Don't use -p flag together with -s in git calls. Latest

git version changes behaviour of such call changing the output.

In all the calls where -p -s was used we're only interested in
list of changesets and not the diff.

rhodecode/controllers/compare.py

@@ -1,265 +1,265 b''
 # -*- coding: utf-8 -*-
 """
     rhodecode.controllers.compare
     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
     compare controller for pylons showing differences between two
     repos, branches, bookmarks or tips
 
     :created_on: May 6, 2012
     :author: marcink
     :copyright: (C) 2010-2012 Marcin Kuzminski <[e-mail]>
     :license: GPLv3, see COPYING for more details.
 """
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
 # the Free Software Foundation, either version 3 of the License, or
 # (at your option) any later version.
 #
 # This program is distributed in the hope that it will be useful,
 # but WITHOUT ANY WARRANTY; without even the implied warranty of
 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 # GNU General Public License for more details.
 #
 # You should have received a copy of the GNU General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
 import logging
 import traceback
 import re
 
 from webob.exc import HTTPNotFound, HTTPBadRequest
 from pylons import request, response, session, tmpl_context as c, url
 from pylons.controllers.util import abort, redirect
 from pylons.i18n.translation import _
 
 from rhodecode.lib.vcs.exceptions import EmptyRepositoryError, RepositoryError
 from rhodecode.lib.vcs.utils import safe_str
 from rhodecode.lib.vcs.utils.hgcompat import scmutil, unionrepo
 from rhodecode.lib import helpers as h
 from rhodecode.lib.base import BaseRepoController, render
 from rhodecode.lib.auth import LoginRequired, HasRepoPermissionAnyDecorator
 from rhodecode.lib import diffs
 from rhodecode.lib.utils2 import safe_str
 from rhodecode.model.db import Repository
 from rhodecode.lib.diffs import LimitedDiffContainer
 
 
 log = logging.getLogger(__name__)
 
 
 class CompareController(BaseRepoController):
 
     def __before__(self):
         super(CompareController, self).__before__()
 
     def __get_rev_or_redirect(self, ref, repo, redirect_after=True,
                              partial=False):
         """
         Safe way to get changeset if error occur it redirects to changeset with
         proper message. If partial is set then don't do redirect raise Exception
         instead
 
         :param rev: revision to fetch
         :param repo: repo instance
         """
 
         rev = ref[1] # default and used for git
         if repo.scm_instance.alias == 'hg':
             # lookup up the exact node id
             _revset_predicates = {
                     'branch': 'branch',
                     'book': 'bookmark',
                     'tag': 'tag',
                     'rev': 'id',
                 }
             rev_spec = "max(%s(%%s))" % _revset_predicates[ref[0]]
             revs = repo.scm_instance._repo.revs(rev_spec, safe_str(ref[1]))
             if revs:
                 rev = revs[-1]
             # else: TODO: just report 'not found'
 
         try:
             return repo.scm_instance.get_changeset(rev).raw_id
         except EmptyRepositoryError, e:
             if not redirect_after:
                 return None
             h.flash(h.literal(_('There are no changesets yet')),
                     category='warning')
             redirect(url('summary_home', repo_name=repo.repo_name))
 
         except RepositoryError, e:
             log.error(traceback.format_exc())
             h.flash(safe_str(e), category='warning')
             if not partial:
                 redirect(h.url('summary_home', repo_name=repo.repo_name))
             raise HTTPBadRequest()
 
     def _get_changesets(self, alias, org_repo, org_rev, other_repo, other_rev, merge):
         """
         Returns a list of changesets that can be merged from org_repo@org_rev
         to other_repo@other_rev ... and the ancestor that would be used for merge
         """
 
         ancestor = None
 
         if org_rev == other_rev:
             changesets = []
             if merge:
                 ancestor = org_rev
 
         elif alias == 'hg':
             #case two independent repos
             if org_repo != other_repo:
                 hgrepo = unionrepo.unionrepository(other_repo.baseui,
                                                    other_repo.path,
                                                    org_repo.path)
                 # all the changesets we are looking for will be in other_repo,
                 # so rev numbers from hgrepo can be used in other_repo
 
             #no remote compare do it on the same repository
             else:
                 hgrepo = other_repo._repo
 
             if merge:
                 revs = hgrepo.revs("ancestors(id(%s)) and not ancestors(id(%s)) and not id(%s)",
                                    other_rev, org_rev, org_rev)
 
                 ancestors = hgrepo.revs("ancestor(id(%s), id(%s))", org_rev, other_rev)
                 if ancestors:
                     # pick arbitrary ancestor - but there is usually only one
                     ancestor = hgrepo[ancestors[0]].hex()
             else:
                 # TODO: have both + and - changesets
                 revs = hgrepo.revs("id(%s) :: id(%s) - id(%s)",
                                    org_rev, other_rev, org_rev)
 
             changesets = [other_repo.get_changeset(rev) for rev in revs]
 
         elif alias == 'git':
             if org_repo != other_repo:
                 raise Exception('Comparing of different GIT repositories is not'
                                 'allowed. Got %s != %s' % (org_repo, other_repo))
 
             so, se = org_repo.run_git_command(
-                'log --reverse --pretty="format: %%H" -s -p %s..%s'
+                'log --reverse --pretty="format: %%H" -s %s..%s'
                     % (org_rev, other_rev)
             )
             changesets = [org_repo.get_changeset(cs)
                           for cs in re.findall(r'[0-9a-fA-F]{40}', so)]
 
         return changesets, ancestor
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.read', 'repository.write',
                                    'repository.admin')
     def index(self, org_ref_type, org_ref, other_ref_type, other_ref):
         # org_ref will be evaluated in org_repo
         org_repo = c.rhodecode_db_repo.repo_name
         org_ref = (org_ref_type, org_ref)
         # other_ref will be evaluated in other_repo
         other_ref = (other_ref_type, other_ref)
         other_repo = request.GET.get('other_repo', org_repo)
         # If merge is True:
         #   Show what org would get if merged with other:
         #   List changesets that are ancestors of other but not of org.
         #   New changesets in org is thus ignored.
         #   Diff will be from common ancestor, and merges of org to other will thus be ignored.
         # If merge is False:
         #   Make a raw diff from org to other, no matter if related or not.
         #   Changesets in one and not in the other will be ignored
         merge = bool(request.GET.get('merge'))
         # fulldiff disables cut_off_limit
         c.fulldiff = request.GET.get('fulldiff')
         # partial uses compare_cs.html template directly
         partial = request.environ.get('HTTP_X_PARTIAL_XHR')
         # as_form puts hidden input field with changeset revisions
         c.as_form = partial and request.GET.get('as_form')
         # swap url for compare_diff page - never partial and never as_form
         c.swap_url = h.url('compare_url',
             repo_name=other_repo,
             org_ref_type=other_ref[0], org_ref=other_ref[1],
             other_repo=org_repo,
             other_ref_type=org_ref[0], other_ref=org_ref[1],
             merge=merge or '')
 
         org_repo = Repository.get_by_repo_name(org_repo)
         other_repo = Repository.get_by_repo_name(other_repo)
 
         if org_repo is None:
             log.error('Could not find org repo %s' % org_repo)
             raise HTTPNotFound
         if other_repo is None:
             log.error('Could not find other repo %s' % other_repo)
             raise HTTPNotFound
 
         if org_repo != other_repo and h.is_git(org_repo):
             log.error('compare of two remote repos not available for GIT REPOS')
             raise HTTPNotFound
 
         if org_repo.scm_instance.alias != other_repo.scm_instance.alias:
             log.error('compare of two different kind of remote repos not available')
             raise HTTPNotFound
 
         org_rev = self.__get_rev_or_redirect(ref=org_ref, repo=org_repo, partial=partial)
         other_rev = self.__get_rev_or_redirect(ref=other_ref, repo=other_repo, partial=partial)
 
         c.org_repo = org_repo
         c.other_repo = other_repo
         c.org_ref = org_ref[1]
         c.other_ref = other_ref[1]
         c.org_ref_type = org_ref[0]
         c.other_ref_type = other_ref[0]
 
         c.cs_ranges, c.ancestor = self._get_changesets(org_repo.scm_instance.alias,
                                                        org_repo.scm_instance, org_rev,
                                                        other_repo.scm_instance, other_rev,
                                                        merge)
 
         c.statuses = c.rhodecode_db_repo.statuses([x.raw_id for x in
                                                    c.cs_ranges])
         if merge and not c.ancestor:
             log.error('Unable to find ancestor revision')
 
         if partial:
             return render('compare/compare_cs.html')
 
         if c.ancestor:
             assert merge
             # case we want a simple diff without incoming changesets,
             # previewing what will be merged.
             # Make the diff on the other repo (which is known to have other_ref)
             log.debug('Using ancestor %s as org_ref instead of %s'
                       % (c.ancestor, org_ref))
             org_rev = c.ancestor
             org_repo = other_repo
 
         diff_limit = self.cut_off_limit if not c.fulldiff else None
 
         log.debug('running diff between %s and %s in %s'
                   % (org_rev, other_rev, org_repo.scm_instance.path))
         txtdiff = org_repo.scm_instance.get_diff(rev1=org_rev, rev2=other_rev)
 
         diff_processor = diffs.DiffProcessor(txtdiff or '', format='gitdiff',
                                              diff_limit=diff_limit)
         _parsed = diff_processor.prepare()
 
         c.limited_diff = False
         if isinstance(_parsed, LimitedDiffContainer):
             c.limited_diff = True
 
         c.files = []
         c.changes = {}
         c.lines_added = 0
         c.lines_deleted = 0
         for f in _parsed:
             st = f['stats']
             if not st['binary']:
                 c.lines_added += st['added']
                 c.lines_deleted += st['deleted']
             fid = h.FID('', f['filename'])
             c.files.append([fid, f['operation'], f['filename'], f['stats']])
             htmldiff = diff_processor.as_html(enable_comments=False, parsed_lines=[f])
             c.changes[fid] = [f['operation'], f['filename'], htmldiff]
 
         return render('compare/compare_diff.html')

rhodecode/lib/vcs/backends/git/changeset.py

@@ -1,554 +1,552 b''
 import re
 from itertools import chain
 from dulwich import objects
 from subprocess import Popen, PIPE
 
 from rhodecode.lib.vcs.conf import settings
 from rhodecode.lib.vcs.backends.base import BaseChangeset, EmptyChangeset
 from rhodecode.lib.vcs.exceptions import (
     RepositoryError, ChangesetError, NodeDoesNotExistError, VCSError,
     ChangesetDoesNotExistError, ImproperArchiveTypeError
 )
 from rhodecode.lib.vcs.nodes import (
     FileNode, DirNode, NodeKind, RootNode, RemovedFileNode, SubModuleNode,
     ChangedFileNodesGenerator, AddedFileNodesGenerator, RemovedFileNodesGenerator
 )
 from rhodecode.lib.vcs.utils import (
     safe_unicode, safe_str, safe_int, date_fromtimestamp
 )
 from rhodecode.lib.vcs.utils.lazy import LazyProperty
 
 
 class GitChangeset(BaseChangeset):
     """
     Represents state of the repository at single revision.
     """
 
     def __init__(self, repository, revision):
         self._stat_modes = {}
         self.repository = repository
 
         try:
             commit = self.repository._repo[revision]
             if isinstance(commit, objects.Tag):
                 revision = commit.object[1]
                 commit = self.repository._repo.get_object(commit.object[1])
         except KeyError:
             raise RepositoryError("Cannot get object with id %s" % revision)
         self.raw_id = revision
         self.id = self.raw_id
         self.short_id = self.raw_id[:12]
         self._commit = commit
         self._tree_id = commit.tree
         self._committer_property = 'committer'
         self._author_property = 'author'
         self._date_property = 'commit_time'
         self._date_tz_property = 'commit_timezone'
         self.revision = repository.revisions.index(revision)
 
         self.nodes = {}
         self._paths = {}
 
     @LazyProperty
     def message(self):
         return safe_unicode(self._commit.message)
 
     @LazyProperty
     def committer(self):
         return safe_unicode(getattr(self._commit, self._committer_property))
 
     @LazyProperty
     def author(self):
         return safe_unicode(getattr(self._commit, self._author_property))
 
     @LazyProperty
     def date(self):
         return date_fromtimestamp(getattr(self._commit, self._date_property),
                                   getattr(self._commit, self._date_tz_property))
 
     @LazyProperty
     def _timestamp(self):
         return getattr(self._commit, self._date_property)
 
     @LazyProperty
     def status(self):
         """
         Returns modified, added, removed, deleted files for current changeset
         """
         return self.changed, self.added, self.removed
 
     @LazyProperty
     def tags(self):
         _tags = []
         for tname, tsha in self.repository.tags.iteritems():
             if tsha == self.raw_id:
                 _tags.append(tname)
         return _tags
 
     @LazyProperty
     def branch(self):
 
         heads = self.repository._heads(reverse=False)
 
         ref = heads.get(self.raw_id)
         if ref:
             return safe_unicode(ref)
 
     def _fix_path(self, path):
         """
         Paths are stored without trailing slash so we need to get rid off it if
         needed.
         """
         if path.endswith('/'):
             path = path.rstrip('/')
         return path
 
     def _get_id_for_path(self, path):
         path = safe_str(path)
         # FIXME: Please, spare a couple of minutes and make those codes cleaner;
         if not path in self._paths:
             path = path.strip('/')
             # set root tree
             tree = self.repository._repo[self._tree_id]
             if path == '':
                 self._paths[''] = tree.id
                 return tree.id
             splitted = path.split('/')
             dirs, name = splitted[:-1], splitted[-1]
             curdir = ''
 
             # initially extract things from root dir
             for item, stat, id in tree.iteritems():
                 if curdir:
                     name = '/'.join((curdir, item))
                 else:
                     name = item
                 self._paths[name] = id
                 self._stat_modes[name] = stat
 
             for dir in dirs:
                 if curdir:
                     curdir = '/'.join((curdir, dir))
                 else:
                     curdir = dir
                 dir_id = None
                 for item, stat, id in tree.iteritems():
                     if dir == item:
                         dir_id = id
                 if dir_id:
                     # Update tree
                     tree = self.repository._repo[dir_id]
                     if not isinstance(tree, objects.Tree):
                         raise ChangesetError('%s is not a directory' % curdir)
                 else:
                     raise ChangesetError('%s have not been found' % curdir)
 
                 # cache all items from the given traversed tree
                 for item, stat, id in tree.iteritems():
                     if curdir:
                         name = '/'.join((curdir, item))
                     else:
                         name = item
                     self._paths[name] = id
                     self._stat_modes[name] = stat
             if not path in self._paths:
                 raise NodeDoesNotExistError("There is no file nor directory "
                     "at the given path '%s' at revision %s"
                     % (path, safe_str(self.short_id)))
         return self._paths[path]
 
     def _get_kind(self, path):
         obj = self.repository._repo[self._get_id_for_path(path)]
         if isinstance(obj, objects.Blob):
             return NodeKind.FILE
         elif isinstance(obj, objects.Tree):
             return NodeKind.DIR
 
     def _get_filectx(self, path):
         path = self._fix_path(path)
         if self._get_kind(path) != NodeKind.FILE:
             raise ChangesetError("File does not exist for revision %s at "
                 " '%s'" % (self.raw_id, path))
         return path
 
     def _get_file_nodes(self):
         return chain(*(t[2] for t in self.walk()))
 
     @LazyProperty
     def parents(self):
         """
         Returns list of parents changesets.
         """
         return [self.repository.get_changeset(parent)
                 for parent in self._commit.parents]
 
     @LazyProperty
     def children(self):
         """
         Returns list of children changesets.
         """
         rev_filter = _git_path = settings.GIT_REV_FILTER
         so, se = self.repository.run_git_command(
             "rev-list %s --children | grep '^%s'" % (rev_filter, self.raw_id)
         )
 
         children = []
         for l in so.splitlines():
             childs = l.split(' ')[1:]
             children.extend(childs)
         return [self.repository.get_changeset(cs) for cs in children]
 
     def next(self, branch=None):
 
         if branch and self.branch != branch:
             raise VCSError('Branch option used on changeset not belonging '
                            'to that branch')
 
         def _next(changeset, branch):
             try:
                 next_ = changeset.revision + 1
                 next_rev = changeset.repository.revisions[next_]
             except IndexError:
                 raise ChangesetDoesNotExistError
             cs = changeset.repository.get_changeset(next_rev)
 
             if branch and branch != cs.branch:
                 return _next(cs, branch)
 
             return cs
 
         return _next(self, branch)
 
     def prev(self, branch=None):
         if branch and self.branch != branch:
             raise VCSError('Branch option used on changeset not belonging '
                            'to that branch')
 
         def _prev(changeset, branch):
             try:
                 prev_ = changeset.revision - 1
                 if prev_ < 0:
                     raise IndexError
                 prev_rev = changeset.repository.revisions[prev_]
             except IndexError:
                 raise ChangesetDoesNotExistError
 
             cs = changeset.repository.get_changeset(prev_rev)
 
             if branch and branch != cs.branch:
                 return _prev(cs, branch)
 
             return cs
 
         return _prev(self, branch)
 
     def diff(self, ignore_whitespace=True, context=3):
         rev1 = self.parents[0] if self.parents else self.repository.EMPTY_CHANGESET
         rev2 = self
         return ''.join(self.repository.get_diff(rev1, rev2,
                                     ignore_whitespace=ignore_whitespace,
                                     context=context))
 
     def get_file_mode(self, path):
         """
         Returns stat mode of the file at the given ``path``.
         """
         # ensure path is traversed
         path = safe_str(path)
         self._get_id_for_path(path)
         return self._stat_modes[path]
 
     def get_file_content(self, path):
         """
         Returns content of the file at given ``path``.
         """
         id = self._get_id_for_path(path)
         blob = self.repository._repo[id]
         return blob.as_pretty_string()
 
     def get_file_size(self, path):
         """
         Returns size of the file at given ``path``.
         """
         id = self._get_id_for_path(path)
         blob = self.repository._repo[id]
         return blob.raw_length()
 
     def get_file_changeset(self, path):
         """
         Returns last commit of the file at the given ``path``.
         """
         return self.get_file_history(path, limit=1)[0]
 
     def get_file_history(self, path, limit=None):
         """
         Returns history of file as reversed list of ``Changeset`` objects for
         which file at given ``path`` has been modified.
 
         TODO: This function now uses os underlying 'git' and 'grep' commands
         which is generally not good. Should be replaced with algorithm
         iterating commits.
         """
         self._get_filectx(path)
         cs_id = safe_str(self.id)
         f_path = safe_str(path)
 
         if limit:
-            cmd = 'log -n %s --pretty="format: %%H" -s -p %s -- "%s"' % (
-                      safe_int(limit, 0), cs_id, f_path
-                   )
+            cmd = 'log -n %s --pretty="format: %%H" -s %s -- "%s"' % (
+                      safe_int(limit, 0), cs_id, f_path)
 
         else:
-            cmd = 'log --pretty="format: %%H" -s -p %s -- "%s"' % (
-                      cs_id, f_path
-                   )
+            cmd = 'log --pretty="format: %%H" -s %s -- "%s"' % (
+                      cs_id, f_path)
         so, se = self.repository.run_git_command(cmd)
         ids = re.findall(r'[0-9a-fA-F]{40}', so)
-        return [self.repository.get_changeset(id) for id in ids]
+        return [self.repository.get_changeset(sha) for sha in ids]
 
     def get_file_history_2(self, path):
         """
         Returns history of file as reversed list of ``Changeset`` objects for
         which file at given ``path`` has been modified.
 
         """
         self._get_filectx(path)
         from dulwich.walk import Walker
         include = [self.id]
         walker = Walker(self.repository._repo.object_store, include,
                         paths=[path], max_entries=1)
         return [self.repository.get_changeset(sha)
                 for sha in (x.commit.id for x in walker)]
 
     def get_file_annotate(self, path):
         """
         Returns a generator of four element tuples with
             lineno, sha, changeset lazy loader and line
 
         TODO: This function now uses os underlying 'git' command which is
         generally not good. Should be replaced with algorithm iterating
         commits.
         """
         cmd = 'blame -l --root -r %s -- "%s"' % (self.id, path)
         # -l     ==> outputs long shas (and we need all 40 characters)
         # --root ==> doesn't put '^' character for bounderies
         # -r sha ==> blames for the given revision
         so, se = self.repository.run_git_command(cmd)
 
         for i, blame_line in enumerate(so.split('\n')[:-1]):
             ln_no = i + 1
             sha, line = re.split(r' ', blame_line, 1)
             yield (ln_no, sha, lambda: self.repository.get_changeset(sha), line)
 
     def fill_archive(self, stream=None, kind='tgz', prefix=None,
                      subrepos=False):
         """
         Fills up given stream.
 
         :param stream: file like object.
         :param kind: one of following: ``zip``, ``tgz`` or ``tbz2``.
             Default: ``tgz``.
         :param prefix: name of root directory in archive.
             Default is repository name and changeset's raw_id joined with dash
             (``repo-tip.<KIND>``).
         :param subrepos: include subrepos in this archive.
 
         :raise ImproperArchiveTypeError: If given kind is wrong.
         :raise VcsError: If given stream is None
 
         """
         allowed_kinds = settings.ARCHIVE_SPECS.keys()
         if kind not in allowed_kinds:
             raise ImproperArchiveTypeError('Archive kind not supported use one'
                 'of %s', allowed_kinds)
 
         if prefix is None:
             prefix = '%s-%s' % (self.repository.name, self.short_id)
         elif prefix.startswith('/'):
             raise VCSError("Prefix cannot start with leading slash")
         elif prefix.strip() == '':
             raise VCSError("Prefix cannot be empty")
 
         if kind == 'zip':
             frmt = 'zip'
         else:
             frmt = 'tar'
         _git_path = settings.GIT_EXECUTABLE_PATH
         cmd = '%s archive --format=%s --prefix=%s/ %s' % (_git_path,
                                                 frmt, prefix, self.raw_id)
         if kind == 'tgz':
             cmd += ' | gzip -9'
         elif kind == 'tbz2':
             cmd += ' | bzip2 -9'
 
         if stream is None:
             raise VCSError('You need to pass in a valid stream for filling'
                            ' with archival data')
         popen = Popen(cmd, stdout=PIPE, stderr=PIPE, shell=True,
             cwd=self.repository.path)
 
         buffer_size = 1024 * 8
         chunk = popen.stdout.read(buffer_size)
         while chunk:
             stream.write(chunk)
             chunk = popen.stdout.read(buffer_size)
         # Make sure all descriptors would be read
         popen.communicate()
 
     def get_nodes(self, path):
         if self._get_kind(path) != NodeKind.DIR:
             raise ChangesetError("Directory does not exist for revision %s at "
                 " '%s'" % (self.revision, path))
         path = self._fix_path(path)
         id = self._get_id_for_path(path)
         tree = self.repository._repo[id]
         dirnodes = []
         filenodes = []
         als = self.repository.alias
         for name, stat, id in tree.iteritems():
             if objects.S_ISGITLINK(stat):
                 dirnodes.append(SubModuleNode(name, url=None, changeset=id,
                                               alias=als))
                 continue
 
             obj = self.repository._repo.get_object(id)
             if path != '':
                 obj_path = '/'.join((path, name))
             else:
                 obj_path = name
             if obj_path not in self._stat_modes:
                 self._stat_modes[obj_path] = stat
             if isinstance(obj, objects.Tree):
                 dirnodes.append(DirNode(obj_path, changeset=self))
             elif isinstance(obj, objects.Blob):
                 filenodes.append(FileNode(obj_path, changeset=self, mode=stat))
             else:
                 raise ChangesetError("Requested object should be Tree "
                                      "or Blob, is %r" % type(obj))
         nodes = dirnodes + filenodes
         for node in nodes:
             if not node.path in self.nodes:
                 self.nodes[node.path] = node
         nodes.sort()
         return nodes
 
     def get_node(self, path):
         if isinstance(path, unicode):
             path = path.encode('utf-8')
         path = self._fix_path(path)
         if not path in self.nodes:
             try:
                 id_ = self._get_id_for_path(path)
             except ChangesetError:
                 raise NodeDoesNotExistError("Cannot find one of parents' "
                     "directories for a given path: %s" % path)
 
             _GL = lambda m: m and objects.S_ISGITLINK(m)
             if _GL(self._stat_modes.get(path)):
                 node = SubModuleNode(path, url=None, changeset=id_,
                                      alias=self.repository.alias)
             else:
                 obj = self.repository._repo.get_object(id_)
 
                 if isinstance(obj, objects.Tree):
                     if path == '':
                         node = RootNode(changeset=self)
                     else:
                         node = DirNode(path, changeset=self)
                     node._tree = obj
                 elif isinstance(obj, objects.Blob):
                     node = FileNode(path, changeset=self)
                     node._blob = obj
                 else:
                     raise NodeDoesNotExistError("There is no file nor directory "
                         "at the given path '%s' at revision %s"
                         % (path, self.short_id))
             # cache node
             self.nodes[path] = node
         return self.nodes[path]
 
     @LazyProperty
     def affected_files(self):
         """
         Gets a fast accessible file changes for given changeset
         """
         added, modified, deleted = self._changes_cache
         return list(added.union(modified).union(deleted))
 
     @LazyProperty
     def _diff_name_status(self):
         output = []
         for parent in self.parents:
             cmd = 'diff --name-status %s %s --encoding=utf8' % (parent.raw_id,
                                                                 self.raw_id)
             so, se = self.repository.run_git_command(cmd)
             output.append(so.strip())
         return '\n'.join(output)
 
     @LazyProperty
     def _changes_cache(self):
         added = set()
         modified = set()
         deleted = set()
         _r = self.repository._repo
 
         parents = self.parents
         if not self.parents:
             parents = [EmptyChangeset()]
         for parent in parents:
             if isinstance(parent, EmptyChangeset):
                 oid = None
             else:
                 oid = _r[parent.raw_id].tree
             changes = _r.object_store.tree_changes(oid, _r[self.raw_id].tree)
             for (oldpath, newpath), (_, _), (_, _) in changes:
                 if newpath and oldpath:
                     modified.add(newpath)
                 elif newpath and not oldpath:
                     added.add(newpath)
                 elif not newpath and oldpath:
                     deleted.add(oldpath)
         return added, modified, deleted
 
     def _get_paths_for_status(self, status):
         """
         Returns sorted list of paths for given ``status``.
 
         :param status: one of: *added*, *modified* or *deleted*
         """
         added, modified, deleted = self._changes_cache
         return sorted({
             'added': list(added),
             'modified': list(modified),
             'deleted': list(deleted)}[status]
         )
 
     @LazyProperty
     def added(self):
         """
         Returns list of added ``FileNode`` objects.
         """
         if not self.parents:
             return list(self._get_file_nodes())
         return AddedFileNodesGenerator([n for n in
                                 self._get_paths_for_status('added')], self)
 
     @LazyProperty
     def changed(self):
         """
         Returns list of modified ``FileNode`` objects.
         """
         if not self.parents:
             return []
         return ChangedFileNodesGenerator([n for n in
                                 self._get_paths_for_status('modified')], self)
 
     @LazyProperty
     def removed(self):
         """
         Returns list of removed ``FileNode`` objects.
         """
         if not self.parents:
             return []
         return RemovedFileNodesGenerator([n for n in
                                 self._get_paths_for_status('deleted')], self)