Learn more  » Push, build, and install  RubyGems npm packages Python packages Maven artifacts PHP packages Go Modules Bower components Debian packages RPM packages NuGet packages

agriconnect / dulwich   python

Repository URL to install this package:

/ walk.py

# walk.py -- General implementation of walking commits and their contents.
# Copyright (C) 2010 Google, Inc.
# Dulwich is dual-licensed under the Apache License, Version 2.0 and the GNU
# General Public License as public by the Free Software Foundation; version 2.0
# or (at your option) any later version. You can redistribute it and/or
# modify it under the terms of either of these two licenses.
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# See the License for the specific language governing permissions and
# limitations under the License.
# You should have received a copy of the licenses; if not, see
# <http://www.gnu.org/licenses/> for a copy of the GNU General Public License
# and <http://www.apache.org/licenses/LICENSE-2.0> for a copy of the Apache
# License, Version 2.0.

"""General implementation of walking commits and their contents."""

from collections import defaultdict

import collections
import heapq
from itertools import chain

from dulwich.diff_tree import (
from dulwich.errors import (
from dulwich.objects import (

ORDER_DATE = 'date'
ORDER_TOPO = 'topo'


# Maximum number of commits to walk past a commit time boundary.

class WalkEntry(object):
    """Object encapsulating a single result from a walk."""

    def __init__(self, walker, commit):
        self.commit = commit
        self._store = walker.store
        self._get_parents = walker.get_parents
        self._changes = {}
        self._rename_detector = walker.rename_detector

    def changes(self, path_prefix=None):
        """Get the tree changes for this entry.

        :param path_prefix: Portion of the path in the repository to
            use to filter changes. Must be a directory name. Must be
            a full, valid, path reference (no partial names or wildcards).
        :return: For commits with up to one parent, a list of TreeChange
            objects; if the commit has no parents, these will be relative to
            the empty tree. For merge commits, a list of lists of TreeChange
            objects; see dulwich.diff.tree_changes_for_merge.
        cached = self._changes.get(path_prefix)
        if cached is None:
            commit = self.commit
            if not self._get_parents(commit):
                changes_func = tree_changes
                parent = None
            elif len(self._get_parents(commit)) == 1:
                changes_func = tree_changes
                parent = self._store[self._get_parents(commit)[0]].tree
                if path_prefix:
                    mode, subtree_sha = parent.lookup_path(
                    parent = self._store[subtree_sha]
                changes_func = tree_changes_for_merge
                parent = [
                        self._store[p].tree for p in self._get_parents(commit)]
                if path_prefix:
                    parent_trees = [self._store[p] for p in parent]
                    parent = []
                    for p in parent_trees:
                            mode, st = p.lookup_path(
                        except KeyError:
            commit_tree_sha = commit.tree
            if path_prefix:
                commit_tree = self._store[commit_tree_sha]
                mode, commit_tree_sha = commit_tree.lookup_path(
            cached = list(changes_func(
              self._store, parent, commit_tree_sha,
            self._changes[path_prefix] = cached
        return self._changes[path_prefix]

    def __repr__(self):
        return '<WalkEntry commit=%s, changes=%r>' % (
          self.commit.id, self.changes())

class _CommitTimeQueue(object):
    """Priority queue of WalkEntry objects by commit time."""

    def __init__(self, walker):
        self._walker = walker
        self._store = walker.store
        self._get_parents = walker.get_parents
        self._excluded = walker.excluded
        self._pq = []
        self._pq_set = set()
        self._seen = set()
        self._done = set()
        self._min_time = walker.since
        self._last = None
        self._extra_commits_left = _MAX_EXTRA_COMMITS
        self._is_finished = False

        for commit_id in chain(walker.include, walker.excluded):

    def _push(self, object_id):
            obj = self._store[object_id]
        except KeyError:
            raise MissingCommitError(object_id)
        if isinstance(obj, Tag):
        # TODO(jelmer): What to do about non-Commit and non-Tag objects?
        commit = obj
        if commit.id not in self._pq_set and commit.id not in self._done:
            heapq.heappush(self._pq, (-commit.commit_time, commit))

    def _exclude_parents(self, commit):
        excluded = self._excluded
        seen = self._seen
        todo = [commit]
        while todo:
            commit = todo.pop()
            for parent in self._get_parents(commit):
                if parent not in excluded and parent in seen:
                    # TODO: This is inefficient unless the object store does
                    # some caching (which DiskObjectStore currently does not).
                    # We could either add caching in this class or pass around
                    # parsed queue entry objects instead of commits.

    def next(self):
        if self._is_finished:
            return None
        while self._pq:
            _, commit = heapq.heappop(self._pq)
            sha = commit.id
            if sha in self._done:

            for parent_id in self._get_parents(commit):

            reset_extra_commits = True
            is_excluded = sha in self._excluded
            if is_excluded:
                if self._pq and all(c.id in self._excluded
                                    for _, c in self._pq):
                    _, n = self._pq[0]
                    if self._last and n.commit_time >= self._last.commit_time:
                        # If the next commit is newer than the last one, we
                        # need to keep walking in case its parents (which we
                        # may not have seen yet) are excluded. This gives the
                        # excluded set a chance to "catch up" while the commit
                        # is still in the Walker's output queue.
                        reset_extra_commits = True
                        reset_extra_commits = False

            if (self._min_time is not None and
                    commit.commit_time < self._min_time):
                # We want to stop walking at min_time, but commits at the
                # boundary may be out of order with respect to their parents.
                # So we walk _MAX_EXTRA_COMMITS more commits once we hit this
                # boundary.
                reset_extra_commits = False

            if reset_extra_commits:
                # We're not at a boundary, so reset the counter.
                self._extra_commits_left = _MAX_EXTRA_COMMITS
                self._extra_commits_left -= 1
                if not self._extra_commits_left:

            if not is_excluded:
                self._last = commit
                return WalkEntry(self._walker, commit)
        self._is_finished = True
        return None

    __next__ = next

class Walker(object):
    """Object for performing a walk of commits in a store.

    Walker objects are initialized with a store and other options and can then
    be treated as iterators of Commit objects.

    def __init__(self, store, include, exclude=None, order=ORDER_DATE,
                 reverse=False, max_entries=None, paths=None,
                 rename_detector=None, follow=False, since=None, until=None,
                 get_parents=lambda commit: commit.parents,

        :param store: ObjectStore instance for looking up objects.
        :param include: Iterable of SHAs of commits to include along with their
        :param exclude: Iterable of SHAs of commits to exclude along with their
            ancestors, overriding includes.
        :param order: ORDER_* constant specifying the order of results.
            Anything other than ORDER_DATE may result in O(n) memory usage.
        :param reverse: If True, reverse the order of output, requiring O(n)
        :param max_entries: The maximum number of entries to yield, or None for
            no limit.
        :param paths: Iterable of file or subtree paths to show entries for.
        :param rename_detector: diff.RenameDetector object for detecting
        :param follow: If True, follow path across renames/copies. Forces a
            default rename_detector.
        :param since: Timestamp to list commits after.
        :param until: Timestamp to list commits before.
        :param get_parents: Method to retrieve the parents of a commit
        :param queue_cls: A class to use for a queue of commits, supporting the
            iterator protocol. The constructor takes a single argument, the
        # Note: when adding arguments to this method, please also update
        # dulwich.repo.BaseRepo.get_walker
        if order not in ALL_ORDERS:
            raise ValueError('Unknown walk order %s' % order)
        self.store = store
        if isinstance(include, bytes):
            # TODO(jelmer): Really, this should require a single type.
            # Print deprecation warning here?
            include = [include]
        self.include = include
        self.excluded = set(exclude or [])
        self.order = order
        self.reverse = reverse
        self.max_entries = max_entries
        self.paths = paths and set(paths) or None
        if follow and not rename_detector:
            rename_detector = RenameDetector(store)
        self.rename_detector = rename_detector
        self.get_parents = get_parents
        self.follow = follow
        self.since = since
        self.until = until

        self._num_entries = 0
        self._queue = queue_cls(self)
        self._out_queue = collections.deque()

    def _path_matches(self, changed_path):
        if changed_path is None:
            return False
        for followed_path in self.paths:
            if changed_path == followed_path:
                return True
            if (changed_path.startswith(followed_path) and
                    changed_path[len(followed_path)] == b'/'[0]):
                return True
        return False

    def _change_matches(self, change):
        if not change:
            return False

        old_path = change.old.path
        new_path = change.new.path
        if self._path_matches(new_path):
            if self.follow and change.type in RENAME_CHANGE_TYPES:
            return True
        elif self._path_matches(old_path):
            return True
        return False

    def _should_return(self, entry):
        """Determine if a walk entry should be returned..

        :param entry: The WalkEntry to consider.
        :return: True if the WalkEntry should be returned by this walk, or
            False otherwise (e.g. if it doesn't match any requested paths).
        commit = entry.commit
        if self.since is not None and commit.commit_time < self.since:
            return False
        if self.until is not None and commit.commit_time > self.until:
            return False
        if commit.id in self.excluded:
            return False

        if self.paths is None:
            return True

        if len(self.get_parents(commit)) > 1:
            for path_changes in entry.changes():
                # For merge commits, only include changes with conflicts for
                # this path. Since a rename conflict may include different
                # old.paths, we have to check all of them.
                for change in path_changes:
                    if self._change_matches(change):
                        return True
Loading ...