diff --git a/swh/loader/dir/converters.py b/swh/loader/dir/converters.py index b37054a..48ae672 100644 --- a/swh/loader/dir/converters.py +++ b/swh/loader/dir/converters.py @@ -1,136 +1,136 @@ # Copyright (C) 2015 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 """Convert dir objects to dictionaries suitable for swh.storage""" from datetime import datetime from swh.loader.dir.git.git import GitType def format_to_minutes(offset_str): """Convert a git string timezone format string (e.g +0200, -0310) to minutes. Args: offset_str: a string representing an offset. Returns: A positive or negative number of minutes of such input """ sign = offset_str[0] hours = int(offset_str[1:3]) minutes = int(offset_str[3:]) + (hours * 60) return minutes if sign == '+' else -1 * minutes -def blob_to_content(obj, objects, log=None, +def blob_to_content(obj, log=None, max_content_size=None, origin_id=None): """Convert to a compliant swh content. """ size = obj['length'] ret = { 'sha1': obj['sha1'], 'sha256': obj['sha256'], 'sha1_git': obj['sha1_git'], 'data': obj['data'], 'length': size, 'perms': obj['perms'].value, 'type': obj['type'].value } if max_content_size and size > max_content_size: if log: log.info('Skipping content %s, too large (%s > %s)' % (obj['sha1_git'], size, max_content_size)) - ret.update({'status': 'absent', - 'reason': 'Content too large', - 'origin': origin_id}) + ret.update({'status': 'absent', + 'reason': 'Content too large', + 'origin': origin_id}) return ret ret.update({ 'status': 'visible' }) return ret # Map of type to swh types _entry_type_map = { GitType.TREE: 'dir', GitType.BLOB: 'file', GitType.COMM: 'rev', } def tree_to_directory(tree, objects, log=None): """Format a tree as a directory """ entries = [] for entry in objects[tree['path']]: entries.append({ 'type': _entry_type_map[entry['type']], 'perms': int(entry['perms'].value), 'name': entry['name'], 'target': entry['sha1_git'] }) return { 'id': tree['sha1_git'], 'entries': entries } def commit_to_revision(commit, objects, log=None): """Format a commit as a revision. """ upper_directory = objects[''][0] return { 'id': commit['sha1_git'], 'date': datetime.fromtimestamp(commit['revision_author_date']), 'date_offset': format_to_minutes(commit['revision_author_offset']), 'committer_date': datetime.fromtimestamp(commit['revision_committer_date']), 'committer_date_offset': format_to_minutes(commit['revision_committer_offset']), 'type': commit['revision_type'], 'directory': upper_directory['sha1_git'], 'message': commit['revision_message'], 'author_name': commit['revision_author_name'], 'author_email': commit['revision_author_email'], 'committer_name': commit['revision_committer_name'], 'committer_email': commit['revision_committer_email'], 'parents': [], } -def annotated_tag_to_release(release, objects, log=None): +def annotated_tag_to_release(release, log=None): """Format a swh release. """ return { 'id': release['sha1_git'], 'revision': release['revision_sha1_git'], 'name': release['release_name'], 'comment': release['release_comment'], 'date': datetime.fromtimestamp(release['release_date']), 'date_offset': format_to_minutes(release['release_offset']), 'author_name': release['release_author_name'], 'author_email': release['release_author_email'], } def origin_url_to_origin(origin_url): """Format a pygit2.Repository as an origin suitable for swh.storage""" return { 'type': 'dir', 'url': origin_url, } diff --git a/swh/loader/dir/loader.py b/swh/loader/dir/loader.py index 895117c..ca6520d 100644 --- a/swh/loader/dir/loader.py +++ b/swh/loader/dir/loader.py @@ -1,529 +1,526 @@ # Copyright (C) 2015 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import logging import uuid import traceback import os import psycopg2 from retrying import retry from swh.core import config from swh.loader.dir import converters from swh.loader.dir.git import git from swh.loader.dir.git.git import GitType def send_in_packets(source_list, formatter, sender, packet_size, packet_size_bytes=None, *args, **kwargs): """Send objects from `source_list`, passed through `formatter` (with extra args *args, **kwargs), using the `sender`, in packets of `packet_size` objects (and of max `packet_size_bytes`). """ formatted_objects = [] count = 0 if not packet_size_bytes: packet_size_bytes = 0 for obj in source_list: formatted_object = formatter(obj, *args, **kwargs) if formatted_object: formatted_objects.append(formatted_object) else: continue if packet_size_bytes: count += formatted_object['length'] if len(formatted_objects) >= packet_size or count > packet_size_bytes: sender(formatted_objects) formatted_objects = [] count = 0 if formatted_objects: sender(formatted_objects) def retry_loading(error): """Retry policy when the database raises an integrity error""" if not isinstance(error, psycopg2.IntegrityError): return False logger = logging.getLogger('swh.loader.git.DirLoader') error_name = error.__module__ + '.' + error.__class__.__name__ logger.warning('Retry loading a batch', exc_info=False, extra={ 'swh_type': 'storage_retry', 'swh_exception_type': error_name, 'swh_exception': traceback.format_exception( error.__class__, error, error.__traceback__, ), }) return True class DirLoader(config.SWHConfig): """A bulk loader for a directory""" DEFAULT_CONFIG = { 'storage_class': ('str', 'remote_storage'), 'storage_args': ('list[str]', ['http://localhost:5000/']), 'send_contents': ('bool', True), 'send_directories': ('bool', True), 'send_revisions': ('bool', True), 'send_releases': ('bool', True), 'send_occurrences': ('bool', True), 'content_packet_size': ('int', 10000), 'content_packet_size_bytes': ('int', 1024 * 1024 * 1024), 'directory_packet_size': ('int', 25000), 'revision_packet_size': ('int', 100000), 'release_packet_size': ('int', 100000), 'occurrence_packet_size': ('int', 100000), # origin information 'origin_url': ('str', 'file:///dev/null'), # occurrence information 'branch': ('str', 'master'), 'authority_id': ('int', 1), 'validity': ('str', '2015-01-01 00:00:00+00'), # revision information 'revision_author_name': ('str', 'swh author'), 'revision_author_email': ('str', 'swh@inria.fr'), 'revision_author_date': ('int', '1444054085'), 'revision_author_offset': ('str', '+0200'), 'revision_committer_name': ('str', 'swh committer'), 'revision_committer_email': ('str', 'swh@inria.fr'), 'revision_committer_date': ('int', '1444054085'), 'revision_committer_offset': ('str', '+0200'), 'revision_type': ('str', 'tar'), 'revision_message': ('str', 'synthetic revision'), # release information 'release_name': ('str', 'v0.0.1'), 'release_date': ('int', '1444054085'), 'release_offset': ('str', '+0200'), 'release_author_name': ('str', 'swh author'), 'release_author_email': ('str', 'swh@inria.fr'), 'release_comment': ('str', 'synthetic release'), } def __init__(self, config): self.config = config if self.config['storage_class'] == 'remote_storage': from swh.storage.api.client import RemoteStorage as Storage else: from swh.storage import Storage self.storage = Storage(*self.config['storage_args']) self.log = logging.getLogger('swh.loader.dir.DirLoader') @retry(retry_on_exception=retry_loading, stop_max_attempt_number=3) def send_contents(self, content_list): """Actually send properly formatted contents to the database""" num_contents = len(content_list) log_id = str(uuid.uuid4()) self.log.debug("Sending %d contents" % num_contents, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'content', 'swh_num': num_contents, 'swh_id': log_id, }) self.storage.content_add(content_list) self.log.debug("Done sending %d contents" % num_contents, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'content', 'swh_num': num_contents, 'swh_id': log_id, }) @retry(retry_on_exception=retry_loading, stop_max_attempt_number=3) def send_directories(self, directory_list): """Actually send properly formatted directories to the database""" num_directories = len(directory_list) log_id = str(uuid.uuid4()) self.log.debug("Sending %d directories" % num_directories, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'directory', 'swh_num': num_directories, 'swh_id': log_id, }) self.storage.directory_add(directory_list) self.log.debug("Done sending %d directories" % num_directories, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'directory', 'swh_num': num_directories, 'swh_id': log_id, }) @retry(retry_on_exception=retry_loading, stop_max_attempt_number=3) def send_revisions(self, revision_list): """Actually send properly formatted revisions to the database""" num_revisions = len(revision_list) log_id = str(uuid.uuid4()) self.log.debug("Sending %d revisions" % num_revisions, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'revision', 'swh_num': num_revisions, 'swh_id': log_id, }) self.storage.revision_add(revision_list) self.log.debug("Done sending %d revisions" % num_revisions, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'revision', 'swh_num': num_revisions, 'swh_id': log_id, }) @retry(retry_on_exception=retry_loading, stop_max_attempt_number=3) def send_releases(self, release_list): """Actually send properly formatted releases to the database""" num_releases = len(release_list) log_id = str(uuid.uuid4()) self.log.debug("Sending %d releases" % num_releases, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'release', 'swh_num': num_releases, 'swh_id': log_id, }) self.storage.release_add(release_list) self.log.debug("Done sending %d releases" % num_releases, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'release', 'swh_num': num_releases, 'swh_id': log_id, }) @retry(retry_on_exception=retry_loading, stop_max_attempt_number=3) def send_occurrences(self, occurrence_list): """Actually send properly formatted occurrences to the database""" num_occurrences = len(occurrence_list) log_id = str(uuid.uuid4()) self.log.debug("Sending %d occurrences" % num_occurrences, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'occurrence', 'swh_num': num_occurrences, 'swh_id': log_id, }) self.storage.occurrence_add(occurrence_list) self.log.debug("Done sending %d occurrences" % num_occurrences, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'occurrence', 'swh_num': num_occurrences, 'swh_id': log_id, }) def get_or_create_origin(self, origin_url): origin = converters.origin_url_to_origin(origin_url) origin['id'] = self.storage.origin_add_one(origin) return origin def dir_origin(self, root_dir, origin_url): log_id = str(uuid.uuid4()) self.log.debug('Creating origin for %s' % origin_url, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'origin', 'swh_num': 1, 'swh_id': log_id }) origin = self.get_or_create_origin(origin_url) self.log.debug('Done creating origin for %s' % origin_url, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'origin', 'swh_num': 1, 'swh_id': log_id }) return origin def dir_revision(self, root_dir, origin_url, revision_date, revision_offset, revision_committer_date, revision_committer_offset, revision_type, revision_message, revision_author, revision_committer): """Create a revision. """ log_id = str(uuid.uuid4()) self.log.debug('Creating origin for %s' % origin_url, extra={ 'swh_type': 'storage_send_start', 'swh_content_type': 'origin', 'swh_num': 1, 'swh_id': log_id }) self.get_or_create_origin(origin_url) self.log.debug('Done creating origin for %s' % origin_url, extra={ 'swh_type': 'storage_send_end', 'swh_content_type': 'origin', 'swh_num': 1, 'swh_id': log_id }) def bulk_send_blobs(self, objects, blobs, origin_id): """Format blobs as swh contents and send them to the database""" packet_size = self.config['content_packet_size'] packet_size_bytes = self.config['content_packet_size_bytes'] max_content_size = self.config['content_size_limit'] send_in_packets(blobs, converters.blob_to_content, self.send_contents, packet_size, packet_size_bytes=packet_size_bytes, - objects=objects, log=self.log, max_content_size=max_content_size, origin_id=origin_id) def bulk_send_trees(self, objects, trees): """Format trees as swh directories and send them to the database""" packet_size = self.config['directory_packet_size'] send_in_packets(trees, converters.tree_to_directory, self.send_directories, packet_size, objects=objects, log=self.log) def bulk_send_commits(self, objects, commits): """Format commits as swh revisions and send them to the database""" packet_size = self.config['revision_packet_size'] send_in_packets(commits, converters.commit_to_revision, self.send_revisions, packet_size, objects=objects, log=self.log) def bulk_send_annotated_tags(self, objects, tags): """Format annotated tags (pygit2.Tag objects) as swh releases and send them to the database """ packet_size = self.config['release_packet_size'] send_in_packets(tags, converters.annotated_tag_to_release, self.send_releases, packet_size, - objects=objects, log=self.log) def bulk_send_refs(self, objects, refs): """Format git references as swh occurrences and send them to the database """ packet_size = self.config['occurrence_packet_size'] send_in_packets(refs, lambda ref: ref, self.send_occurrences, packet_size) def compute_dir_ref(self, root_dir, branch, revision_hash, origin_id, authority_id, validity): """List all the refs from the given root directory root_dir. Args: - root_dir: the root directory - branch: occurrence's branch name - revision_hash: the revision hash - origin_id (int): the id of the origin from which the root_dir is taken - validity (datetime.datetime): the validity date for the repository's refs - authority_id (int): the id of the authority on `validity`. Returns: One dictionary with keys: - branch (str): name of the ref - revision (sha1_git): revision pointed at by the ref - origin (int) - validity (datetime.DateTime) - authority (int) Compatible with occurrence_add. """ - log_id = str(uuid.uuid4()) self.log.debug("Computing occurrence %s representation at %s" % ( branch, revision_hash), extra={ 'swh_type': 'computing_occurrence_dir', 'swh_name': branch, 'swh_target': str(revision_hash), 'swh_id': log_id, }) return { 'branch': branch, 'revision': revision_hash, 'origin': origin_id, 'validity': validity, 'authority': authority_id, } def list_repo_objs(self, root_dir, info): """List all objects from root_dir. Args: - root_dir (path): the directory to list Returns: a dict containing lists of `Oid`s with keys for each object type: - CONTENT - DIRECTORY """ def get_objects_per_object_type(objects_per_path): m = { GitType.BLOB: [], GitType.TREE: [], GitType.COMM: [], GitType.RELE: [] } for tree_path in objects_per_path: objs = objects_per_path[tree_path] # print('tree_path: %s, objs: %s' % (tree_path, objs)) for obj in objs: m[obj['type']].append(obj) return m log_id = str(uuid.uuid4()) self.log.info("Started listing %s" % root_dir, extra={ 'swh_type': 'git_list_objs_start', 'swh_repo': root_dir, 'swh_id': log_id, }) objects_per_path = git.walk_and_compute_sha1_from_directory(root_dir) objects = get_objects_per_object_type(objects_per_path) tree_hash = objects_per_path[''][0]['sha1_git'] revision = git.compute_revision_git_sha1(tree_hash, info) objects.update({ GitType.COMM: [revision] }) revision_hash = revision['sha1_git'] release = git.compute_release(revision_hash, info) objects.update({ GitType.RELE: [release] }) self.log.info("Done listing the objects in %s: %d contents, " "%d directories, %d revisions, %d releases" % ( root_dir, len(objects[GitType.BLOB]), len(objects[GitType.TREE]), len(objects[GitType.COMM]), len(objects[GitType.RELE]) ), extra={ 'swh_type': 'git_list_objs_end', 'swh_repo': root_dir, 'swh_num_blobs': len(objects[GitType.BLOB]), 'swh_num_trees': len(objects[GitType.TREE]), 'swh_num_commits': len(objects[GitType.COMM]), 'swh_num_releases': len(objects[GitType.RELE]), 'swh_id': log_id, }) return objects, objects_per_path def load_dir(self, root_dir, objects, objects_per_path, refs, origin_id): if self.config['send_contents']: self.bulk_send_blobs(objects_per_path, objects[GitType.BLOB], origin_id) else: self.log.info('Not sending contents') if self.config['send_directories']: self.bulk_send_trees(objects_per_path, objects[GitType.TREE]) else: self.log.info('Not sending directories') if self.config['send_revisions']: self.bulk_send_commits(objects_per_path, objects[GitType.COMM]) else: self.log.info('Not sending revisions') if self.config['send_releases']: self.bulk_send_annotated_tags(objects_per_path, objects[GitType.RELE]) else: self.log.info('Not sending releases') if self.config['send_occurrences']: self.bulk_send_refs(objects_per_path, refs) else: self.log.info('Not sending occurrences') def process(self, root_dir): if not os.path.exists(root_dir): self.log.info('Skipping inexistant directory %s' % root_dir, extra={ 'swh_type': 'dir_repo_list_refs', 'swh_repo': root_dir, 'swh_num_refs': 0, }) return files = os.listdir(root_dir) if not(files): self.log.info('Skipping empty directory %s' % root_dir, extra={ 'swh_type': 'dir_repo_list_refs', 'swh_repo': root_dir, 'swh_num_refs': 0, }) return # Add origin to storage if needed, use the one from config if not origin = self.dir_origin(root_dir, self.config['origin_url']) # We want to load the repository, walk all the objects objects, objects_per_path = self.list_repo_objs(root_dir, self.config) # Compute revision information (mixed from outside input + dir content) revision = objects[GitType.COMM][0] # Parse all the refs from our root_dir ref = self.compute_dir_ref(root_dir, self.config['branch'], revision['sha1_git'], origin['id'], self.config['authority_id'], self.config['validity']) # Finally, load the repository self.load_dir(root_dir, objects, objects_per_path, [ref], origin['id']) diff --git a/swh/loader/dir/tests/test_converters.py b/swh/loader/dir/tests/test_converters.py new file mode 100644 index 0000000..a3ab907 --- /dev/null +++ b/swh/loader/dir/tests/test_converters.py @@ -0,0 +1,206 @@ +# Copyright (C) 2015 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 unittest + +from nose.tools import istest +from datetime import datetime + +from swh.loader.dir import converters +from swh.loader.dir.git.git import GitType, GitPerm + + +class TestConverters(unittest.TestCase): + @istest + def format_to_minutes(self): + self.assertEquals(converters.format_to_minutes('+0100'), 60) + self.assertEquals(converters.format_to_minutes('-0200'), -120) + self.assertEquals(converters.format_to_minutes('+1250'), 12*60+50) + self.assertEquals(converters.format_to_minutes('+0000'), 0) + self.assertEquals(converters.format_to_minutes('-0000'), 0) + + @istest + def origin_url_to_origin(self): + # given + origin_url = 'foobar' + + # when + self.assertDictEqual({ + 'type': 'dir', + 'url': origin_url, + }, converters.origin_url_to_origin(origin_url)) + + @istest + def annotated_tag_to_release(self): + # given + release = { + 'sha1_git': '123', + 'revision_sha1_git': '456', + 'release_name': 'some-release', + 'release_comment': 'some-comment-on-release', + 'release_date': 1444054085, + 'release_offset': '-0300', + 'release_author_name': 'someone', + 'release_author_email': 'someone@whatelse.eu' + } + + expected_release = { + 'id': '123', + 'revision': '456', + 'name': 'some-release', + 'comment': 'some-comment-on-release', + 'date': datetime.fromtimestamp(1444054085), + 'date_offset': -180, + 'author_name': 'someone', + 'author_email': 'someone@whatelse.eu', + } + + # when + actual_release = converters.annotated_tag_to_release(release) + + # then + self.assertDictEqual( + expected_release, + actual_release) + + @istest + def blob_to_content_visible(self): + obj = { + 'length': 9, + 'data': b'some-data', + 'sha1': b'sha1', + 'sha1_git': b'sha1-git', + 'sha256': b'sha256', + 'perms': GitPerm.BLOB, + 'type': GitType.BLOB + } + + expected_content = { + 'length': 9, + 'data': b'some-data', + 'sha1': b'sha1', + 'sha1_git': b'sha1-git', + 'sha256': b'sha256', + 'perms': GitPerm.BLOB.value, + 'type': GitType.BLOB.value, + 'status': 'visible' + } + + # when + actual_content = converters.blob_to_content(obj) + + # then + self.assertEqual(expected_content, actual_content) + + @istest + def blob_to_content_absent(self): + obj = { + 'length': 9, + 'data': b'some-data', + 'sha1': b'sha1', + 'sha1_git': b'sha1-git', + 'sha256': b'sha256', + 'perms': GitPerm.BLOB, + 'type': GitType.BLOB + } + + expected_content = { + 'length': 9, + 'data': b'some-data', + 'sha1': b'sha1', + 'sha1_git': b'sha1-git', + 'sha256': b'sha256', + 'perms': GitPerm.BLOB.value, + 'type': GitType.BLOB.value, + 'status': 'absent', + 'reason': 'Content too large', + 'origin': 3} + + # when + actual_content = converters.blob_to_content(obj, + max_content_size=5, + origin_id=3) + + # then + self.assertEqual(expected_content, actual_content) + + @istest + def tree_to_directory_no_entries(self): + # given + tree = { + 'path': 'foo', + 'sha1_git': b'tree_sha1_git' + } + objects = { + 'foo': [{'type': GitType.TREE, + 'perms': GitPerm.TREE, + 'name': 'bar', + 'sha1_git': b'sha1-target'}, + {'type': GitType.BLOB, + 'perms': GitPerm.BLOB, + 'name': 'file-foo', + 'sha1_git': b'file-foo-sha1-target'}] + } + + expected_directory = { + 'id': b'tree_sha1_git', + 'entries': [{'type': 'dir', + 'perms': int(GitPerm.TREE.value), + 'name': 'bar', + 'target': b'sha1-target'}, + {'type': 'file', + 'perms': int(GitPerm.BLOB.value), + 'name': 'file-foo', + 'target': b'file-foo-sha1-target'}] + } + + # when + actual_directory = converters.tree_to_directory(tree, objects) + + # then + self.assertEqual(actual_directory, expected_directory) + + @istest + def commit_to_revision(self): + # given + commit = { + 'sha1_git': 'commit-git-sha1', + 'revision_author_date': 1444054085, + 'revision_author_offset': '+0000', + 'revision_committer_date': 1444054085, + 'revision_committer_offset': '-0000', + 'revision_type': 'tar', + 'revision_message': 'synthetic-message-input', + 'revision_author_name': 'author-name', + 'revision_author_email': 'author-email', + 'revision_committer_name': 'committer-name', + 'revision_committer_email': 'committer-email', + } + + objects = { + '': [{'sha1_git': 'targeted-tree-sha1'}] + } + + expected_revision = { + 'id': 'commit-git-sha1', + 'date': datetime.fromtimestamp(1444054085), + 'date_offset': 0, + 'committer_date': datetime.fromtimestamp(1444054085), + 'committer_date_offset': 0, + 'type': 'tar', + 'directory': 'targeted-tree-sha1', + 'message': 'synthetic-message-input', + 'author_name': 'author-name', + 'author_email': 'author-email', + 'committer_name': 'committer-name', + 'committer_email': 'committer-email', + 'parents': [], + } + + # when + actual_revision = converters.commit_to_revision(commit, objects) + + # then + self.assertEquals(actual_revision, expected_revision)