIndexer` instead. Note: :class:`ContentRangeIndexer` is not an instantiable object. To use it, one should inherit from this class and override the methods mentioned in the :class:`BaseIndexer` class. """ @abc.abstractmethod - def indexed_contents_in_range(self, start, end): + def indexed_contents_in_range( + self, start: bytes, end: bytes + ) -> Any: """Retrieve indexed contents within range [start, end]. Args: - start (bytes): Starting bound from range identifier - end (bytes): End range identifier + start: Starting bound from range identifier + end: End range identifier Yields: bytes: Content identifier present in the range ``[start, end]`` """ pass - def _list_contents_to_index(self, start, end, indexed): + def _list_contents_to_index( + self, start: bytes, end: bytes, indexed: Set[bytes] + ) -> Generator[bytes, None, None]: """Compute from storage the new contents to index in the range [start, end]. The already indexed contents are skipped. Args: - start (bytes): Starting bound from range identifier - end (bytes): End range identifier - indexed (Set[bytes]): Set of content already indexed. + start: Starting bound from range identifier + end: End range identifier + indexed: Set of content already indexed. Yields: bytes: Identifier of contents to index. """ if not isinstance(start, bytes) or not isinstance(end, bytes): raise TypeError('identifiers must be bytes, not %r and %r.' % (start, end)) while start: result = self.storage.content_get_range(start, end) contents = result['contents'] for c in contents: _id = hashutil.hash_to_bytes(c['sha1']) if _id in indexed: continue yield _id start = result['next'] - def _index_contents(self, start, end, indexed, **kwargs): + def _index_contents( + self, start: bytes, end: bytes, indexed: Set[bytes], **kwargs: Any + ) -> Generator[Dict, None, None]: """Index the contents from within range [start, end] Args: - start (bytes): Starting bound from range identifier - end (bytes): End range identifier - indexed (Set[bytes]): Set of content already indexed. + start: Starting bound from range identifier + end: End range identifier + indexed: Set of content already indexed. Yields: dict: Data indexed to persist using the indexer storage """ for sha1 in self._list_contents_to_index(start, end, indexed): try: raw_content = self.objstorage.get(sha1) except ObjNotFoundError: self.log.warning('Content %s not found in objstorage' % hashutil.hash_to_hex(sha1)) continue - res = self.index(sha1, raw_content, **kwargs) + res = self.index(sha1, raw_content, **kwargs) # type: ignore if res: if not isinstance(res['id'], bytes): raise TypeError( '%r.index should return ids as bytes, not %r' % (self.__class__.__name__, res['id'])) yield res - def _index_with_skipping_already_done(self, start, end): + def _index_with_skipping_already_done( + self, start: bytes, end: bytes + ) -> Generator[Dict, None, None]: """Index not already indexed contents in range [start, end]. Args: - start** (Union[bytes, str]): Starting range identifier - end (Union[bytes, str]): Ending range identifier + start: Starting range identifier + end: Ending range identifier Yields: - bytes: Content identifier present in the range + dict: Content identifier present in the range ``[start, end]`` which are not already indexed. """ while start: indexed_page = self.indexed_contents_in_range(start, end) contents = indexed_page['ids'] _end = contents[-1] if contents else end yield from self._index_contents( start, _end, contents) start = indexed_page['next'] def run(self, start, end, skip_existing=True, **kwargs): """Given a range of content ids, compute the indexing computations on the contents within. Either the indexer is incremental (filter out existing computed data) or not (compute everything from scratch). Args: start (Union[bytes, str]): Starting range identifier end (Union[bytes, str]): Ending range identifier skip_existing (bool): Skip existing indexed data (default) or not **kwargs: passed to the `index` method Returns: bool: True if data was indexed, False otherwise. """ with_indexed_data = False try: if isinstance(start, str): start = hashutil.hash_to_bytes(start) if isinstance(end, str): end = hashutil.hash_to_bytes(end) if skip_existing: gen = self._index_with_skipping_already_done(start, end) else: gen = self._index_contents(start, end, indexed=[]) for results in utils.grouper(gen, n=self.config['write_batch_size']): self.persist_index_computations( results, policy_update='update-dups') with_indexed_data = True except Exception: if not self.catch_exceptions: raise self.log.exception( 'Problem when computing metadata.') finally: return with_indexed_data class OriginIndexer(BaseIndexer): """An object type indexer, inherits from the :class:`BaseIndexer` and implements Origin indexing using the run method Note: the :class:`OriginIndexer` is not an instantiable object. To use it in another context one should inherit from this class and override the methods mentioned in the :class:`BaseIndexer` class. """ def run(self, origin_urls, policy_update='update-dups', next_step=None, **kwargs): """Given a list of origin urls: - retrieve origins from storage - execute the indexing computations - store the results (according to policy_update) Args: origin_urls ([str]): list of origin urls. policy_update (str): either 'update-dups' or 'ignore-dups' to respectively update duplicates (default) or ignore them next_step (dict): a dict in the form expected by `scheduler.backend.SchedulerBackend.create_tasks` without `next_run`, plus an optional `result_name` key. parse_ids (bool): Do we need to parse id or not (default) **kwargs: passed to the `index` method """ results = self.index_list(origin_urls, **kwargs) self.persist_index_computations(results, policy_update) self.results = results return self.next_step(results, task=next_step) - def index_list(self, origins, **kwargs): + def index_list(self, origins: List[Any], **kwargs: Any) -> List[Dict]: results = [] for origin in origins: try: res = self.index(origin, **kwargs) if res: # If no results, skip it results.append(res) except Exception: if not self.catch_exceptions: raise self.log.exception( 'Problem when processing origin %s', origin['id']) return results class RevisionIndexer(BaseIndexer): """An object type indexer, inherits from the :class:`BaseIndexer` and implements Revision indexing using the run method Note: the :class:`RevisionIndexer` is not an instantiable object. To use it in another context one should inherit from this class and override the methods mentioned in the :class:`BaseIndexer` class. """ def run(self, ids, policy_update, next_step=None): """Given a list of sha1_gits: - retrieve revisions from storage - execute the indexing computations - store the results (according to policy_update) Args: ids ([bytes or str]): sha1_git's identifier list policy_update (str): either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them """ results = [] ids = [hashutil.hash_to_bytes(id_) if isinstance(id_, str) else id_ for id_ in ids] revs = self.storage.revision_get(ids) for rev in revs: if not rev: self.log.warning('Revisions %s not found in storage' % list(map(hashutil.hash_to_hex, ids))) continue try: res = self.index(rev) if res: # If no results, skip it results.append(res) except Exception: if not self.catch_exceptions: raise self.log.exception( 'Problem when processing revision') self.persist_index_computations(results, policy_update) self.results = results return self.next_step(results, task=next_step) diff --git a/swh/indexer/metadata.py b/swh/indexer/metadata.py index 2900221..9b3781e 100644 --- a/swh/indexer/metadata.py +++ b/swh/indexer/metadata.py @@ -1,354 +1,366 @@ -# Copyright (C) 2017-2018 The Software Heritage developers +# Copyright (C) 2017-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from copy import deepcopy +from typing import Any, List, Dict, Tuple, Callable, Generator + from swh.core.utils import grouper from swh.indexer.codemeta import merge_documents from swh.indexer.indexer import ContentIndexer, RevisionIndexer, OriginIndexer from swh.indexer.origin_head import OriginHeadIndexer from swh.indexer.metadata_dictionary import MAPPINGS from swh.indexer.metadata_detector import detect_metadata from swh.indexer.storage import INDEXER_CFG_KEY from swh.model import hashutil REVISION_GET_BATCH_SIZE = 10 ORIGIN_GET_BATCH_SIZE = 10 -def call_with_batches(f, args, batch_size): +def call_with_batches( + f: Callable[[List[Dict[str, Any]]], Dict['str', Any]], + args: List[Dict[str, str]], batch_size: int +) -> Generator[str, None, None]: """Calls a function with batches of args, and concatenates the results. """ groups = grouper(args, batch_size) for group in groups: yield from f(list(group)) class ContentMetadataIndexer(ContentIndexer): """Content-level indexer This indexer is in charge of: - filtering out content already indexed in content_metadata - reading content from objstorage with the content's id sha1 - computing metadata by given context - using the metadata_dictionary as the 'swh-metadata-translator' tool - store result in content_metadata table """ def filter(self, ids): """Filter out known sha1s and return only missing ones. """ yield from self.idx_storage.content_metadata_missing(( { 'id': sha1, 'indexer_configuration_id': self.tool['id'], } for sha1 in ids )) def index(self, id, data, log_suffix='unknown revision'): """Index sha1s' content and store result. Args: id (bytes): content's identifier data (bytes): raw content in bytes Returns: dict: dictionary representing a content_metadata. If the translation wasn't successful the metadata keys will be returned as None """ result = { 'id': id, 'indexer_configuration_id': self.tool['id'], 'metadata': None } try: mapping_name = self.tool['tool_configuration']['context'] log_suffix += ', content_id=%s' % hashutil.hash_to_hex(id) result['metadata'] = \ MAPPINGS[mapping_name](log_suffix).translate(data) except Exception: self.log.exception( "Problem during metadata translation " "for content %s" % hashutil.hash_to_hex(id)) if result['metadata'] is None: return None return result - def persist_index_computations(self, results, policy_update): + def persist_index_computations( + self, results: List[Dict], policy_update: str + ) -> None: """Persist the results in storage. Args: - results ([dict]): list of content_metadata, dict with the + results: list of content_metadata, dict with the following keys: - id (bytes): content's identifier (sha1) - metadata (jsonb): detected metadata - policy_update ([str]): either 'update-dups' or 'ignore-dups' to + policy_update: either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them """ self.idx_storage.content_metadata_add( results, conflict_update=(policy_update == 'update-dups')) class RevisionMetadataIndexer(RevisionIndexer): """Revision-level indexer This indexer is in charge of: - filtering revisions already indexed in revision_intrinsic_metadata table with defined computation tool - retrieve all entry_files in root directory - use metadata_detector for file_names containing metadata - compute metadata translation if necessary and possible (depends on tool) - send sha1s to content indexing if possible - store the results for revision """ ADDITIONAL_CONFIG = { 'tools': ('dict', { 'name': 'swh-metadata-detector', 'version': '0.0.2', 'configuration': { }, }), } def filter(self, sha1_gits): """Filter out known sha1s and return only missing ones. """ yield from self.idx_storage.revision_intrinsic_metadata_missing(( { 'id': sha1_git, 'indexer_configuration_id': self.tool['id'], } for sha1_git in sha1_gits )) def index(self, rev): """Index rev by processing it and organizing result. use metadata_detector to iterate on filenames - if one filename detected -> sends file to content indexer - if multiple file detected -> translation needed at revision level Args: rev (dict): revision artifact from storage Returns: dict: dictionary representing a revision_intrinsic_metadata, with keys: - id (str): rev's identifier (sha1_git) - indexer_configuration_id (bytes): tool used - metadata: dict of retrieved metadata """ result = { 'id': rev['id'], 'indexer_configuration_id': self.tool['id'], 'mappings': None, 'metadata': None } try: root_dir = rev['directory'] dir_ls = list(self.storage.directory_ls(root_dir, recursive=False)) if [entry['type'] for entry in dir_ls] == ['dir']: # If the root is just a single directory, recurse into it # eg. PyPI packages, GNU tarballs subdir = dir_ls[0]['target'] dir_ls = self.storage.directory_ls(subdir, recursive=False) files = [entry for entry in dir_ls if entry['type'] == 'file'] detected_files = detect_metadata(files) (mappings, metadata) = self.translate_revision_intrinsic_metadata( detected_files, log_suffix='revision=%s' % hashutil.hash_to_hex(rev['id'])) result['mappings'] = mappings result['metadata'] = metadata except Exception as e: self.log.exception( 'Problem when indexing rev: %r', e) return result - def persist_index_computations(self, results, policy_update): + def persist_index_computations( + self, results: List[Dict], policy_update: str + ) -> None: """Persist the results in storage. Args: - results ([dict]): list of content_mimetype, dict with the + results: list of content_mimetype, dict with the following keys: - id (bytes): content's identifier (sha1) - mimetype (bytes): mimetype in bytes - encoding (bytes): encoding in bytes - policy_update ([str]): either 'update-dups' or 'ignore-dups' to + policy_update: either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them """ # TODO: add functions in storage to keep data in # revision_intrinsic_metadata self.idx_storage.revision_intrinsic_metadata_add( results, conflict_update=(policy_update == 'update-dups')) def translate_revision_intrinsic_metadata( - self, detected_files, log_suffix): + self, detected_files: Dict[str, List[Any]], log_suffix: str + ) -> Tuple[List[Any], List[Any]]: """ Determine plan of action to translate metadata when containing one or multiple detected files: Args: - detected_files (dict): dictionary mapping context names (e.g., + detected_files: dictionary mapping context names (e.g., "npm", "authors") to list of sha1 Returns: (List[str], dict): list of mappings used and dict with translated metadata according to the CodeMeta vocabulary """ used_mappings = [MAPPINGS[context].name for context in detected_files] metadata = [] tool = { 'name': 'swh-metadata-translator', 'version': '0.0.2', 'configuration': { }, } # TODO: iterate on each context, on each file # -> get raw_contents # -> translate each content config = { k: self.config[k] for k in [INDEXER_CFG_KEY, 'objstorage', 'storage'] } config['tools'] = [tool] for context in detected_files.keys(): cfg = deepcopy(config) cfg['tools'][0]['configuration']['context'] = context c_metadata_indexer = ContentMetadataIndexer(config=cfg) # sha1s that are in content_metadata table sha1s_in_storage = [] metadata_generator = self.idx_storage.content_metadata_get( detected_files[context]) for c in metadata_generator: # extracting metadata sha1 = c['id'] sha1s_in_storage.append(sha1) local_metadata = c['metadata'] # local metadata is aggregated if local_metadata: metadata.append(local_metadata) sha1s_filtered = [item for item in detected_files[context] if item not in sha1s_in_storage] if sha1s_filtered: # content indexing try: c_metadata_indexer.run(sha1s_filtered, policy_update='ignore-dups', log_suffix=log_suffix) # on the fly possibility: for result in c_metadata_indexer.results: local_metadata = result['metadata'] metadata.append(local_metadata) except Exception: self.log.exception( "Exception while indexing metadata on contents") metadata = merge_documents(metadata) return (used_mappings, metadata) class OriginMetadataIndexer(OriginIndexer): ADDITIONAL_CONFIG = RevisionMetadataIndexer.ADDITIONAL_CONFIG USE_TOOLS = False - def __init__(self, config=None, **kwargs): + def __init__(self, config=None, **kwargs) -> None: super().__init__(config=config, **kwargs) self.origin_head_indexer = OriginHeadIndexer(config=config) self.revision_metadata_indexer = RevisionMetadataIndexer(config=config) def index_list(self, origin_urls): head_rev_ids = [] origins_with_head = [] origins = list(call_with_batches( self.storage.origin_get, [{'url': url} for url in origin_urls], ORIGIN_GET_BATCH_SIZE)) for origin in origins: head_result = self.origin_head_indexer.index(origin['url']) if head_result: origins_with_head.append(origin) head_rev_ids.append(head_result['revision_id']) head_revs = list(call_with_batches( self.storage.revision_get, head_rev_ids, REVISION_GET_BATCH_SIZE)) assert len(head_revs) == len(head_rev_ids) results = [] for (origin, rev) in zip(origins_with_head, head_revs): if not rev: self.log.warning('Missing head revision of origin %r', origin['url']) continue rev_metadata = self.revision_metadata_indexer.index(rev) orig_metadata = { 'from_revision': rev_metadata['id'], 'id': origin['url'], 'metadata': rev_metadata['metadata'], 'mappings': rev_metadata['mappings'], 'indexer_configuration_id': rev_metadata['indexer_configuration_id'], } results.append((orig_metadata, rev_metadata)) return results - def persist_index_computations(self, results, policy_update): + def persist_index_computations( + self, results: List[Dict], policy_update: str + ) -> None: conflict_update = (policy_update == 'update-dups') # Deduplicate revisions - rev_metadata = [] - orig_metadata = [] - revs_to_delete = [] - origs_to_delete = [] + rev_metadata: List[Any] = [] + orig_metadata: List[Any] = [] + revs_to_delete: List[Any] = [] + origs_to_delete: List[Any] = [] for (orig_item, rev_item) in results: assert rev_item['metadata'] == orig_item['metadata'] if not rev_item['metadata'] or \ rev_item['metadata'].keys() <= {'@context'}: # If we didn't find any metadata, don't store a DB record # (and delete existing ones, if any) if rev_item not in revs_to_delete: revs_to_delete.append(rev_item) if orig_item not in origs_to_delete: origs_to_delete.append(orig_item) else: if rev_item not in rev_metadata: rev_metadata.append(rev_item) if orig_item not in orig_metadata: orig_metadata.append(orig_item) if rev_metadata: self.idx_storage.revision_intrinsic_metadata_add( rev_metadata, conflict_update=conflict_update) if orig_metadata: self.idx_storage.origin_intrinsic_metadata_add( orig_metadata, conflict_update=conflict_update) # revs_to_delete should always be empty unless we changed a mapping # to detect less files or less content. # However, origs_to_delete may be empty whenever an upstream deletes # a metadata file. if origs_to_delete: self.idx_storage.origin_intrinsic_metadata_delete(origs_to_delete) if revs_to_delete: self.idx_storage.revision_intrinsic_metadata_delete(revs_to_delete) diff --git a/swh/indexer/mimetype.py b/swh/indexer/mimetype.py index d8dda33..0d391bb 100644 --- a/swh/indexer/mimetype.py +++ b/swh/indexer/mimetype.py @@ -1,147 +1,152 @@ -# Copyright (C) 2016-2018 The Software Heritage developers +# Copyright (C) 2016-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information +from typing import Optional, Dict, Any, List import magic -from typing import Optional - from .indexer import ContentIndexer, ContentRangeIndexer if not hasattr(magic.Magic, 'from_buffer'): raise ImportError( 'Expected "import magic" to import python-magic, but file_magic ' 'was imported instead.') -def compute_mimetype_encoding(raw_content): +def compute_mimetype_encoding(raw_content: bytes) -> Dict[str, bytes]: """Determine mimetype and encoding from the raw content. Args: - raw_content (bytes): content's raw data + raw_content: content's raw data Returns: - dict: mimetype and encoding key and corresponding values - (as bytes). + dict: mimetype and encoding key and corresponding values. """ m = magic.Magic(mime=True, mime_encoding=True) res = m.from_buffer(raw_content) (mimetype, encoding) = res.split('; charset=') return { 'mimetype': mimetype, 'encoding': encoding, } class MixinMimetypeIndexer: """Mixin mimetype indexer. See :class:`MimetypeIndexer` and :class:`MimetypeRangeIndexer` """ + tool: Dict[str, Any] + idx_storage: Any ADDITIONAL_CONFIG = { 'tools': ('dict', { 'name': 'file', 'version': '1:5.30-1+deb9u1', 'configuration': { "type": "library", "debian-package": "python3-magic" }, }), 'write_batch_size': ('int', 1000), } CONFIG_BASE_FILENAME = 'indexer/mimetype' # type: Optional[str] - def index(self, id, data): + def index(self, id: bytes, data: bytes) -> Dict[str, Any]: """Index sha1s' content and store result. Args: - id (bytes): content's identifier - data (bytes): raw content in bytes + id: content's identifier + data: raw content in bytes Returns: dict: content's mimetype; dict keys being - - **id** (bytes): content's identifier (sha1) - - **mimetype** (bytes): mimetype in bytes - - **encoding** (bytes): encoding in bytes + - id: content's identifier (sha1) + - mimetype: mimetype in bytes + - encoding: encoding in bytes """ properties = compute_mimetype_encoding(data) properties.update({ 'id': id, 'indexer_configuration_id': self.tool['id'], - }) + }) return properties - def persist_index_computations(self, results, policy_update): + def persist_index_computations( + self, results: List[Dict], policy_update: List[str] + ) -> None: """Persist the results in storage. Args: - results ([dict]): list of content's mimetype dicts + results: list of content's mimetype dicts (see :meth:`.index`) - policy_update ([str]): either 'update-dups' or 'ignore-dups' to + policy_update: either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them """ self.idx_storage.content_mimetype_add( results, conflict_update=(policy_update == 'update-dups')) class MimetypeIndexer(MixinMimetypeIndexer, ContentIndexer): """Mimetype Indexer working on list of content identifiers. It: - (optionally) filters out content already indexed (cf. :meth:`.filter`) - reads content from objstorage per the content's id (sha1) - computes {mimetype, encoding} from that content - stores result in storage """ def filter(self, ids): """Filter out known sha1s and return only missing ones. """ yield from self.idx_storage.content_mimetype_missing(( { 'id': sha1, 'indexer_configuration_id': self.tool['id'], } for sha1 in ids )) class MimetypeRangeIndexer(MixinMimetypeIndexer, ContentRangeIndexer): """Mimetype Range Indexer working on range of content identifiers. It: - (optionally) filters out content already indexed (cf :meth:`.indexed_contents_in_range`) - reads content from objstorage per the content's id (sha1) - computes {mimetype, encoding} from that content - stores result in storage """ - def indexed_contents_in_range(self, start, end): + + def indexed_contents_in_range( + self, start: bytes, end: bytes + ) -> Dict[str, Optional[bytes]]: """Retrieve indexed content id within range [start, end]. Args: - start (bytes): Starting bound from range identifier - end (bytes): End range identifier + start: Starting bound from range identifier + end: End range identifier Returns: dict: a dict with keys: - - **ids** [bytes]: iterable of content ids within the range. - - **next** (Optional[bytes]): The next range of sha1 starts at + - ids: iterable of content ids within the range. + - next: The next range of sha1 starts at this sha1 if any """ return self.idx_storage.content_mimetype_get_range( start, end, self.tool['id']) diff --git a/swh/indexer/origin_head.py b/swh/indexer/origin_head.py index 5a6431b..707d643 100644 --- a/swh/indexer/origin_head.py +++ b/swh/indexer/origin_head.py @@ -1,150 +1,158 @@ -# Copyright (C) 2018 The Software Heritage developers +# Copyright (C) 2018-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information +from typing import List, Tuple, Any, Dict, Union + import re import click import logging from swh.indexer.indexer import OriginIndexer class OriginHeadIndexer(OriginIndexer): """Origin-level indexer. This indexer is in charge of looking up the revision that acts as the "head" of an origin. In git, this is usually the commit pointed to by the 'master' branch.""" USE_TOOLS = False - def persist_index_computations(self, results, policy_update): + def persist_index_computations( + self, results: Any, policy_update: str + ) -> None: """Do nothing. The indexer's results are not persistent, they should only be piped to another indexer.""" pass # Dispatch def index(self, origin_url): latest_visit = self.storage.origin_visit_get_latest( origin_url, allowed_statuses=['full'], require_snapshot=True) if latest_visit is None: return None latest_snapshot = self.storage.snapshot_get(latest_visit['snapshot']) method = getattr( self, '_try_get_%s_head' % latest_visit['type'], self._try_get_head_generic) rev_id = method(latest_snapshot) if rev_id is not None: return { 'origin_url': origin_url, 'revision_id': rev_id, } # could not find a head revision return None # Tarballs _archive_filename_re = re.compile( rb'^' rb'(?P.*)[-_]' rb'(?P[0-9]+(\.[0-9])*)' rb'(?P[-+][a-zA-Z0-9.~]+?)?' rb'(?P(\.[a-zA-Z0-9]+)+)' rb'$') @classmethod - def _parse_version(cls, filename): + def _parse_version( + cls: Any, filename: str + ) -> Tuple[Union[float, int], ...]: """Extracts the release version from an archive filename, to get an ordering whose maximum is likely to be the last version of the software >>> OriginHeadIndexer._parse_version(b'foo') (-inf,) >>> OriginHeadIndexer._parse_version(b'foo.tar.gz') (-inf,) >>> OriginHeadIndexer._parse_version(b'gnu-hello-0.0.1.tar.gz') (0, 0, 1, 0) >>> OriginHeadIndexer._parse_version(b'gnu-hello-0.0.1-beta2.tar.gz') (0, 0, 1, -1, 'beta2') >>> OriginHeadIndexer._parse_version(b'gnu-hello-0.0.1+foobar.tar.gz') (0, 0, 1, 1, 'foobar') """ res = cls._archive_filename_re.match(filename) if res is None: return (float('-infinity'),) version = [int(n) for n in res.group('version').decode().split('.')] if res.group('preversion') is None: version.append(0) else: preversion = res.group('preversion').decode() if preversion.startswith('-'): version.append(-1) version.append(preversion[1:]) elif preversion.startswith('+'): version.append(1) version.append(preversion[1:]) else: assert False, res.group('preversion') return tuple(version) - def _try_get_ftp_head(self, snapshot): + def _try_get_ftp_head(self, snapshot: Dict[str, Any]) -> Any: archive_names = list(snapshot['branches']) max_archive_name = max(archive_names, key=self._parse_version) r = self._try_resolve_target(snapshot['branches'], max_archive_name) return r # Generic - def _try_get_head_generic(self, snapshot): + def _try_get_head_generic( + self, snapshot: Dict[str, Any] + ) -> Any: # Works on 'deposit', 'pypi', and VCSs. try: branches = snapshot['branches'] except KeyError: return None else: return ( self._try_resolve_target(branches, b'HEAD') or self._try_resolve_target(branches, b'master') ) - def _try_resolve_target(self, branches, target_name): + def _try_resolve_target(self, branches: Dict, target_name: bytes) -> Any: try: target = branches[target_name] if target is None: return None while target['target_type'] == 'alias': target = branches[target['target']] if target is None: return None if target['target_type'] == 'revision': return target['target'] elif target['target_type'] == 'content': return None # TODO elif target['target_type'] == 'directory': return None # TODO elif target['target_type'] == 'release': return None # TODO else: assert False except KeyError: return None @click.command() @click.option('--origins', '-i', help='Origins to lookup, in the "type+url" format', multiple=True) -def main(origins): +def main(origins: List[str]) -> None: rev_metadata_indexer = OriginHeadIndexer() rev_metadata_indexer.run(origins) if __name__ == '__main__': logging.basicConfig(level=logging.INFO) main() diff --git a/swh/indexer/rehash.py b/swh/indexer/rehash.py index 3ef4ab3..038b63e 100644 --- a/swh/indexer/rehash.py +++ b/swh/indexer/rehash.py @@ -1,173 +1,177 @@ -# Copyright (C) 2017-2018 The Software Heritage developers +# Copyright (C) 2017-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import logging import itertools from collections import defaultdict +from typing import Dict, Any, Tuple, List, Generator from swh.core import utils from swh.core.config import SWHConfig from swh.model import hashutil from swh.objstorage import get_objstorage from swh.objstorage.exc import ObjNotFoundError from swh.storage import get_storage class RecomputeChecksums(SWHConfig): """Class in charge of (re)computing content's hashes. Hashes to compute are defined across 2 configuration options: compute_checksums ([str]) list of hash algorithms that py:func:`swh.model.hashutil.MultiHash.from_data` function should be able to deal with. For variable-length checksums, a desired checksum length should also be provided. Their format is : e.g: blake2:512 recompute_checksums (bool) a boolean to notify that we also want to recompute potential existing hashes specified in compute_checksums. Default to False. """ DEFAULT_CONFIG = { # The storage to read from or update metadata to 'storage': ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5002/' }, }), # The objstorage to read contents' data from 'objstorage': ('dict', { 'cls': 'pathslicing', 'args': { 'root': '/srv/softwareheritage/objects', 'slicing': '0:2/2:4/4:6', }, }), # the set of checksums that should be computed. # Examples: 'sha1_git', 'blake2b512', 'blake2s256' 'compute_checksums': ( 'list[str]', []), # whether checksums that already exist in the DB should be # recomputed/updated or left untouched 'recompute_checksums': ('bool', False), # Number of contents to retrieve blobs at the same time 'batch_size_retrieve_content': ('int', 10), # Number of contents to update at the same time 'batch_size_update': ('int', 100), } CONFIG_BASE_FILENAME = 'indexer/rehash' - def __init__(self): + def __init__(self) -> None: self.config = self.parse_config_file() self.storage = get_storage(**self.config['storage']) self.objstorage = get_objstorage(**self.config['objstorage']) self.compute_checksums = self.config['compute_checksums'] self.recompute_checksums = self.config[ 'recompute_checksums'] self.batch_size_retrieve_content = self.config[ 'batch_size_retrieve_content'] self.batch_size_update = self.config[ 'batch_size_update'] self.log = logging.getLogger('swh.indexer.rehash') if not self.compute_checksums: raise ValueError('Checksums list should not be empty.') - def _read_content_ids(self, contents): + def _read_content_ids( + self, contents: List[Dict[str, Any]] + ) -> Generator[bytes, Any, None]: """Read the content identifiers from the contents. """ for c in contents: h = c['sha1'] if isinstance(h, str): h = hashutil.hash_to_bytes(h) yield h - def get_new_contents_metadata(self, all_contents): + def get_new_contents_metadata( + self, all_contents: List[Dict[str, Any]] + ) -> Generator[Tuple[Dict[str, Any], List[Any]], Any, None]: """Retrieve raw contents and compute new checksums on the contents. Unknown or corrupted contents are skipped. Args: - all_contents ([dict]): List of contents as dictionary with + all_contents: List of contents as dictionary with the necessary primary keys - checksum_algorithms ([str]): List of checksums to compute Yields: tuple: tuple of (content to update, list of checksums computed) """ content_ids = self._read_content_ids(all_contents) for contents in utils.grouper(content_ids, self.batch_size_retrieve_content): contents_iter = itertools.tee(contents, 2) try: content_metadata = self.storage.content_get_metadata( [s for s in contents_iter[0]]) except Exception: self.log.exception( 'Problem when reading contents metadata.') continue for content in content_metadata: # Recompute checksums provided in compute_checksums options if self.recompute_checksums: checksums_to_compute = list(self.compute_checksums) else: # Compute checksums provided in compute_checksums # options not already defined for that content checksums_to_compute = [h for h in self.compute_checksums if not content.get(h)] if not checksums_to_compute: # Nothing to recompute continue try: raw_content = self.objstorage.get(content['sha1']) except ObjNotFoundError: self.log.warning('Content %s not found in objstorage!' % content['sha1']) continue content_hashes = hashutil.MultiHash.from_data( raw_content, hash_names=checksums_to_compute).digest() content.update(content_hashes) yield content, checksums_to_compute - def run(self, contents): + def run(self, contents: List[Dict[str, Any]]) -> None: """Given a list of content: - (re)compute a given set of checksums on contents available in our object storage - update those contents with the new metadata Args: - contents (dict): contents as dictionary with necessary keys. + contents: contents as dictionary with necessary keys. key present in such dictionary should be the ones defined in the 'primary_key' option. """ for data in utils.grouper( self.get_new_contents_metadata(contents), self.batch_size_update): - groups = defaultdict(list) + groups: Dict[str, List[Any]] = defaultdict(list) for content, keys_to_update in data: keys = ','.join(keys_to_update) groups[keys].append(content) for keys_to_update, contents in groups.items(): keys = keys_to_update.split(',') try: self.storage.content_update(contents, keys=keys) except Exception: self.log.exception('Problem during update.') continue