diff --git a/swh/lister/bitbucket/lister.py b/swh/lister/bitbucket/lister.py index 7bc2c1a..6885c8e 100644 --- a/swh/lister/bitbucket/lister.py +++ b/swh/lister/bitbucket/lister.py @@ -1,36 +1,37 @@ # Copyright (C) 2017-2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from urllib import parse from swh.lister.bitbucket.models import BitBucketModel from swh.lister.core.indexing_lister import SWHIndexingHttpLister class BitBucketLister(SWHIndexingHttpLister): PATH_TEMPLATE = '/repositories?after=%s' MODEL = BitBucketModel + LISTER_NAME = 'bitbucket.com' def get_model_from_repo(self, repo): return { 'uid': repo['uuid'], 'indexable': repo['created_on'], 'name': repo['name'], 'full_name': repo['full_name'], 'html_url': repo['links']['html']['href'], 'origin_url': repo['links']['clone'][0]['href'], 'origin_type': repo['scm'], 'description': repo['description'] } def get_next_target_from_response(self, response): body = response.json() if 'next' in body: return parse.unquote(body['next'].split('after=')[1]) else: return None def transport_response_simplified(self, response): repos = response.json()['values'] return [self.get_model_from_repo(repo) for repo in repos] diff --git a/swh/lister/bitbucket/tasks.py b/swh/lister/bitbucket/tasks.py index 74eab7b..e480994 100644 --- a/swh/lister/bitbucket/tasks.py +++ b/swh/lister/bitbucket/tasks.py @@ -1,28 +1,27 @@ # Copyright (C) 2017-2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from swh.lister.core.tasks import (IndexingDiscoveryListerTask, RangeListerTask, IndexingRefreshListerTask, ListerTaskBase) from .lister import BitBucketLister class BitBucketListerTask(ListerTaskBase): def new_lister(self): - return BitBucketLister(lister_name='bitbucket.com', - api_baseurl='https://api.bitbucket.org/2.0') + return BitBucketLister(api_baseurl='https://api.bitbucket.org/2.0') class IncrementalBitBucketLister(BitBucketListerTask, IndexingDiscoveryListerTask): task_queue = 'swh_lister_bitbucket_discover' class RangeBitBucketLister(BitBucketListerTask, RangeListerTask): task_queue = 'swh_lister_bitbucket_refresh' class FullBitBucketRelister(BitBucketListerTask, IndexingRefreshListerTask): task_queue = 'swh_lister_bitbucket_refresh' diff --git a/swh/lister/cli.py b/swh/lister/cli.py index 86b4bd8..1f0d6f1 100644 --- a/swh/lister/cli.py +++ b/swh/lister/cli.py @@ -1,59 +1,56 @@ # Copyright (C) 2018 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 @click.command() @click.option( '--db-url', '-d', default='postgres:///lister-gitlab.com', help='SQLAlchemy DB URL; see ' '') # noqa @click.option('--lister', required=1, type=click.Choice(['github', 'gitlab', 'bitbucket']), help='Lister to act upon') @click.option('--create-tables', is_flag=True, default=False, help='create tables') @click.option('--drop-tables', is_flag=True, default=False, help='Drop tables') def cli(db_url, lister, create_tables, drop_tables): """Initialize db model according to lister. """ supported_listers = ['github', 'gitlab', 'bitbucket'] override_conf = {'lister_db_url': db_url} if lister == 'github': from .github import models from .github.lister import GitHubLister - _lister = GitHubLister(lister_name='github.com', - api_baseurl='https://api.github.com', + _lister = GitHubLister(api_baseurl='https://api.github.com', override_config=override_conf) elif lister == 'bitbucket': from .bitbucket import models from .bitbucket.lister import BitBucketLister - _lister = BitBucketLister(lister_name='bitbucket.com', - api_baseurl='https://api.bitbucket.org/2.0', + _lister = BitBucketLister(api_baseurl='https://api.bitbucket.org/2.0', override_config=override_conf) elif lister == 'gitlab': from .gitlab import models from .gitlab.lister import GitLabLister - _lister = GitLabLister(lister_name='gitlab.com', - api_baseurl='https://gitlab.com/api/v4/', + _lister = GitLabLister(api_baseurl='https://gitlab.com/api/v4/', override_config=override_conf) else: raise ValueError('Only supported listers are %s' % supported_listers) if drop_tables: models.ModelBase.metadata.drop_all(_lister.db_engine) if create_tables: models.ModelBase.metadata.create_all(_lister.db_engine) if __name__ == '__main__': cli() diff --git a/swh/lister/core/indexing_lister.py b/swh/lister/core/indexing_lister.py index 8166900..d086382 100644 --- a/swh/lister/core/indexing_lister.py +++ b/swh/lister/core/indexing_lister.py @@ -1,212 +1,210 @@ # Copyright (C) 2015-2017 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import logging from sqlalchemy import func from .lister_transports import SWHListerHttpTransport from .lister_base import SWHListerBase class SWHIndexingLister(SWHListerBase): """Lister* intermediate class for any service that follows the pattern: - The service must report at least one stable unique identifier, known herein as the UID value, for every listed repository. - If the service splits the list of repositories into sublists, it must report at least one stable and sorted index identifier for every listed repository, known herein as the indexable value, which can be used as part of the service endpoint query to request a sublist beginning from that index. This might be the UID if the UID is monotonic. - Client sends a request to list repositories starting from a given index. - Client receives structured (json/xml/etc) response with information about a sequential series of repositories starting from that index and, if necessary/available, some indication of the URL or index for fetching the next series of repository data. See :class:`swh.lister.core.lister_base.SWHListerBase` for more details. This class cannot be instantiated. To create a new Lister for a source code listing service that follows the model described above, you must subclass this class and provide the required overrides in addition to any unmet implementation/override requirements of this class's base. (see parent class and member docstrings for details) Required Overrides:: def get_next_target_from_response """ @abc.abstractmethod def get_next_target_from_response(self, response): """Find the next server endpoint identifier given the entire response. Implementation of this method depends on the server API spec and the shape of the network response object returned by the transport_request method. Args: response (transport response): response page from the server Returns: index of next page, possibly extracted from a next href url """ pass # You probably don't need to override anything below this line. def filter_before_inject(self, models_list): """Overrides SWHListerBase.filter_before_inject Bounds query results by this Lister's set max_index. """ models_list = [ m for m in models_list if self.is_within_bounds(m['indexable'], None, self.max_index) ] return models_list def db_query_range(self, start, end): """Look in the db for a range of repositories with indexable values in the range [start, end] Args: start (model indexable type): start of desired indexable range end (model indexable type): end of desired indexable range Returns: a list of sqlalchemy.ext.declarative.declarative_base objects with indexable values within the given range """ retlist = self.db_session.query(self.MODEL) if start is not None: retlist = retlist.filter(self.MODEL.indexable >= start) if end is not None: retlist = retlist.filter(self.MODEL.indexable <= end) return retlist def db_partition_indices(self, partition_size): """Describe an index-space compartmentalization of the db table in equal sized chunks. This is used to describe min&max bounds for parallelizing fetch tasks. Args: partition_size (int): desired size to make each partition Returns: a list of tuples (begin, end) of indexable value that declare approximately equal-sized ranges of existing repos """ n = self.db_num_entries() partitions = [] partition_size = min(partition_size, n) prev_index = None for i in range(0, n-1, partition_size): # indexable column from the ith row index = self.db_session.query(self.MODEL.indexable) \ .order_by(self.MODEL.indexable).offset(i).first() if index is not None and prev_index is not None: partitions.append((prev_index, index)) prev_index = index partitions.append((prev_index, self.db_last_index())) return partitions def db_last_index(self): """Look in the db for the largest indexable value Returns: the largest indexable value of all repos in the db """ t = self.db_session.query(func.max(self.MODEL.indexable)).first() if t: return t[0] else: return None def disable_deleted_repo_tasks(self, start, end, keep_these): """Disable tasks for repos that no longer exist between start and end. Args: start: beginning of range to disable end: end of range to disable keep_these (uid list): do not disable repos with uids in this list """ if end is None: end = self.db_last_index() if not self.is_within_bounds(end, None, self.max_index): end = self.max_index deleted_repos = self.winnow_models( self.db_query_range(start, end), self.MODEL.uid, keep_these ) tasks_to_disable = [repo.task_id for repo in deleted_repos if repo.task_id is not None] if tasks_to_disable: self.scheduler.disable_tasks(tasks_to_disable) for repo in deleted_repos: repo.task_id = None def run(self, min_index=None, max_index=None): """Main entry function. Sequentially fetches repository data from the service according to the basic outline in the class docstring, continually fetching sublists until either there is no next index reference given or the given next index is greater than the desired max_index. Args: min_index (indexable type): optional index to start from max_index (indexable type): optional index to stop at Returns: nothing """ index = min_index or '' loop_count = 0 self.min_index = min_index self.max_index = max_index while self.is_within_bounds(index, self.min_index, self.max_index): logging.info('listing repos starting at %s' % index) response, injected_repos = self.ingest_data(index) next_index = self.get_next_target_from_response(response) # Determine if any repos were deleted, and disable their tasks. keep_these = [k for k in injected_repos.keys()] self.disable_deleted_repo_tasks(index, next_index, keep_these) # termination condition if (next_index is None) or (next_index == index): logging.info('stopping after index %s, no next link found' % index) break else: index = next_index loop_count += 1 if loop_count == 20: logging.info('flushing updates') loop_count = 0 self.db_session.commit() self.db_session = self.mk_session() self.db_session.commit() self.db_session = self.mk_session() class SWHIndexingHttpLister(SWHListerHttpTransport, SWHIndexingLister): """Convenience class for ensuring right lookup and init order when combining SWHIndexingLister and SWHListerHttpTransport.""" - def __init__(self, lister_name=None, api_baseurl=None, - override_config=None): + def __init__(self, api_baseurl=None, override_config=None): SWHListerHttpTransport.__init__(self, api_baseurl=api_baseurl) - SWHIndexingLister.__init__(self, lister_name=lister_name, - override_config=override_config) + SWHIndexingLister.__init__(self, override_config=override_config) diff --git a/swh/lister/core/lister_base.py b/swh/lister/core/lister_base.py index ecda6b6..4fa2b28 100644 --- a/swh/lister/core/lister_base.py +++ b/swh/lister/core/lister_base.py @@ -1,496 +1,494 @@ # Copyright (C) 2015-2017 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import datetime import gzip import logging import os import re import time from sqlalchemy import create_engine, func from sqlalchemy.orm import sessionmaker from swh.core import config from swh.scheduler.backend import SchedulerBackend from swh.storage import get_storage from .abstractattribute import AbstractAttribute def utcnow(): return datetime.datetime.now(tz=datetime.timezone.utc) class FetchError(RuntimeError): def __init__(self, response): self.response = response def __str__(self): return repr(self.response) class SWHListerBase(abc.ABC, config.SWHConfig): """Lister core base class. Generally a source code hosting service provides an API endpoint for listing the set of stored repositories. A Lister is the discovery service responsible for finding this list, all at once or sequentially by parts, and queueing local tasks to fetch and ingest the referenced repositories. The core method in this class is ingest_data. Any subclasses should be calling this method one or more times to fetch and ingest data from API endpoints. See swh.lister.core.lister_base.SWHIndexingLister for example usage. This class cannot be instantiated. Any instantiable Lister descending from SWHListerBase must provide at least the required overrides. (see member docstrings for details): Required Overrides: MODEL def transport_request def transport_response_to_string def transport_response_simplified def transport_quota_check Optional Overrides: def filter_before_inject def is_within_bounds """ MODEL = AbstractAttribute('Subclass type (not instance)' ' of swh.lister.core.models.ModelBase' ' customized for a specific service.') + LISTER_NAME = AbstractAttribute("Lister's name") @abc.abstractmethod def transport_request(self, identifier): """Given a target endpoint identifier to query, try once to request it. Implementation of this method determines the network request protocol. Args: identifier (string): unique identifier for an endpoint query. e.g. If the service indexes lists of repositories by date and time of creation, this might be that as a formatted string. Or it might be an integer UID. Or it might be nothing. It depends on what the service needs. Returns: the entire request response Raises: Will catch internal transport-dependent connection exceptions and raise swh.lister.core.lister_base.FetchError instead. Other non-connection exceptions should propagate unchanged. """ pass @abc.abstractmethod def transport_response_to_string(self, response): """Convert the server response into a formatted string for logging. Implementation of this method depends on the shape of the network response object returned by the transport_request method. Args: response: the server response Returns: a pretty string of the response """ pass @abc.abstractmethod def transport_response_simplified(self, response): """Convert the server response into list of a dict for each repo in the response, mapping columns in the lister's MODEL class to repo data. Implementation of this method depends on the server API spec and the shape of the network response object returned by the transport_request method. Args: response: response object from the server. Returns: list of repo MODEL dicts ( eg. [{'uid': r['id'], etc.} for r in response.json()] ) """ pass @abc.abstractmethod def transport_quota_check(self, response): """Check server response to see if we're hitting request rate limits. Implementation of this method depends on the server communication protocol and API spec and the shape of the network response object returned by the transport_request method. Args: response (session response): complete API query response Returns: 1) must retry request? True/False 2) seconds to delay if True """ pass def filter_before_inject(self, models_list): """Function run after transport_response_simplified but before injection into the local db and creation of workers. Can be used to eliminate some of the results if necessary. MAY BE OVERRIDDEN if an intermediate Lister class needs to filter results before injection without requiring every child class to do so. Args: models_list: list of dicts returned by transport_response_simplified. Returns: models_list with entries changed according to custom logic. """ return models_list def is_within_bounds(self, inner, lower=None, upper=None): """See if a sortable value is inside the range [lower,upper]. MAY BE OVERRIDDEN, for example if the server indexable* key is technically sortable but not automatically so. * - ( see: swh.lister.core.indexing_lister.SWHIndexingLister ) Args: inner (sortable type): the value being checked lower (sortable type): optional lower bound upper (sortable type): optional upper bound Returns: whether inner is confined by the optional lower and upper bounds """ try: if lower is None and upper is None: return True elif lower is None: ret = inner <= upper elif upper is None: ret = inner >= lower else: ret = lower <= inner <= upper self.string_pattern_check(inner, lower, upper) except Exception as e: logging.error(str(e) + ': %s, %s, %s' % (('inner=%s%s' % (type(inner), inner)), ('lower=%s%s' % (type(lower), lower)), ('upper=%s%s' % (type(upper), upper))) ) raise return ret # You probably don't need to override anything below this line. DEFAULT_CONFIG = { 'storage': ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5002/' }, }), 'scheduling_db': ('str', 'dbname=softwareheritage-scheduler-dev'), } @property def CONFIG_BASE_FILENAME(self): # noqa: N802 - return 'lister-%s' % self.lister_name + return 'lister-%s' % self.LISTER_NAME @property def ADDITIONAL_CONFIG(self): # noqa: N802 return { 'lister_db_url': - ('str', 'postgresql:///lister-%s' % self.lister_name), + ('str', 'postgresql:///lister-%s' % self.LISTER_NAME), 'credentials': ('list[dict]', []), 'cache_responses': ('bool', False), 'cache_dir': - ('str', '~/.cache/swh/lister/%s' % self.lister_name), + ('str', '~/.cache/swh/lister/%s' % self.LISTER_NAME), } INITIAL_BACKOFF = 10 MAX_RETRIES = 7 CONN_SLEEP = 10 - def __init__(self, lister_name=None, override_config=None): + def __init__(self, override_config=None): self.backoff = self.INITIAL_BACKOFF - if lister_name is None: - raise NameError("Every lister must be assigned a lister_name.") - self.lister_name = lister_name # 'github?', 'bitbucket?', 'foo.com?' self.config = self.parse_config_file( base_filename=self.CONFIG_BASE_FILENAME, additional_configs=[self.ADDITIONAL_CONFIG] ) self.config['cache_dir'] = os.path.expanduser(self.config['cache_dir']) if self.config['cache_responses']: config.prepare_folders(self.config, 'cache_dir') if override_config: self.config.update(override_config) self.storage = get_storage(**self.config['storage']) self.scheduler = SchedulerBackend( scheduling_db=self.config['scheduling_db'], ) self.db_engine = create_engine(self.config['lister_db_url']) self.mk_session = sessionmaker(bind=self.db_engine) self.db_session = self.mk_session() def reset_backoff(self): """Reset exponential backoff timeout to initial level.""" self.backoff = self.INITIAL_BACKOFF def back_off(self): """Get next exponential backoff timeout.""" ret = self.backoff self.backoff *= 10 return ret def safely_issue_request(self, identifier): """Make network request with retries, rate quotas, and response logs. Protocol is handled by the implementation of the transport_request method. Args: identifier: resource identifier Returns: server response """ retries_left = self.MAX_RETRIES do_cache = self.config['cache_responses'] while retries_left > 0: try: r = self.transport_request(identifier) except FetchError: # network-level connection error, try again logging.warn('connection error on %s: sleep for %d seconds' % (identifier, self.CONN_SLEEP)) time.sleep(self.CONN_SLEEP) retries_left -= 1 continue if do_cache: self.save_response(r) # detect throttling must_retry, delay = self.transport_quota_check(r) if must_retry: logging.warn('rate limited on %s: sleep for %f seconds' % (identifier, delay)) time.sleep(delay) else: # request ok break retries_left -= 1 if not retries_left: logging.warn('giving up on %s: max retries exceeded' % identifier) return r def db_query_equal(self, key, value): """Look in the db for a row with key == value Args: key: column key to look at value: value to look for in that column Returns: sqlalchemy.ext.declarative.declarative_base object with the given key == value """ if isinstance(key, str): key = self.MODEL.__dict__[key] return self.db_session.query(self.MODEL) \ .filter(key == value).first() def winnow_models(self, mlist, key, to_remove): """Given a list of models, remove any with matching some member of a list of values. Args: mlist (list of model rows): the initial list of models key (column): the column to filter on to_remove (list): if anything in mlist has column equal to one of the values in to_remove, it will be removed from the result Returns: A list of model rows starting from mlist minus any matching rows """ if isinstance(key, str): key = self.MODEL.__dict__[key] if to_remove: return mlist.filter(~key.in_(to_remove)).all() else: return mlist.all() def db_num_entries(self): """Return the known number of entries in the lister db""" return self.db_session.query(func.count('*')).select_from(self.MODEL) \ .scalar() def db_inject_repo(self, model_dict): """Add/update a new repo to the db and mark it last_seen now. Args: model_dict: dictionary mapping model keys to values Returns: new or updated sqlalchemy.ext.declarative.declarative_base object associated with the injection """ sql_repo = self.db_query_equal('uid', model_dict['uid']) if not sql_repo: sql_repo = self.MODEL(**model_dict) self.db_session.add(sql_repo) else: for k in model_dict: setattr(sql_repo, k, model_dict[k]) sql_repo.last_seen = utcnow() return sql_repo def origin_dict(self, origin_type, origin_url): """Return special dict format for the origins list Args: origin_type (string) origin_url (string) Returns: the same information in a different form """ return { 'type': origin_type, 'url': origin_url, } def task_dict(self, origin_type, origin_url): """Return special dict format for the tasks list Args: origin_type (string) origin_url (string) Returns: the same information in a different form """ return { 'type': 'origin-update-%s' % origin_type, 'arguments': { 'args': [ origin_url, ], 'kwargs': {}, }, 'next_run': utcnow(), } def string_pattern_check(self, a, b, c=None): """When comparing indexable types in is_within_bounds, complex strings may not be allowed to differ in basic structure. If they do, it could be a sign of not understanding the data well. For instance, an ISO 8601 time string cannot be compared against its urlencoded equivalent, but this is an easy mistake to accidentally make. This method acts as a friendly sanity check. Args: a (string): inner component of the is_within_bounds method b (string): lower component of the is_within_bounds method c (string): upper component of the is_within_bounds method Returns: nothing Raises: TypeError if strings a, b, and c don't conform to the same basic pattern. """ if isinstance(a, str): a_pattern = re.sub('[a-zA-Z0-9]', '[a-zA-Z0-9]', re.escape(a)) if (isinstance(b, str) and (re.match(a_pattern, b) is None) or isinstance(c, str) and (re.match(a_pattern, c) is None)): logging.debug(a_pattern) raise TypeError('incomparable string patterns detected') def inject_repo_data_into_db(self, models_list): """Inject data into the db. Args: models_list: list of dicts mapping keys from the db model for each repo to be injected Returns: dict of uid:sql_repo pairs """ injected_repos = {} for m in models_list: injected_repos[m['uid']] = self.db_inject_repo(m) return injected_repos def create_missing_origins_and_tasks(self, models_list, injected_repos): """Find any newly created db entries that don't yet have tasks or origin objects assigned. Args: models_list: a list of dicts mapping keys in the db model for each repo injected_repos: dict of uid:sql_repo pairs that have just been created Returns: Nothing. Modifies injected_repos. """ for m in models_list: ir = injected_repos[m['uid']] if not ir.origin_id: ir.origin_id = self.storage.origin_add_one( self.origin_dict(m['origin_type'], m['origin_url']) ) if not ir.task_id: ir.task_id = self.scheduler.create_tasks( [self.task_dict(m['origin_type'], m['origin_url'])] )[0]['id'] def ingest_data(self, identifier): """The core data fetch sequence. Request server endpoint. Simplify and filter response list of repositories. Inject repo information into local db. Queue loader tasks for linked repositories. Args: identifier: Resource identifier. """ # Request (partial?) list of repositories info response = self.safely_issue_request(identifier) models_list = self.transport_response_simplified(response) models_list = self.filter_before_inject(models_list) # inject into local db injected = self.inject_repo_data_into_db(models_list) # queue workers self.create_missing_origins_and_tasks(models_list, injected) return response, injected def save_response(self, response): """Log the response from a server request to a cache dir. Args: response: full server response cache_dir: system path for cache dir Returns: nothing """ datepath = utcnow().isoformat() fname = os.path.join( self.config['cache_dir'], datepath + '.gz', ) with gzip.open(fname, 'w') as f: f.write(bytes( self.transport_response_to_string(response), 'UTF-8' )) diff --git a/swh/lister/core/paging_lister.py b/swh/lister/core/paging_lister.py index 34457b2..c4044f0 100644 --- a/swh/lister/core/paging_lister.py +++ b/swh/lister/core/paging_lister.py @@ -1,138 +1,136 @@ # Copyright (C) 2015-2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import logging from .lister_transports import SWHListerHttpTransport from .lister_base import SWHListerBase class SWHPagingLister(SWHListerBase): """Lister* intermediate class for any service that follows the simple pagination page pattern. - Client sends a request to list repositories starting from a given page identifier. - Client receives structured (json/xml/etc) response with information about a sequential series of repositories (per page) starting from a given index. And, if available, some indication of the next page index for fetching the remaining repository data. See :class:`swh.lister.core.lister_base.SWHListerBase` for more details. This class cannot be instantiated. To create a new Lister for a source code listing service that follows the model described above, you must subclass this class. Then provide the required overrides in addition to any unmet implementation/override requirements of this class's base (see parent class and member docstrings for details). Required Overrides:: def get_next_target_from_response """ @abc.abstractmethod def get_next_target_from_response(self, response): """Find the next server endpoint page given the entire response. Implementation of this method depends on the server API spec and the shape of the network response object returned by the transport_request method. For example, some api can use the headers links to provide the next page. Args: response (transport response): response page from the server Returns: index of next page, possibly extracted from a next href url """ pass @abc.abstractmethod def get_pages_information(self): """Find the total number of pages. Implementation of this method depends on the server API spec and the shape of the network response object returned by the transport_request method. For example, some api can use dedicated headers: - x-total-pages to provide the total number of pages - x-total to provide the total number of repositories - x-per-page to provide the number of elements per page Returns: tuple (total number of repositories, total number of pages, per_page) """ pass # You probably don't need to override anything below this line. def run(self, min_index=None, max_index=None): """Main entry function. Sequentially fetches repository data from the service according to the basic outline in the class docstring. Continually fetching sublists until either there is no next index reference given or the given next index is greater than the desired max_index. Args: min_index (indexable type): optional index to start from max_index (indexable type): optional index to stop at Returns: nothing """ index = min_index or '' loop_count = 0 self.min_index = min_index self.max_index = max_index while self.is_within_bounds(index, self.min_index, self.max_index): logging.info('listing repos starting at %s' % index) response, injected_repos = self.ingest_data(index) next_index = self.get_next_target_from_response(response) # termination condition if (next_index is None) or (next_index == index): logging.info('stopping after index %s, no next link found' % index) break else: index = next_index loop_count += 1 if loop_count == 20: logging.info('flushing updates') loop_count = 0 self.db_session.commit() self.db_session = self.mk_session() self.db_session.commit() self.db_session = self.mk_session() class SWHPagingHttpLister(SWHListerHttpTransport, SWHPagingLister): """Convenience class for ensuring right lookup and init order when combining SWHPagingLister and SWHListerHttpTransport. """ - def __init__(self, lister_name=None, api_baseurl=None, - override_config=None): + def __init__(self, api_baseurl=None, override_config=None): SWHListerHttpTransport.__init__(self, api_baseurl=api_baseurl) - SWHPagingLister.__init__(self, lister_name=lister_name, - override_config=override_config) + SWHPagingLister.__init__(self, override_config=override_config) diff --git a/swh/lister/core/tests/test_lister.py b/swh/lister/core/tests/test_lister.py index ebe9d40..dd9a750 100644 --- a/swh/lister/core/tests/test_lister.py +++ b/swh/lister/core/tests/test_lister.py @@ -1,241 +1,241 @@ # Copyright (C) 2017-2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import time from unittest import TestCase from unittest.mock import Mock, patch import requests_mock from testing.postgresql import Postgresql from nose.tools import istest from sqlalchemy import create_engine from swh.lister.core.abstractattribute import AbstractAttribute def noop(*args, **kwargs): pass @requests_mock.Mocker() class HttpListerTesterBase(abc.ABC): """Base testing class for subclasses of swh.lister.core.indexing_lister.SWHIndexingHttpLister. swh.lister.core.paging_lister.SWHPagingHttpLister See swh.lister.github.tests.test_gh_lister for an example of how to customize for a specific listing service. """ Lister = AbstractAttribute('The lister class to test') test_re = AbstractAttribute('Compiled regex matching the server url. Must' ' capture the index value.') lister_subdir = AbstractAttribute('bitbucket, github, etc.') good_api_response_file = AbstractAttribute('Example good response body') bad_api_response_file = AbstractAttribute('Example bad response body') first_index = AbstractAttribute('First index in good_api_response') last_index = AbstractAttribute('Last index in good_api_response') entries_per_page = AbstractAttribute('Number of results in good response') + LISTER_NAME = 'fake-lister' # May need to override this if the headers are used for something def response_headers(self, request): return {} # May need to override this if the server uses non-standard rate limiting # method. # Please keep the requested retry delay reasonably low. def mock_rate_quota(self, n, request, context): self.rate_limit += 1 context.status_code = 429 context.headers['Retry-After'] = '1' return '{"error":"dummy"}' def __init__(self, *args, **kwargs): super().__init__(*args, **kwargs) self.rate_limit = 1 self.response = None self.fl = None self.helper = None if self.__class__ != HttpListerTesterBase: self.run = TestCase.run.__get__(self, self.__class__) else: self.run = noop def request_index(self, request): m = self.test_re.search(request.path_url) if m and (len(m.groups()) > 0): return m.group(1) else: return None def mock_response(self, request, context): self.fl.reset_backoff() self.rate_limit = 1 context.status_code = 200 custom_headers = self.response_headers(request) context.headers.update(custom_headers) if self.request_index(request) == str(self.first_index): with open('swh/lister/%s/tests/%s' % (self.lister_subdir, self.good_api_response_file), 'r', encoding='utf-8') as r: return r.read() else: with open('swh/lister/%s/tests/%s' % (self.lister_subdir, self.bad_api_response_file), 'r', encoding='utf-8') as r: return r.read() def mock_limit_n_response(self, n, request, context): self.fl.reset_backoff() if self.rate_limit <= n: return self.mock_rate_quota(n, request, context) else: return self.mock_response(request, context) def mock_limit_once_response(self, request, context): return self.mock_limit_n_response(1, request, context) def mock_limit_twice_response(self, request, context): return self.mock_limit_n_response(2, request, context) def get_fl(self, override_config=None): """Retrieve an instance of fake lister (fl). """ if override_config or self.fl is None: with patch( 'swh.scheduler.backend.SchedulerBackend.reconnect', noop ): - self.fl = self.Lister(lister_name='fakelister', - api_baseurl='https://fakeurl', + self.fl = self.Lister(api_baseurl='https://fakeurl', override_config=override_config) self.fl.INITIAL_BACKOFF = 1 self.fl.reset_backoff() return self.fl def get_api_response(self): fl = self.get_fl() if self.response is None: self.response = fl.safely_issue_request(self.first_index) return self.response @istest def test_is_within_bounds(self, http_mocker): fl = self.get_fl() self.assertFalse(fl.is_within_bounds(1, 2, 3)) self.assertTrue(fl.is_within_bounds(2, 1, 3)) self.assertTrue(fl.is_within_bounds(1, 1, 1)) self.assertTrue(fl.is_within_bounds(1, None, None)) self.assertTrue(fl.is_within_bounds(1, None, 2)) self.assertTrue(fl.is_within_bounds(1, 0, None)) self.assertTrue(fl.is_within_bounds("b", "a", "c")) self.assertFalse(fl.is_within_bounds("a", "b", "c")) self.assertTrue(fl.is_within_bounds("a", None, "c")) self.assertTrue(fl.is_within_bounds("a", None, None)) self.assertTrue(fl.is_within_bounds("b", "a", None)) self.assertFalse(fl.is_within_bounds("a", "b", None)) self.assertTrue(fl.is_within_bounds("aa:02", "aa:01", "aa:03")) self.assertFalse(fl.is_within_bounds("aa:12", None, "aa:03")) with self.assertRaises(TypeError): fl.is_within_bounds(1.0, "b", None) with self.assertRaises(TypeError): fl.is_within_bounds("A:B", "A::B", None) @istest def test_api_request(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_limit_twice_response) with patch.object(time, 'sleep', wraps=time.sleep) as sleepmock: self.get_api_response() self.assertEqual(sleepmock.call_count, 2) @istest def test_repos_list(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_response) li = self.get_fl().transport_response_simplified( self.get_api_response() ) self.assertIsInstance(li, list) self.assertEqual(len(li), self.entries_per_page) @istest def test_model_map(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_response) fl = self.get_fl() li = fl.transport_response_simplified(self.get_api_response()) di = li[0] self.assertIsInstance(di, dict) pubs = [k for k in vars(fl.MODEL).keys() if not k.startswith('_')] for k in pubs: if k not in ['last_seen', 'task_id', 'origin_id', 'id']: self.assertIn(k, di) def disable_storage_and_scheduler(self, fl): fl.create_missing_origins_and_tasks = Mock(return_value=None) def disable_db(self, fl): fl.winnow_models = Mock(return_value=[]) fl.db_inject_repo = Mock(return_value=fl.MODEL()) fl.disable_deleted_repo_tasks = Mock(return_value=None) @istest def test_fetch_none_nodb(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_response) fl = self.get_fl() self.disable_storage_and_scheduler(fl) self.disable_db(fl) fl.run(min_index=1, max_index=1) # stores no results @istest def test_fetch_one_nodb(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_response) fl = self.get_fl() self.disable_storage_and_scheduler(fl) self.disable_db(fl) fl.run(min_index=self.first_index, max_index=self.first_index) @istest def test_fetch_multiple_pages_nodb(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_response) fl = self.get_fl() self.disable_storage_and_scheduler(fl) self.disable_db(fl) fl.run(min_index=self.first_index) def init_db(self, db, model): engine = create_engine(db.url()) model.metadata.create_all(engine) @istest def test_fetch_multiple_pages_yesdb(self, http_mocker): http_mocker.get(self.test_re, text=self.mock_response) initdb_args = Postgresql.DEFAULT_SETTINGS['initdb_args'] initdb_args = ' '.join([initdb_args, '-E UTF-8']) db = Postgresql(initdb_args=initdb_args) fl = self.get_fl(override_config={'lister_db_url': db.url()}) self.init_db(db, fl.MODEL) self.disable_storage_and_scheduler(fl) # FIXME: Separate the tests properly for the gitlab lister # did not succeed yet if not hasattr(fl, 'db_last_index'): # gitlab lister cannot pass here return fl.run(min_index=self.first_index) self.assertEqual(fl.db_last_index(), self.last_index) partitions = fl.db_partition_indices(5) self.assertGreater(len(partitions), 0) for k in partitions: self.assertLessEqual(len(k), 5) self.assertGreater(len(k), 0) diff --git a/swh/lister/debian/lister.py b/swh/lister/debian/lister.py index 913971b..d23e693 100644 --- a/swh/lister/debian/lister.py +++ b/swh/lister/debian/lister.py @@ -1,236 +1,236 @@ # Copyright (C) 2017 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import bz2 from collections import defaultdict import datetime import gzip import lzma import logging from debian.deb822 import Sources from sqlalchemy.orm import joinedload, load_only from sqlalchemy.schema import CreateTable, DropTable from swh.storage.schemata.distribution import ( AreaSnapshot, Distribution, DistributionSnapshot, Package, TempPackage, ) from swh.lister.core.lister_base import SWHListerBase, FetchError from swh.lister.core.lister_transports import SWHListerHttpTransport decompressors = { 'gz': lambda f: gzip.GzipFile(fileobj=f), 'bz2': bz2.BZ2File, 'xz': lzma.LZMAFile, } class DebianLister(SWHListerHttpTransport, SWHListerBase): MODEL = Package PATH_TEMPLATE = None + LISTER_NAME = 'debian' def __init__(self, override_config=None): SWHListerHttpTransport.__init__(self, api_baseurl="bogus") - SWHListerBase.__init__(self, lister_name='debian', - override_config=override_config) + SWHListerBase.__init__(self, override_config=override_config) def transport_request(self, identifier): """Subvert SWHListerHttpTransport.transport_request, to try several index URIs in turn. The Debian repository format supports several compression algorithms across the ages, so we try several URIs. Once we have found a working URI, we break and set `self.decompressor` to the one that matched. Returns: a requests Response object. Raises: FetchError: when all the URIs failed to be retrieved. """ response = None compression = None for uri, compression in self.area.index_uris(): response = super().transport_request(uri) if response.status_code == 200: break else: raise FetchError( "Could not retrieve index for %s" % self.area ) self.decompressor = decompressors.get(compression) return response def request_uri(self, identifier): # In the overridden transport_request, we pass # SWHListerBase.transport_request() the full URI as identifier, so we # need to return it here. return identifier def request_params(self, identifier): # Enable streaming to allow wrapping the response in the decompressor # in transport_response_simplified. params = super().request_params(identifier) params['stream'] = True return params def transport_response_simplified(self, response): """Decompress and parse the package index fetched in `transport_request`. For each package, we "pivot" the file list entries (Files, Checksums-Sha1, Checksums-Sha256), to return a files dict mapping filenames to their checksums. """ if self.decompressor: data = self.decompressor(response.raw) else: data = response.raw for src_pkg in Sources.iter_paragraphs(data.readlines()): files = defaultdict(dict) for field in src_pkg._multivalued_fields: if field.startswith('checksums-'): sum_name = field[len('checksums-'):] else: sum_name = 'md5sum' if field in src_pkg: for entry in src_pkg[field]: name = entry['name'] files[name]['name'] = entry['name'] files[name]['size'] = int(entry['size'], 10) files[name][sum_name] = entry[sum_name] yield { 'name': src_pkg['Package'], 'version': src_pkg['Version'], 'directory': src_pkg['Directory'], 'files': files, } def inject_repo_data_into_db(self, models_list): """Generate the Package entries that didn't previously exist. Contrary to SWHListerBase, we don't actually insert the data in database. `create_missing_origins_and_tasks` does it once we have the origin and task identifiers. """ by_name_version = {} temp_packages = [] area_id = self.area.id for model in models_list: name = model['name'] version = model['version'] temp_packages.append({ 'area_id': area_id, 'name': name, 'version': version, }) by_name_version[name, version] = model # Add all the listed packages to a temporary table self.db_session.execute(CreateTable(TempPackage.__table__)) self.db_session.bulk_insert_mappings(TempPackage, temp_packages) def exists_tmp_pkg(db_session, model): return ( db_session.query(model) .filter(Package.area_id == TempPackage.area_id) .filter(Package.name == TempPackage.name) .filter(Package.version == TempPackage.version) .exists() ) # Filter out the packages that already exist in the main Package table new_packages = self.db_session\ .query(TempPackage)\ .options(load_only('name', 'version'))\ .filter(~exists_tmp_pkg(self.db_session, Package))\ .all() self.old_area_packages = self.db_session.query(Package).filter( exists_tmp_pkg(self.db_session, TempPackage) ).all() self.db_session.execute(DropTable(TempPackage.__table__)) added_packages = [] for package in new_packages: model = by_name_version[package.name, package.version] added_packages.append(Package(area=self.area, **model)) self.db_session.add_all(added_packages) return added_packages def create_missing_origins_and_tasks(self, models_list, added_packages): """We create tasks at the end of the full snapshot processing""" return def create_tasks_for_snapshot(self, snapshot): tasks = [ snapshot.task_for_package(name, versions) for name, versions in snapshot.get_packages().items() ] return self.scheduler.create_tasks(tasks) def run(self, distribution, date=None): """Run the lister for a given (distribution, area) tuple. Args: distribution (str): name of the distribution (e.g. "Debian") date (datetime.datetime): date the snapshot is taken (defaults to now) """ distribution = self.db_session\ .query(Distribution)\ .options(joinedload(Distribution.areas))\ .filter(Distribution.name == distribution)\ .one_or_none() if not distribution: raise ValueError("Distribution %s is not registered" % distribution) if not distribution.type == 'deb': raise ValueError("Distribution %s is not a Debian derivative" % distribution) date = date or datetime.datetime.now(tz=datetime.timezone.utc) logging.debug('Creating snapshot for distribution %s on date %s' % (distribution, date)) snapshot = DistributionSnapshot(date=date, distribution=distribution) self.db_session.add(snapshot) for area in distribution.areas: if not area.active: continue self.area = area logging.debug('Processing area %s' % area) _, new_area_packages = self.ingest_data(None) area_snapshot = AreaSnapshot(snapshot=snapshot, area=area) self.db_session.add(area_snapshot) area_snapshot.packages.extend(new_area_packages) area_snapshot.packages.extend(self.old_area_packages) self.create_tasks_for_snapshot(snapshot) self.db_session.commit() return True diff --git a/swh/lister/github/lister.py b/swh/lister/github/lister.py index 30d4290..f841f60 100644 --- a/swh/lister/github/lister.py +++ b/swh/lister/github/lister.py @@ -1,51 +1,52 @@ # Copyright (C) 2017-2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import re import time from swh.lister.core.indexing_lister import SWHIndexingHttpLister from swh.lister.github.models import GitHubModel class GitHubLister(SWHIndexingHttpLister): PATH_TEMPLATE = '/repositories?since=%d' MODEL = GitHubModel API_URL_INDEX_RE = re.compile(r'^.*/repositories\?since=(\d+)') + LISTER_NAME = 'github.com' def get_model_from_repo(self, repo): return { 'uid': repo['id'], 'indexable': repo['id'], 'name': repo['name'], 'full_name': repo['full_name'], 'html_url': repo['html_url'], 'origin_url': repo['html_url'], 'origin_type': 'git', 'description': repo['description'], 'fork': repo['fork'], } def transport_quota_check(self, response): reqs_remaining = int(response.headers['X-RateLimit-Remaining']) if response.status_code == 403 and reqs_remaining == 0: reset_at = int(response.headers['X-RateLimit-Reset']) delay = min(reset_at - time.time(), 3600) return True, delay else: return False, 0 def get_next_target_from_response(self, response): if 'next' in response.links: next_url = response.links['next']['url'] return int(self.API_URL_INDEX_RE.match(next_url).group(1)) else: return None def transport_response_simplified(self, response): repos = response.json() return [self.get_model_from_repo(repo) for repo in repos] def request_headers(self): return {'Accept': 'application/vnd.github.v3+json'} diff --git a/swh/lister/github/tasks.py b/swh/lister/github/tasks.py index f4c5b36..ba04e8c 100644 --- a/swh/lister/github/tasks.py +++ b/swh/lister/github/tasks.py @@ -1,27 +1,26 @@ # Copyright (C) 2017-2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from swh.lister.core.tasks import (IndexingDiscoveryListerTask, RangeListerTask, IndexingRefreshListerTask, ListerTaskBase) from .lister import GitHubLister class GitHubListerTask(ListerTaskBase): def new_lister(self): - return GitHubLister(lister_name='github.com', - api_baseurl='https://api.github.com') + return GitHubLister(api_baseurl='https://api.github.com') class IncrementalGitHubLister(GitHubListerTask, IndexingDiscoveryListerTask): task_queue = 'swh_lister_github_discover' class RangeGitHubLister(GitHubListerTask, RangeListerTask): task_queue = 'swh_lister_github_refresh' class FullGitHubRelister(GitHubListerTask, IndexingRefreshListerTask): task_queue = 'swh_lister_github_refresh' diff --git a/swh/lister/gitlab/lister.py b/swh/lister/gitlab/lister.py index 78f7736..1cbfc7a 100644 --- a/swh/lister/gitlab/lister.py +++ b/swh/lister/gitlab/lister.py @@ -1,130 +1,124 @@ # Copyright (C) 2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import random import re import time from ..core.paging_lister import SWHPagingHttpLister from .models import GitLabModel class GitLabLister(SWHPagingHttpLister): # Template path expecting an integer that represents the page id PATH_TEMPLATE = '/projects?page=%d&order_by=id&sort=asc&simple=true' API_URL_INDEX_RE = re.compile(r'^.*/projects.*page=(\d+).*') MODEL = GitLabModel + LISTER_NAME = 'gitlab' - @property - def CONFIG_BASE_FILENAME(self): - """One gitlab lister for all instances. We discriminate between the - origin on a per instance basis in the table. - - """ - return 'lister-gitlab' + def __init__(self, api_baseurl=None, instance=None, + override_config=None, sort='asc'): + super().__init__(api_baseurl=api_baseurl, + override_config=override_config) + self.instance = instance + self.PATH_TEMPLATE = '%s&sort=%s' % (self.PATH_TEMPLATE, sort) @property def ADDITIONAL_CONFIG(self): """Override additional config as the 'credentials' structure change between the ancestor classes and this class. cf. request_params method below """ - return { - 'lister_db_url': - ('str', 'postgresql:///lister-gitlab'), - 'credentials': # credentials is a dict - ('dict', {}), - 'cache_responses': - ('bool', False), - 'cache_dir': - ('str', '~/.cache/swh/lister/%s' % self.lister_name), - } + default_config = super().ADDITIONAL_CONFIG + # 'credentials' is a dict of (instance, {username, password}) dict + default_config['credentials'] = ('dict', {}) + return default_config def request_params(self, identifier): """Get the full parameters passed to requests given the transport_request identifier. For the gitlab lister, the 'credentials' entries is configured per instance. For example: - credentials: - gitlab.com: - username: user0 password: - username: user1 password: - ... - other-gitlab-instance: ... """ params = { 'headers': self.request_headers() or {} } # Retrieve the credentials per instance creds = self.config['credentials'] if creds: - creds_lister = creds[self.lister_name] + creds_lister = creds[self.instance] auth = random.choice(creds_lister) if creds else None if auth: params['auth'] = (auth['username'], auth['password']) return params def get_model_from_repo(self, repo): return { - 'instance': self.lister_name, + 'instance': self.instance, 'uid': repo['id'], 'indexable': repo['id'], 'name': repo['name'], 'full_name': repo['path_with_namespace'], 'html_url': repo['web_url'], 'origin_url': repo['http_url_to_repo'], 'origin_type': 'git', 'description': repo['description'], } def transport_quota_check(self, response): """Deal with rate limit if any. """ # not all gitlab instance have rate limit if 'RateLimit-Remaining' in response.headers: reqs_remaining = int(response.headers['RateLimit-Remaining']) if response.status_code == 403 and reqs_remaining == 0: reset_at = int(response.headers['RateLimit-Reset']) delay = min(reset_at - time.time(), 3600) return True, delay return False, 0 def get_next_target_from_response(self, response): """Determine the next page identifier. """ if 'next' in response.links: next_url = response.links['next']['url'] return int(self.API_URL_INDEX_RE.match(next_url).group(1)) return None def get_pages_information(self): """Determine some pages information. """ response = self.transport_head(identifier=1) h = response.headers total = h.get('x-total', h.get('X-Total')) total_pages = h.get('x-total-pages', h.get('X-Total-Pages')) per_page = h.get('x-per-page', h.get('X-Per-Page')) if total is not None: total = int(total) if total_pages is not None: total_pages = int(total_pages) if per_page is not None: per_page = int(per_page) return total, total_pages, per_page def transport_response_simplified(self, response): repos = response.json() return [self.get_model_from_repo(repo) for repo in repos] diff --git a/swh/lister/gitlab/tasks.py b/swh/lister/gitlab/tasks.py index 63bc179..ba6290e 100644 --- a/swh/lister/gitlab/tasks.py +++ b/swh/lister/gitlab/tasks.py @@ -1,44 +1,43 @@ # Copyright (C) 2018 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import random from celery import group from ..core.tasks import ListerTaskBase, RangeListerTask from .lister import GitLabLister class GitLabListerTask(ListerTaskBase): - def new_lister(self, lister_name='gitlab.com', - api_baseurl='https://gitlab.com/api/v4'): - return GitLabLister( - lister_name=lister_name, api_baseurl=api_baseurl) + def new_lister(self, api_baseurl='https://gitlab.com/api/v4', + instance='gitlab.com'): + return GitLabLister(api_baseurl=api_baseurl, instance=instance) class RangeGitLabLister(GitLabListerTask, RangeListerTask): """GitLab lister working on specified range (start, end) arguments. """ task_queue = 'swh_lister_gitlab_refresh' class FullGitLabRelister(GitLabListerTask): task_queue = 'swh_lister_gitlab_refresh' def run_task(self, *args, **kwargs): lister = self.new_lister(*args, **kwargs) total, _, per_page = lister.get_pages_information() ranges = [] prev_index = None for index in range(0, total, per_page): if index is not None and prev_index is not None: ranges.append((prev_index, index)) prev_index = index random.shuffle(ranges) range_task = RangeGitLabLister() group(range_task.s(minv, maxv, *args, **kwargs) for minv, maxv in ranges)()