diff --git a/swh/lister/bitbucket/lister.py b/swh/lister/bitbucket/lister.py index ae5866c..04a20aa 100644 --- a/swh/lister/bitbucket/lister.py +++ b/swh/lister/bitbucket/lister.py @@ -1,71 +1,72 @@ -# Copyright (C) 2017-2018 the Software Heritage developers +# Copyright (C) 2017-2019 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 import logging import iso8601 from swh.lister.bitbucket.models import BitBucketModel from swh.lister.core.indexing_lister import SWHIndexingHttpLister logger = logging.getLogger(__name__) class BitBucketLister(SWHIndexingHttpLister): PATH_TEMPLATE = '/repositories?after=%s' MODEL = BitBucketModel LISTER_NAME = 'bitbucket' + instance = 'bitbucket' 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] def request_uri(self, identifier): return super().request_uri(identifier or '1970-01-01') def is_within_bounds(self, inner, lower=None, upper=None): # values are expected to be str dates try: inner = iso8601.parse_date(inner) if lower: lower = iso8601.parse_date(lower) if upper: upper = iso8601.parse_date(upper) 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 except Exception as e: logger.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 diff --git a/swh/lister/core/lister_base.py b/swh/lister/core/lister_base.py index b215810..a742453 100644 --- a/swh/lister/core/lister_base.py +++ b/swh/lister/core/lister_base.py @@ -1,552 +1,549 @@ # 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 datetime import gzip import json 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 import get_scheduler, utils from swh.storage import get_storage from .abstractattribute import AbstractAttribute logger = logging.getLogger(__name__) 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 do_additional_checks(self, models_list): """Execute some additional checks on the model list. For example, to check for existing repositories in the db. MAY BE OVERRIDDEN if an intermediate Lister class needs to check some more the results before injection. Checks are fine by default, returns the models_list as is by default. Args: models_list: list of dicts returned by transport_response_simplified. Returns: models_list with entries if checks ok, False otherwise """ 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: logger.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/' }, }), 'scheduler': ('dict', { 'cls': 'remote', 'args': { 'url': 'http://localhost:5008/' }, }), 'lister': ('dict', { 'cls': 'local', 'args': { 'db': 'postgresql:///lister', }, }), } @property def CONFIG_BASE_FILENAME(self): # noqa: N802 return 'lister_%s' % self.LISTER_NAME @property def ADDITIONAL_CONFIG(self): # noqa: N802 return { - 'credentials': - ('list[dict]', []), - 'cache_responses': - ('bool', False), - 'cache_dir': - ('str', '~/.cache/swh/lister/%s' % self.LISTER_NAME), + 'credentials': ('dict', {}), + 'cache_responses': ('bool', False), + 'cache_dir': ('str', '~/.cache/swh/lister/%s' % self.LISTER_NAME), } INITIAL_BACKOFF = 10 MAX_RETRIES = 7 CONN_SLEEP = 10 def __init__(self, override_config=None): self.backoff = self.INITIAL_BACKOFF logger.debug('Loading config from %s' % self.CONFIG_BASE_FILENAME) 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) logger.debug('%s CONFIG=%s' % (self, self.config)) self.storage = get_storage(**self.config['storage']) self.scheduler = get_scheduler(**self.config['scheduler']) self.db_engine = create_engine(self.config['lister']['args']['db']) 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'] r = None while retries_left > 0: try: r = self.transport_request(identifier) except FetchError: # network-level connection error, try again logger.warning( '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: logger.warning( 'rate limited on %s: sleep for %f seconds' % (identifier, delay)) time.sleep(delay) else: # request ok break retries_left -= 1 if not retries_left: logger.warning( '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, **kwargs): """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, **kwargs): """Return special dict format for the tasks list Args: origin_type (string) origin_url (string) Returns: the same information in a different form """ _type = 'load-%s' % origin_type _policy = 'recurring' return utils.create_task_dict(_type, _policy, origin_url) 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)): logger.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. """ origins = {} tasks = {} def _origin_key(m): _type = m.get('origin_type', m.get('type')) _url = m.get('origin_url', m.get('url')) return '%s-%s' % (_type, _url) def _task_key(m): return '%s-%s' % (m['type'], json.dumps(m['arguments'], sort_keys=True)) for m in models_list: ir = injected_repos[m['uid']] if not ir.origin_id: origin_dict = self.origin_dict(**m) origins[_origin_key(m)] = (ir, m, origin_dict) if not ir.task_id: task_dict = self.task_dict(**m) tasks[_task_key(task_dict)] = (ir, m, task_dict) new_origins = self.storage.origin_add( (origin_dicts for (_, _, origin_dicts) in origins.values())) for origin in new_origins: ir, m, _ = origins[_origin_key(origin)] ir.origin_id = origin['id'] new_tasks = self.scheduler.create_tasks( (task_dicts for (_, _, task_dicts) in tasks.values())) for task in new_tasks: ir, m, _ = tasks[_task_key(task)] ir.task_id = task['id'] def ingest_data(self, identifier, checks=False): """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. checks (bool): Additional checks required """ # Request (partial?) list of repositories info response = self.safely_issue_request(identifier) if not response: return response, [] models_list = self.transport_response_simplified(response) models_list = self.filter_before_inject(models_list) if checks: models_list = self.do_additional_checks(models_list) if not models_list: return response, [] # 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/lister_transports.py b/swh/lister/core/lister_transports.py index 20f841f..75bc6ad 100644 --- a/swh/lister/core/lister_transports.py +++ b/swh/lister/core/lister_transports.py @@ -1,180 +1,207 @@ # 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 random from datetime import datetime from email.utils import parsedate from pprint import pformat import logging import requests import xmltodict try: from swh.lister._version import __version__ except ImportError: __version__ = 'devel' from .abstractattribute import AbstractAttribute from .lister_base import FetchError logger = logging.getLogger(__name__) class SWHListerHttpTransport(abc.ABC): """Use the Requests library for making Lister endpoint requests. To be used in conjunction with SWHListerBase or a subclass of it. """ PATH_TEMPLATE = AbstractAttribute('string containing a python string' ' format pattern that produces the API' ' endpoint path for listing stored' ' repositories when given an index.' ' eg. "/repositories?after=%s".' 'To be implemented in the API-specific' ' class inheriting this.') EXPECTED_STATUS_CODES = (200, 429, 403, 404) def request_headers(self): """Returns dictionary of any request headers needed by the server. MAY BE OVERRIDDEN if request headers are needed. """ return { 'User-Agent': 'Software Heritage lister (%s)' % self.lister_version } def request_uri(self, identifier): """Get the full request URI given the transport_request identifier. MAY BE OVERRIDDEN if something more complex than the PATH_TEMPLATE is required. """ path = self.PATH_TEMPLATE % identifier return self.api_baseurl + path def request_params(self, identifier): """Get the full parameters passed to requests given the transport_request identifier. + This uses credentials if any are provided. The 'credentials' + configuration is expected to be a dict of multiple levels. The first + level is the lister's name, the second is the lister's instance name. + + For example: + + credentials: + github: # github lister + github: # has only one instance (so far) + - username: some + password: somekey + - username: one + password: onekey + - ... + gitlab: # gitlab lister + riseup: # has many instances + - username: someone + password: ... + - ... + gitlab: + - username: someone + password: ... + - ... + + MAY BE OVERRIDDEN if something more complex than the request headers is needed. """ params = {} params['headers'] = self.request_headers() or {} - creds = self.config['credentials'] + all_creds = self.config['credentials'] + lister_creds = all_creds.get(self.LISTER_NAME, {}) + creds = lister_creds.get(self.instance, {}) auth = random.choice(creds) if creds else None if auth: params['auth'] = (auth['username'], auth['password']) return params def transport_quota_check(self, response): """Implements SWHListerBase.transport_quota_check with standard 429 code check for HTTP with Requests library. MAY BE OVERRIDDEN if the server notifies about rate limits in a non-standard way that doesn't use HTTP 429 and the Retry-After response header. ( https://tools.ietf.org/html/rfc6585#section-4 ) """ if response.status_code == 429: # HTTP too many requests retry_after = response.headers.get('Retry-After', self.back_off()) try: # might be seconds return True, float(retry_after) except Exception: # might be http-date at_date = datetime(*parsedate(retry_after)[:6]) from_now = (at_date - datetime.today()).total_seconds() + 5 return True, max(0, from_now) else: # response ok self.reset_backoff() return False, 0 def __init__(self, api_baseurl=None): if not api_baseurl: raise NameError('HTTP Lister Transport requires api_baseurl.') self.api_baseurl = api_baseurl # eg. 'https://api.github.com' self.session = requests.Session() self.lister_version = __version__ def _transport_action(self, identifier, method='get'): """Permit to ask information to the api prior to actually executing query. """ path = self.request_uri(identifier) params = self.request_params(identifier) try: if method == 'head': response = self.session.head(path, **params) else: response = self.session.get(path, **params) except requests.exceptions.ConnectionError as e: logger.warning('Failed to fetch %s: %s', path, e) raise FetchError(e) else: if response.status_code not in self.EXPECTED_STATUS_CODES: raise FetchError(response) return response def transport_head(self, identifier): """Retrieve head information on api. """ return self._transport_action(identifier, method='head') def transport_request(self, identifier): """Implements SWHListerBase.transport_request for HTTP using Requests. Retrieve get information on api. """ return self._transport_action(identifier) def transport_response_to_string(self, response): """Implements SWHListerBase.transport_response_to_string for HTTP given Requests responses. """ s = pformat(response.request.path_url) s += '\n#\n' + pformat(response.request.headers) s += '\n#\n' + pformat(response.status_code) s += '\n#\n' + pformat(response.headers) s += '\n#\n' try: # json? s += pformat(response.json()) except Exception: # not json try: # xml? s += pformat(xmltodict.parse(response.text)) except Exception: # not xml s += pformat(response.text) return s class ListerOnePageApiTransport(SWHListerHttpTransport): """Leverage requests library to retrieve basic html page and parse result. To be used in conjunction with SWHListerBase or a subclass of it. """ PAGE = AbstractAttribute("The server api's unique page to retrieve and " "parse for information") PATH_TEMPLATE = None # we do not use it def __init__(self, api_baseurl=None): self.session = requests.Session() self.lister_version = __version__ def request_uri(self, _): """Get the full request URI given the transport_request identifier. """ return self.PAGE diff --git a/swh/lister/debian/lister.py b/swh/lister/debian/lister.py index d23e693..5b72c6f 100644 --- a/swh/lister/debian/lister.py +++ b/swh/lister/debian/lister.py @@ -1,236 +1,237 @@ # 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' + instance = 'debian' def __init__(self, override_config=None): SWHListerHttpTransport.__init__(self, api_baseurl="bogus") 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 c0e49c4..245afde 100644 --- a/swh/lister/github/lister.py +++ b/swh/lister/github/lister.py @@ -1,52 +1,53 @@ -# Copyright (C) 2017-2018 the Software Heritage developers +# Copyright (C) 2017-2019 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' + instance = 'github' # There is only 1 instance of such lister 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/gitlab/lister.py b/swh/lister/gitlab/lister.py index 8e09e23..41cb7c6 100644 --- a/swh/lister/gitlab/lister.py +++ b/swh/lister/gitlab/lister.py @@ -1,126 +1,113 @@ -# Copyright (C) 2018 the Software Heritage developers +# Copyright (C) 2018-2019 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 time from urllib3.util import parse_url from ..core.page_by_page_lister import PageByPageHttpLister from .models import GitLabModel class GitLabLister(PageByPageHttpLister): # Template path expecting an integer that represents the page id PATH_TEMPLATE = '/projects?page=%d&order_by=id' MODEL = GitLabModel LISTER_NAME = 'gitlab' def __init__(self, api_baseurl, instance=None, override_config=None, sort='asc', per_page=20): super().__init__(api_baseurl=api_baseurl, override_config=override_config) if instance is None: instance = parse_url(api_baseurl).host self.instance = instance self.PATH_TEMPLATE = '%s&sort=%s' % (self.PATH_TEMPLATE, sort) if per_page != 20: self.PATH_TEMPLATE = '%s&per_page=%s' % ( self.PATH_TEMPLATE, per_page) - @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 - - """ - 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 {} } creds_lister = self.config['credentials'].get(self.instance) if creds_lister: auth = random.choice(creds_lister) if auth: params['auth'] = (auth['username'], auth['password']) return params def uid(self, repo): return '%s/%s' % (self.instance, repo['path_with_namespace']) def get_model_from_repo(self, repo): return { 'instance': self.instance, 'uid': self.uid(repo), '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_int(self, headers, key): _val = headers.get(key) if _val: return int(_val) def get_next_target_from_response(self, response): """Determine the next page identifier. """ return self._get_int(response.headers, 'x-next-page') def get_pages_information(self): """Determine pages information. """ response = self.transport_head(identifier=1) if not response.ok: raise ValueError( 'Problem during information fetch: %s' % response.status_code) h = response.headers return (self._get_int(h, 'x-total'), self._get_int(h, 'x-total-pages'), self._get_int(h, 'x-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/npm/lister.py b/swh/lister/npm/lister.py index 3c64de3..d686851 100644 --- a/swh/lister/npm/lister.py +++ b/swh/lister/npm/lister.py @@ -1,157 +1,158 @@ -# Copyright (C) 2018 the Software Heritage developers +# Copyright (C) 2018-2019 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.parse import quote from swh.lister.core.indexing_lister import SWHIndexingHttpLister from swh.lister.npm.models import NpmModel from swh.scheduler.utils import create_task_dict class NpmListerBase(SWHIndexingHttpLister): """List packages available in the npm registry in a paginated way """ MODEL = NpmModel LISTER_NAME = 'npm' + instance = 'npm' def __init__(self, api_baseurl='https://replicate.npmjs.com', per_page=1000, override_config=None): super().__init__(api_baseurl=api_baseurl, override_config=override_config) self.per_page = per_page + 1 self.PATH_TEMPLATE += '&limit=%s' % self.per_page @property def ADDITIONAL_CONFIG(self): """(Override) Add extra configuration """ default_config = super().ADDITIONAL_CONFIG default_config['loading_task_policy'] = ('str', 'recurring') return default_config def get_model_from_repo(self, repo_name): """(Override) Transform from npm package name to model """ package_url, package_metadata_url = self._compute_urls(repo_name) return { 'uid': repo_name, 'indexable': repo_name, 'name': repo_name, 'full_name': repo_name, 'html_url': package_metadata_url, 'origin_url': package_url, 'origin_type': 'npm', 'description': None } def task_dict(self, origin_type, origin_url, **kwargs): """(Override) Return task dict for loading a npm package into the archive This is overridden from the lister_base as more information is needed for the ingestion task creation. """ task_type = 'load-%s' % origin_type task_policy = self.config['loading_task_policy'] package_name = kwargs.get('name') package_metadata_url = kwargs.get('html_url') return create_task_dict(task_type, task_policy, package_name, origin_url, package_metadata_url=package_metadata_url) def request_headers(self): """(Override) Set requests headers to send when querying the npm registry """ return {'User-Agent': 'Software Heritage npm lister', 'Accept': 'application/json'} def _compute_urls(self, repo_name): """Return a tuple (package_url, package_metadata_url) """ return ( 'https://www.npmjs.com/package/%s' % repo_name, # package metadata url needs to be escaped otherwise some requests # may fail (for instance when a package name contains '/') '%s/%s' % (self.api_baseurl, quote(repo_name, safe='')) ) def string_pattern_check(self, inner, lower, upper=None): """ (Override) Inhibit the effect of that method as packages indices correspond to package names and thus do not respect any kind of fixed length string pattern """ pass class NpmLister(NpmListerBase): """List all packages available in the npm registry in a paginated way """ PATH_TEMPLATE = '/_all_docs?startkey="%s"' def get_next_target_from_response(self, response): """(Override) Get next npm package name to continue the listing """ repos = response.json()['rows'] return repos[-1]['id'] if len(repos) == self.per_page else None def transport_response_simplified(self, response): """(Override) Transform npm registry response to list for model manipulation """ repos = response.json()['rows'] if len(repos) == self.per_page: repos = repos[:-1] return [self.get_model_from_repo(repo['id']) for repo in repos] class NpmIncrementalLister(NpmListerBase): """List packages in the npm registry, updated since a specific update_seq value of the underlying CouchDB database, in a paginated way """ PATH_TEMPLATE = '/_changes?since=%s' @property def CONFIG_BASE_FILENAME(self): # noqa: N802 return 'lister_npm_incremental' def get_next_target_from_response(self, response): """(Override) Get next npm package name to continue the listing """ repos = response.json()['results'] return repos[-1]['seq'] if len(repos) == self.per_page else None def transport_response_simplified(self, response): """(Override) Transform npm registry response to list for model manipulation """ repos = response.json()['results'] if len(repos) == self.per_page: repos = repos[:-1] return [self.get_model_from_repo(repo['id']) for repo in repos] def filter_before_inject(self, models_list): """(Override) Filter out documents in the CouchDB database not related to a npm package """ models_filtered = [] for model in models_list: package_name = model['name'] # document related to CouchDB internals if package_name.startswith('_design/'): continue models_filtered.append(model) return models_filtered def disable_deleted_repo_tasks(self, start, end, keep_these): """(Override) Disable the processing performed by that method as it is not relevant in this incremental lister context and it raises and exception due to a different index type (int instead of str) """ pass diff --git a/swh/lister/phabricator/lister.py b/swh/lister/phabricator/lister.py index dfb77e3..d6e062e 100644 --- a/swh/lister/phabricator/lister.py +++ b/swh/lister/phabricator/lister.py @@ -1,138 +1,143 @@ # Copyright (C) 2019 the Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information +import urllib.parse from swh.lister.core.indexing_lister import SWHIndexingHttpLister from swh.lister.phabricator.models import PhabricatorModel from collections import defaultdict class PhabricatorLister(SWHIndexingHttpLister): PATH_TEMPLATE = '&order=oldest&attachments[uris]=1&after=%s' MODEL = PhabricatorModel LISTER_NAME = 'phabricator' - def __init__(self, forge_url, api_token, override_config=None): + def __init__(self, forge_url, api_token, instance=None, + override_config=None): if forge_url.endswith("/"): forge_url = forge_url[:-1] self.forge_url = forge_url api_endpoint = ('api/diffusion.repository.' 'search?api.token=%s') % api_token api_baseurl = '%s/%s' % (forge_url, api_endpoint) + if not instance: + instance = urllib.parse.urlparse(forge_url).hostname + self.instance = instance super().__init__(api_baseurl=api_baseurl, override_config=override_config) def request_headers(self): """ (Override) Set requests headers to send when querying the Phabricator API """ return {'User-Agent': 'Software Heritage phabricator lister', 'Accept': 'application/json'} def get_model_from_repo(self, repo): url = get_repo_url(repo['attachments']['uris']['uris']) if url is None: return None return { 'uid': self.forge_url + str(repo['id']), 'indexable': repo['id'], 'name': repo['fields']['shortName'], 'full_name': repo['fields']['name'], 'html_url': url, 'origin_url': url, 'description': None, 'origin_type': repo['fields']['vcs'] } def get_next_target_from_response(self, response): body = response.json()['result']['cursor'] if body['after'] != 'null': return body['after'] else: return None def transport_response_simplified(self, response): repos = response.json() if repos['result'] is None: raise ValueError( 'Problem during information fetch: %s' % repos['error_code']) repos = repos['result']['data'] return [self.get_model_from_repo(repo) for repo in repos] def filter_before_inject(self, models_list): """ (Overrides) SWHIndexingLister.filter_before_inject Bounds query results by this Lister's set max_index. """ models_list = [m for m in models_list if m is not None] return super().filter_before_inject(models_list) def _bootstrap_repositories_listing(self): """ Method called when no min_bound value has been provided to the lister. Its purpose is to: 1. get the first repository data hosted on the Phabricator instance 2. inject them into the lister database 3. return the first repository index to start the listing after that value Returns: int: The first repository index """ params = '&order=oldest&limit=1' response = self.safely_issue_request(params) models_list = self.transport_response_simplified(response) self.max_index = models_list[0]['indexable'] models_list = self.filter_before_inject(models_list) injected = self.inject_repo_data_into_db(models_list) self.create_missing_origins_and_tasks(models_list, injected) return self.max_index def run(self, min_bound=None, max_bound=None): """ (Override) Run the lister on the specified Phabricator instance Args: min_bound (int): Optional repository index to start the listing after it max_bound (int): Optional repository index to stop the listing after it """ # initial call to the lister, we need to bootstrap it in that case if min_bound is None: min_bound = self._bootstrap_repositories_listing() super().run(min_bound, max_bound) def get_repo_url(attachments): """ Return url for a hosted repository from its uris attachments according to the following priority lists: * protocol: https > http * identifier: shortname > callsign > id """ processed_urls = defaultdict(dict) for uri in attachments: protocol = uri['fields']['builtin']['protocol'] url = uri['fields']['uri']['effective'] identifier = uri['fields']['builtin']['identifier'] if protocol in ('http', 'https'): processed_urls[protocol][identifier] = url elif protocol is None: for protocol in ('https', 'http'): if url.startswith(protocol): processed_urls[protocol]['undefined'] = url break for protocol in ['https', 'http']: for identifier in ['shortname', 'callsign', 'id', 'undefined']: if (protocol in processed_urls and identifier in processed_urls[protocol]): return processed_urls[protocol][identifier] return None diff --git a/swh/lister/phabricator/tasks.py b/swh/lister/phabricator/tasks.py index 83ecd67..3ebb981 100644 --- a/swh/lister/phabricator/tasks.py +++ b/swh/lister/phabricator/tasks.py @@ -1,28 +1,29 @@ # Copyright (C) 2019 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.scheduler.celery_backend.config import app from swh.lister.phabricator.lister import PhabricatorLister -def new_lister( - forge_url='https://forge.softwareheritage.org', api_token='', **kw): - return PhabricatorLister(forge_url=forge_url, api_token=api_token, **kw) +def new_lister(forge_url='https://forge.softwareheritage.org', api_token='', + instance='swh', **kw): + return PhabricatorLister(forge_url=forge_url, api_token=api_token, + instance=instance, **kw) @app.task(name=__name__ + '.IncrementalPhabricatorLister') def incremental_phabricator_lister(**lister_args): lister = new_lister(**lister_args) lister.run(min_bound=lister.db_last_index()) @app.task(name=__name__ + '.FullPhabricatorLister') def full_phabricator_lister(**lister_args): lister = new_lister(**lister_args) lister.run() @app.task(name=__name__ + '.ping') def ping(): return 'OK' diff --git a/swh/lister/phabricator/tests/test_lister.py b/swh/lister/phabricator/tests/test_lister.py index 29fbcec..6f565ab 100644 --- a/swh/lister/phabricator/tests/test_lister.py +++ b/swh/lister/phabricator/tests/test_lister.py @@ -1,53 +1,55 @@ # Copyright (C) 2019 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 json import unittest from swh.lister.core.tests.test_lister import HttpListerTester from swh.lister.phabricator.lister import PhabricatorLister from swh.lister.phabricator.lister import get_repo_url class PhabricatorListerTester(HttpListerTester, unittest.TestCase): Lister = PhabricatorLister test_re = re.compile(r'\&after=([^?&]+)') lister_subdir = 'phabricator' good_api_response_file = 'api_response.json' good_api_response_undefined_protocol = 'api_response_undefined_'\ 'protocol.json' bad_api_response_file = 'api_empty_response.json' first_index = 1 last_index = 12 entries_per_page = 10 def get_fl(self, override_config=None): """(Override) Retrieve an instance of fake lister (fl). + """ if override_config or self.fl is None: - self.fl = self.Lister(forge_url='https://fakeurl', api_token='a-1', + self.fl = self.Lister(forge_url='https://fakeurl', instance='fake', + api_token='a-1', override_config=override_config) self.fl.INITIAL_BACKOFF = 1 self.fl.reset_backoff() return self.fl def test_get_repo_url(self): f = open('swh/lister/%s/tests/%s' % (self.lister_subdir, self.good_api_response_file)) api_response = json.load(f) repos = api_response['result']['data'] for repo in repos: self.assertEqual( 'https://forge.softwareheritage.org/source/%s.git' % (repo['fields']['shortName']), get_repo_url(repo['attachments']['uris']['uris'])) f = open('swh/lister/%s/tests/%s' % (self.lister_subdir, self.good_api_response_undefined_protocol)) repo = json.load(f) self.assertEqual( 'https://svn.blender.org/svnroot/bf-blender/', get_repo_url(repo['attachments']['uris']['uris'])) diff --git a/swh/lister/phabricator/tests/test_tasks.py b/swh/lister/phabricator/tests/test_tasks.py index 160efcc..a97196f 100644 --- a/swh/lister/phabricator/tests/test_tasks.py +++ b/swh/lister/phabricator/tests/test_tasks.py @@ -1,29 +1,30 @@ from unittest.mock import patch def test_ping(swh_app, celery_session_worker): res = swh_app.send_task( 'swh.lister.phabricator.tasks.ping') assert res res.wait() assert res.successful() assert res.result == 'OK' @patch('swh.lister.phabricator.tasks.PhabricatorLister') def test_incremental(lister, swh_app, celery_session_worker): # setup the mocked PhabricatorLister lister.return_value = lister lister.db_last_index.return_value = 42 lister.run.return_value = None res = swh_app.send_task( 'swh.lister.phabricator.tasks.IncrementalPhabricatorLister') assert res res.wait() assert res.successful() lister.assert_called_once_with( - api_token='', forge_url='https://forge.softwareheritage.org') + api_token='', forge_url='https://forge.softwareheritage.org', + instance='swh') lister.db_last_index.assert_called_once_with() lister.run.assert_called_once_with(min_bound=42) diff --git a/swh/lister/pypi/lister.py b/swh/lister/pypi/lister.py index a488850..fedffd0 100644 --- a/swh/lister/pypi/lister.py +++ b/swh/lister/pypi/lister.py @@ -1,76 +1,77 @@ -# Copyright (C) 2018 the Software Heritage developers +# Copyright (C) 2018-2019 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 xmltodict from .models import PyPIModel from swh.scheduler import utils from swh.lister.core.simple_lister import SimpleLister from swh.lister.core.lister_transports import ListerOnePageApiTransport class PyPILister(ListerOnePageApiTransport, SimpleLister): MODEL = PyPIModel LISTER_NAME = 'pypi' PAGE = 'https://pypi.org/simple/' + instance = 'pypi' # As of today only the main pypi.org is used def __init__(self, override_config=None): ListerOnePageApiTransport .__init__(self) SimpleLister.__init__(self, override_config=override_config) def task_dict(self, origin_type, origin_url, **kwargs): """(Override) Return task format dict This is overridden from the lister_base as more information is needed for the ingestion task creation. """ _type = 'load-%s' % origin_type _policy = 'recurring' project_name = kwargs.get('name') project_metadata_url = kwargs.get('html_url') return utils.create_task_dict( _type, _policy, project_name, origin_url, project_metadata_url=project_metadata_url) def list_packages(self, response): """(Override) List the actual pypi origins from the response. """ result = xmltodict.parse(response.content) _packages = [p['#text'] for p in result['html']['body']['a']] random.shuffle(_packages) return _packages def _compute_urls(self, repo_name): """Returns a tuple (project_url, project_metadata_url) """ return ( 'https://pypi.org/project/%s/' % repo_name, 'https://pypi.org/pypi/%s/json' % repo_name ) def get_model_from_repo(self, repo_name): """(Override) Transform from repository representation to model """ project_url, project_url_meta = self._compute_urls(repo_name) return { 'uid': repo_name, 'name': repo_name, 'full_name': repo_name, 'html_url': project_url_meta, 'origin_url': project_url, 'origin_type': 'pypi', 'description': None, } def transport_response_simplified(self, response): """(Override) Transform response to list for model manipulation """ return [self.get_model_from_repo(repo_name) for repo_name in response]