diff --git a/swh/indexer/indexer.py b/swh/indexer/indexer.py index 378893a..1755953 100644 --- a/swh/indexer/indexer.py +++ b/swh/indexer/indexer.py @@ -1,418 +1,427 @@ # Copyright (C) 2016-2017 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 +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.model import hashutil from swh.scheduler.utils import get_task from swh.indexer.storage import get_indexer_storage, INDEXER_CFG_KEY class DiskIndexer: """Mixin intended to be used with other SomethingIndexer classes. Indexers inheriting from this class are a category of indexers which needs the disk for their computations. Note: This expects `self.working_directory` variable defined at runtime. """ def write_to_temp(self, filename, data): """Write the sha1's content in a temporary file. Args: sha1 (str): the sha1 name filename (str): one of sha1's many filenames data (bytes): the sha1's content to write in temporary file Returns: The path to the temporary file created. That file is filled in with the raw content's data. """ os.makedirs(self.working_directory, exist_ok=True) temp_dir = tempfile.mkdtemp(dir=self.working_directory) content_path = os.path.join(temp_dir, filename) with open(content_path, 'wb') as f: f.write(data) return content_path def cleanup(self, content_path): """Remove content_path from working directory. Args: content_path (str): the file to remove """ temp_dir = os.path.dirname(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 the process of indexation: :func:`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` (later on :class:`OriginIndexer` will also be available) Then you need to implement the following functions: :func:`filter`: filter out data already indexed (in storage). This function is used by the orchestrator and not directly by the indexer (cf. swh.indexer.orchestrator.BaseOrchestratorIndexer). :func:`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. :func:`persist_index_computations`: persist the results of multiple index computations in the storage. The new indexer implementation can also override the following functions: :func:`prepare`: Configuration preparation for the indexer. When overriding, this must call the `super().prepare()` instruction. :func:`check`: Configuration check for the indexer. When overriding, this must call the `super().check()` instruction. :func:`register_tools`: This should return a dict of the tool(s) to use when indexing or filtering. """ CONFIG = 'indexer/base' DEFAULT_CONFIG = { INDEXER_CFG_KEY: ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5007/' } }), # queue to reschedule if problem (none for no rescheduling, # the default) 'rescheduling_task': ('str', None), + 'storage': ('dict', { + 'cls': 'remote', + 'args': { + 'url': 'http://localhost:5002/', + } + }), 'objstorage': ('dict', { 'cls': 'multiplexer', 'args': { 'objstorages': [{ 'cls': 'filtered', 'args': { 'storage_conf': { 'cls': 'azure', 'args': { 'account_name': '0euwestswh', 'api_secret_key': 'secret', 'container_name': 'contents' } }, 'filters_conf': [ {'type': 'readonly'}, {'type': 'prefix', 'prefix': '0'} ] } }, { 'cls': 'filtered', 'args': { 'storage_conf': { 'cls': 'azure', 'args': { 'account_name': '1euwestswh', 'api_secret_key': 'secret', 'container_name': 'contents' } }, 'filters_conf': [ {'type': 'readonly'}, {'type': 'prefix', 'prefix': '1'} ] } }] }, }), } ADDITIONAL_CONFIG = {} def __init__(self): """Prepare and check that the indexer is ready to run. """ super().__init__() self.prepare() self.check() def prepare(self): """Prepare the indexer's needed runtime configuration. Without this step, the indexer cannot possibly run. """ self.config = self.parse_config_file( additional_configs=[self.ADDITIONAL_CONFIG]) + if self.config['storage']: + self.storage = get_storage(**self.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) rescheduling_task = self.config['rescheduling_task'] if rescheduling_task: self.rescheduling_task = get_task(rescheduling_task) else: self.rescheduling_task = None _log = logging.getLogger('requests.packages.urllib3.connectionpool') _log.setLevel(logging.WARN) self.log = logging.getLogger('swh.indexer') self.tools = list(self.register_tools(self.config['tools'])) def check(self): """Check the indexer's configuration is ok before proceeding. If ok, does nothing. If not raise error. """ if not self.tools: raise ValueError('Tools %s is unknown, cannot continue' % self.tools) def _prepare_tool(self, tool): """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): """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. Returns: List of dict with additional id key. Raises: ValueError if not a list nor a dict. """ tools = self.config['tools'] if isinstance(tools, list): tools = 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!') return self.idx_storage.indexer_configuration_add(tools) @abc.abstractmethod def filter(self, ids): """Filter missing ids for that particular indexer. Args: ids ([bytes]): list of ids Yields: iterator of missing ids """ pass @abc.abstractmethod def index(self, id, data): """Index computation for the id and associated raw data. Args: id (bytes): identifier data (bytes): id's data from storage or objstorage depending on object type Returns: a dict that makes sense for the persist_index_computations function. """ pass @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): """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 by index function. Returns: None """ pass @abc.abstractmethod def run(self, ids, policy_update): """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 """ pass class ContentIndexer(BaseIndexer): """An object type indexer, inherits from the :class:`BaseIndexer` and implements the process of indexation for Contents using the run method Note: the :class:`ContentIndexer` 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): """Given a list of ids: - retrieve the content from the storage - execute the indexing computations - store the results (according to policy_update) Args: ids ([bytes]): sha1's identifier list policy_update ([str]): either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them """ results = [] try: for sha1 in ids: try: raw_content = self.objstorage.get(sha1) except ObjNotFoundError: self.log.warn('Content %s not found in objstorage' % hashutil.hash_to_hex(sha1)) continue res = self.index(sha1, raw_content) if res: # If no results, skip it results.append(res) self.persist_index_computations(results, policy_update) self.next_step(results) except Exception: self.log.exception( 'Problem when reading contents metadata.') if self.rescheduling_task: self.log.warn('Rescheduling batch') self.rescheduling_task.delay(ids, policy_update) class RevisionIndexer(BaseIndexer): """An object type indexer, inherits from the :class:`BaseIndexer` and implements the process of indexation for Revisions 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): """Given a list of sha1_gits: - retrieve revisions from storage - execute the indexing computations - store the results (according to policy_update) Args: ids ([bytes]): sha1_git's identifier list policy_update ([str]): either 'update-dups' or 'ignore-dups' to respectively update duplicates or ignore them """ results = [] revs = self.storage.revision_get(ids) for rev in revs: if not rev: self.log.warn('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: self.log.exception( 'Problem when processing revision') self.persist_index_computations(results, policy_update) diff --git a/swh/indexer/metadata.py b/swh/indexer/metadata.py index 6f41e50..77fdfb9 100644 --- a/swh/indexer/metadata.py +++ b/swh/indexer/metadata.py @@ -1,305 +1,294 @@ # Copyright (C) 2017 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 click import logging from swh.indexer.indexer import ContentIndexer, RevisionIndexer from swh.indexer.metadata_dictionary import compute_metadata from swh.indexer.metadata_detector import detect_metadata from swh.indexer.metadata_detector import extract_minimal_metadata_dict from swh.indexer.storage import INDEXER_CFG_KEY -from swh.storage import get_storage from swh.model import hashutil 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 translated_metadata by given context - using the metadata_dictionary as the 'swh-metadata-translator' tool - store result in content_metadata table """ CONFIG_BASE_FILENAME = 'indexer/metadata' def __init__(self, tool, config): # twisted way to use the exact same config of RevisionMetadataIndexer # object that uses internally ContentMetadataIndexer self.config = config self.config['tools'] = tool super().__init__() def prepare(self): self.results = [] if self.config[INDEXER_CFG_KEY]: self.idx_storage = self.config[INDEXER_CFG_KEY] if self.config['objstorage']: self.objstorage = self.config['objstorage'] _log = logging.getLogger('requests.packages.urllib3.connectionpool') _log.setLevel(logging.WARN) self.log = logging.getLogger('swh.indexer') self.tools = self.register_tools(self.config['tools']) # NOTE: only one tool so far, change when no longer true self.tool = self.tools[0] 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): """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 translated_metadata keys will be returned as None """ result = { 'id': id, 'indexer_configuration_id': self.tool['id'], 'translated_metadata': None } try: context = self.tool['tool_configuration']['context'] result['translated_metadata'] = compute_metadata(context, data) # a twisted way to keep result with indexer object for get_results self.results.append(result) except Exception: self.log.exception( "Problem during tool retrieval of metadata translation") return result def persist_index_computations(self, results, policy_update): """Persist the results in storage. Args: results ([dict]): list of content_metadata, dict with the following keys: - id (bytes): content's identifier (sha1) - translated_metadata (jsonb): detected metadata policy_update ([str]): 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')) def get_results(self): """can be called only if run method was called before Returns: list: list of content_metadata entries calculated by current indexer """ return self.results class RevisionMetadataIndexer(RevisionIndexer): """Revision-level indexer This indexer is in charge of: - filtering revisions already indexed in revision_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 """ CONFIG_BASE_FILENAME = 'indexer/metadata' ADDITIONAL_CONFIG = { - 'storage': ('dict', { - 'cls': 'remote', - 'args': { - 'url': 'http://localhost:5002/', - } - }), 'tools': ('dict', { 'name': 'swh-metadata-detector', 'version': '0.0.1', 'configuration': { 'type': 'local', 'context': ['npm', 'codemeta'] }, }), } - def __init__(self, *args, **kwargs): - super().__init__(*args, **kwargs) - self.storage = get_storage(**self.config['storage']) - def prepare(self): super().prepare() self.tool = self.tools[0] def filter(self, sha1_gits): """Filter out known sha1s and return only missing ones. """ yield from self.idx_storage.revision_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 (bytes): revision artifact from storage Returns: dict: dictionary representing a revision_metadata, with keys: - id (bytes): rev's identifier (sha1_git) - indexer_configuration_id (bytes): tool used - translated_metadata (bytes): dict of retrieved metadata """ try: result = { 'id': rev['id'], 'indexer_configuration_id': self.tool['id'], 'translated_metadata': None } root_dir = rev['directory'] dir_ls = self.storage.directory_ls(root_dir, recursive=False) files = (entry for entry in dir_ls if entry['type'] == 'file') detected_files = detect_metadata(files) result['translated_metadata'] = self.translate_revision_metadata( detected_files) except Exception as e: self.log.exception( 'Problem when indexing rev') return result def persist_index_computations(self, results, policy_update): """Persist the results in storage. Args: results ([dict]): 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 respectively update duplicates or ignore them """ # TODO: add functions in storage to keep data in revision_metadata self.idx_storage.revision_metadata_add( results, conflict_update=(policy_update == 'update-dups')) def translate_revision_metadata(self, detected_files): """ Determine plan of action to translate metadata when containing one or multiple detected files: Args: detected_files (dict): dictionary mapping context names (e.g., "npm", "authors") to list of sha1 Returns: dict: dict with translated metadata according to the CodeMeta vocabulary """ translated_metadata = [] tool = { 'name': 'swh-metadata-translator', 'version': '0.0.1', 'configuration': { 'type': 'local', 'context': None }, } # TODO: iterate on each context, on each file # -> get raw_contents # -> translate each content config = { INDEXER_CFG_KEY: self.idx_storage, 'objstorage': self.objstorage } for context in detected_files.keys(): tool['configuration']['context'] = context c_metadata_indexer = ContentMetadataIndexer(tool, config) # 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 translated_metadata sha1 = c['id'] sha1s_in_storage.append(sha1) local_metadata = c['translated_metadata'] # local metadata is aggregated if local_metadata: translated_metadata.append(local_metadata) sha1s_filtered = [item for item in detected_files[context] if item not in sha1s_in_storage] if sha1s_filtered: # schedule indexation of content try: c_metadata_indexer.run(sha1s_filtered, policy_update='ignore-dups') # on the fly possibility: results = c_metadata_indexer.get_results() for result in results: local_metadata = result['translated_metadata'] translated_metadata.append(local_metadata) except Exception as e: self.log.warn("""Exception while indexing content""", e) # transform translated_metadata into min set with swh-metadata-detector min_metadata = extract_minimal_metadata_dict(translated_metadata) return min_metadata @click.command() @click.option('--revs', '-i', default=['8dbb6aeb036e7fd80664eb8bfd1507881af1ba9f', '026040ea79dec1b49b4e3e7beda9132b6b26b51b', '9699072e21eded4be8d45e3b8d543952533fa190'], help='Default sha1_git to lookup', multiple=True) def main(revs): _git_sha1s = list(map(hashutil.hash_to_bytes, revs)) rev_metadata_indexer = RevisionMetadataIndexer() rev_metadata_indexer.run(_git_sha1s, 'update-dups') if __name__ == '__main__': logging.basicConfig(level=logging.INFO) main()