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(

                    % (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)

    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:

        else:

        so, se = self.repository.run_git_command(cmd)

        ids = re.findall(r'[0-9a-fA-F]{40}', so)

    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)