'index does not exist')

        else:

            self.initial = False

    def _get_index_revision(self, repo):

        db_repo = Repository.get_by_repo_name(repo.name_unicode)

        landing_rev = 'tip'

        if db_repo:

            _rev_type, _rev = db_repo.landing_rev

            landing_rev = _rev

        return landing_rev

    def _get_index_changeset(self, repo, index_rev=None):

        if not index_rev:

            index_rev = self._get_index_revision(repo)

        cs = repo.get_changeset(index_rev)

        return cs

    def get_paths(self, repo):

        """

        recursive walk in root dir and return a set of all path in that dir

        based on repository walk function

        """

        index_paths_ = set()

        try:

            cs = self._get_index_changeset(repo)

            for _topnode, _dirs, files in cs.walk('/'):

                for f in files:

                    index_paths_.add(os.path.join(safe_str(repo.path), safe_str(f.path)))

        except RepositoryError:

            log.debug(traceback.format_exc())

            pass

        return index_paths_

    def get_node(self, repo, path, index_rev=None):

        """

        gets a filenode based on given full path. It operates on string for

        hg git compatibility.

        :param repo: scm repo instance

        :param path: full path including root location

        :return: FileNode

        """

        # FIXME: paths should be normalized ... or even better: don't include repo.path

        path = safe_str(path)

        repo_path = safe_str(repo.path)

        assert path.startswith(repo_path)

        assert path[len(repo_path)] in (os.path.sep, os.path.altsep)

        node_path = path[len(repo_path) + 1:]

        cs = self._get_index_changeset(repo, index_rev=index_rev)

        node = cs.get_node(node_path)

        return node

    def is_indexable_node(self, node):

        """

        Just index the content of chosen files, skipping binary files

        """

        return (node.extension in INDEX_EXTENSIONS or node.name in INDEX_FILENAMES) and \

               not node.is_binary

    def get_node_mtime(self, node):

        return mktime(node.last_changeset.date.timetuple())

    def add_doc(self, writer, path, repo, repo_name, index_rev=None):

        """

        Adding doc to writer this function itself fetches data from

        the instance of vcs backend

        """

        try:

            node = self.get_node(repo, path, index_rev)

        except (ChangesetError, NodeDoesNotExistError):

            log.debug("couldn't add doc - %s did not have %r at %s", repo, path, index_rev)

            return 0, 0

        indexed = indexed_w_content = 0

        if self.is_indexable_node(node):

            u_content = node.content

            if not isinstance(u_content, unicode):

                log.warning('  >> %s Could not get this content as unicode '

                            'replacing with empty content' % path)

                u_content = u''

            else:

                log.debug('    >> %s [WITH CONTENT]', path)

                indexed_w_content += 1

        else:

            log.debug('    >> %s', path)

            # just index file name without it's content

            u_content = u''

            indexed += 1

        p = safe_unicode(path)

        writer.add_document(

            fileid=p,

            owner=unicode(repo.contact),

            repository=safe_unicode(repo_name),

            path=p,

            content=u_content,

            modtime=self.get_node_mtime(node),

            extension=node.extension

        )

        return indexed, indexed_w_content

    def index_changesets(self, writer, repo_name, repo, start_rev=None):

        """

        Add all changeset in the vcs repo starting at start_rev

        to the index writer

        :param writer: the whoosh index writer to add to

        :param repo_name: name of the repository from whence the

          changeset originates including the repository group

        :param repo: the vcs repository instance to index changesets for,

          the presumption is the repo has changesets to index

        :param start_rev=None: the full sha id to start indexing from

          if start_rev is None then index from the first changeset in

          the repo

        """

        if start_rev is None:

            start_rev = repo[0].raw_id

        log.debug('indexing changesets in %s starting at rev: %s',

                  repo_name, start_rev)

        indexed = 0

        cs_iter = repo.get_changesets(start=start_rev)

        total = len(cs_iter)

        for cs in cs_iter:

            log.debug('    >> %s/%s', cs, total)

            writer.add_document(

                raw_id=unicode(cs.raw_id),

                owner=unicode(repo.contact),

                date=cs._timestamp,

                repository=safe_unicode(repo_name),

                author=cs.author,

                message=cs.message,

                last=cs.last,

                added=u' '.join([safe_unicode(node.path) for node in cs.added]).lower(),

                removed=u' '.join([safe_unicode(node.path) for node in cs.removed]).lower(),

                changed=u' '.join([safe_unicode(node.path) for node in cs.changed]).lower(),

                parents=u' '.join([cs.raw_id for cs in cs.parents]),

            )

            indexed += 1

        log.debug('indexed %d changesets for repo %s', indexed, repo_name)

        return indexed

    def index_files(self, file_idx_writer, repo_name, repo):

        """

        Index files for given repo_name

        :param file_idx_writer: the whoosh index writer to add to

        :param repo_name: name of the repository we're indexing

        :param repo: instance of vcs repo

        """

        i_cnt = iwc_cnt = 0

        log.debug('building index for %s @revision:%s', repo.path,

                                                self._get_index_revision(repo))

        index_rev = self._get_index_revision(repo)

        for idx_path in self.get_paths(repo):

            i, iwc = self.add_doc(file_idx_writer, idx_path, repo, repo_name, index_rev)

            i_cnt += i

            iwc_cnt += iwc

        log.debug('added %s files %s with content for repo %s',

                  i_cnt + iwc_cnt, iwc_cnt, repo.path)

        return i_cnt, iwc_cnt

    def update_changeset_index(self):

        idx = open_dir(self.index_location, indexname=CHGSET_IDX_NAME)

        with idx.searcher() as searcher:

            writer = idx.writer()

            writer_is_dirty = False

            try:

                indexed_total = 0

                repo_name = None

                for repo_name, repo in self.repo_paths.items():

                    # skip indexing if there aren't any revs in the repo

                    num_of_revs = len(repo)

                    if num_of_revs < 1:

                        continue

                    qp = QueryParser('repository', schema=CHGSETS_SCHEMA)

                    q = qp.parse(u"last:t AND %s" % repo_name)

                    results = searcher.search(q)

                    # default to scanning the entire repo

                    last_rev = 0

                    start_id = None

                    if len(results) > 0:

                        # assuming that there is only one result, if not this

                        # may require a full re-index.

                        start_id = results[0]['raw_id']

                        last_rev = repo.get_changeset(revision=start_id).revision

                    # there are new changesets to index or a new repo to index

                    if last_rev == 0 or num_of_revs > last_rev + 1:

                        # delete the docs in the index for the previous

                        # last changeset(s)

                        for hit in results:

                            q = qp.parse(u"last:t AND %s AND raw_id:%s" %

                                            (repo_name, hit['raw_id']))

                            writer.delete_by_query(q)

                        # index from the previous last changeset + all new ones

                        indexed_total += self.index_changesets(writer,

                                                repo_name, repo, start_id)

                        writer_is_dirty = True

                log.debug('indexed %s changesets for repo %s',

                          indexed_total, repo_name

                )

            finally:

                if writer_is_dirty:

                    log.debug('>> COMMITING CHANGES TO CHANGESET INDEX<<')

                    writer.commit(merge=True)

                    log.debug('>>> FINISHED REBUILDING CHANGESET INDEX <<<')

                else:

                    log.debug('>> NOTHING TO COMMIT TO CHANGESET INDEX<<')

    def update_file_index(self):

        log.debug((u'STARTING INCREMENTAL INDEXING UPDATE FOR EXTENSIONS %s '

                   'AND REPOS %s') % (INDEX_EXTENSIONS, self.repo_paths.keys()))

        idx = open_dir(self.index_location, indexname=self.indexname)

        # The set of all paths in the index

        indexed_paths = set()

    ~~~~~~~~~~~~~~~~~

    Base for all available scm backends

    :created_on: Apr 8, 2010

    :copyright: (c) 2010-2011 by Marcin Kuzminski, Lukasz Balcerzak.

"""

import datetime

import itertools

from kallithea.lib.vcs.utils import author_name, author_email, safe_unicode

from kallithea.lib.vcs.utils.lazy import LazyProperty

from kallithea.lib.vcs.utils.helpers import get_dict_for_attrs

from kallithea.lib.vcs.conf import settings

from kallithea.lib.vcs.exceptions import (

    ChangesetError, EmptyRepositoryError, NodeAlreadyAddedError,

    NodeAlreadyChangedError, NodeAlreadyExistsError, NodeAlreadyRemovedError,

    NodeDoesNotExistError, NodeNotChangedError, RepositoryError

)

class BaseRepository(object):

    """

    Base Repository for final backends

    **Attributes**

        ``DEFAULT_BRANCH_NAME``

            name of default branch (i.e. "trunk" for svn, "master" for git etc.

        ``scm``

            alias of scm, i.e. *git* or *hg*

        ``repo``

            object from external api

        ``revisions``

            list of all available revisions' ids, in ascending order

        ``changesets``

            storage dict caching returned changesets

        ``path``

            absolute path to the repository

        ``branches``

            branches as list of changesets

        ``tags``

            tags as list of changesets

    """

    scm = None

    DEFAULT_BRANCH_NAME = None

    EMPTY_CHANGESET = '0' * 40

    def __init__(self, repo_path, create=False, **kwargs):

        """

        Initializes repository. Raises RepositoryError if repository could

        not be find at the given ``repo_path`` or directory at ``repo_path``

        exists and ``create`` is set to True.

        :param repo_path: local path of the repository

        :param create=False: if set to True, would try to create repository.

        :param src_url=None: if set, should be proper url from which repository

          would be cloned; requires ``create`` parameter to be set to True -

          raises RepositoryError if src_url is set and create evaluates to

          False

        """

        raise NotImplementedError

    def __str__(self):

        return '<%s at %s>' % (self.__class__.__name__, self.path)

    def __repr__(self):

        return self.__str__()

    def __len__(self):

        return self.count()

    def __eq__(self, other):

        same_instance = isinstance(other, self.__class__)

        return same_instance and getattr(other, 'path', None) == self.path

    def __ne__(self, other):

        return not self.__eq__(other)

    @LazyProperty

    def alias(self):

        for k, v in settings.BACKENDS.items():

            if v.split('.')[-1] == str(self.__class__.__name__):

                return k

    @LazyProperty

    def name(self):

        raise NotImplementedError

    @property

    def name_unicode(self):

        return safe_unicode(self.name)

    @LazyProperty

    def owner(self):

        raise NotImplementedError

    @LazyProperty

    def description(self):

        raise NotImplementedError

    @LazyProperty

    def size(self):

        """

        Returns combined size in bytes for all repository files

        """

        size = 0

        try:

            tip = self.get_changeset()

            for topnode, dirs, files in tip.walk('/'):

                for f in files:

                    size += tip.get_file_size(f.path)

        except RepositoryError as e:

            pass

        return size

    def is_valid(self):

        """

        Validates repository.

        """

        raise NotImplementedError

    def is_empty(self):

        return self._empty

    #==========================================================================

    # CHANGESETS

    #==========================================================================

    def get_changeset(self, revision=None):

        """

        Returns instance of ``Changeset`` class. If ``revision`` is None, most

        recent changeset is returned.

        :raises ``EmptyRepositoryError``: if there are no revisions

        """

        raise NotImplementedError

    def __iter__(self):

        """

        Allows Repository objects to be iterated.

        *Requires* implementation of ``__getitem__`` method.

        """

        for revision in self.revisions:

            yield self.get_changeset(revision)

    def get_changesets(self, start=None, end=None, start_date=None,

                       end_date=None, branch_name=None, reverse=False):

        """

        Returns iterator of ``MercurialChangeset`` objects from start to end

        not inclusive This should behave just like a list, ie. end is not

        inclusive

        :param start: None or str

        :param end: None or str

        :param start_date:

        :param end_date:

        :param branch_name:

        :param reversed:

        """

        raise NotImplementedError

    def __getslice__(self, i, j):

        """

        Returns a iterator of sliced repository

        """

        for rev in self.revisions[i:j]:

            yield self.get_changeset(rev)

    def __getitem__(self, key):

        return self.get_changeset(key)

    def count(self):

        return len(self.revisions)

    def tag(self, name, user, revision=None, message=None, date=None, **opts):

        """

        Creates and returns a tag for the given ``revision``.

        :param name: name for new tag

                                 author='this is it <this-is-it@foo.bar.com>',

                                 vcs_type='hg',

                                 parent=prev,

                                 newfile=True)

repos = [

    # reponame,              init func or fork base, groupname

    (u'indexing_test',       init_indexing_test,     None),

    (u'indexing_test-fork',  u'indexing_test',       None),

    (u'group/indexing_test', u'indexing_test',       u'group'),

    (u'this-is-it',          u'indexing_test',       None),

    (u'indexing_test-foo',   u'indexing_test',       None),

    (u'indexing_test-FOO',   u'indexing_test',       None),

    (u'stopword_test',       init_stopword_test,     None),

]

# map: name => id

repoids = {}

groupids = {}

def rebuild_index(full_index):

    with mock.patch('kallithea.lib.indexers.daemon.log.debug',

                    lambda *args, **kwargs: None):

        # The more revisions managed repositories have, the more

        # memory capturing "log.debug()" output in "indexers.daemon"

        # requires. This may cause unintentional failure of subsequent

        # tests, if ENOMEM at forking "git" prevents from rebuilding

        # index for search.

        # Therefore, "log.debug()" is disabled regardless of logging

        # level while rebuilding index.

        # (FYI, ENOMEM occurs at forking "git" with python 2.7.3,

        # Linux 3.2.78-1 x86_64, 3GB memory, and no ulimit

        # configuration for memory)

        create_test_index(TESTS_TMP_PATH, CONFIG, full_index=full_index)

class TestSearchControllerIndexing(TestController):

    @classmethod

    def setup_class(cls):

        for reponame, init_or_fork, groupname in repos:

            if groupname and groupname not in groupids:

                group = fixture.create_repo_group(groupname)

                groupids[groupname] = group.group_id

            if callable(init_or_fork):

                repo = fixture.create_repo(reponame,

                                           repo_group=groupname)

                init_or_fork(repo)

            else:

                repo = fixture.create_fork(init_or_fork, reponame,

                                           repo_group=groupname)

            repoids[reponame] = repo.repo_id

        # treat "it" as indexable filename

        filenames_mock = list(INDEX_FILENAMES)

        filenames_mock.append('it')

        with mock.patch('kallithea.lib.indexers.daemon.INDEX_FILENAMES',

                        filenames_mock):

            rebuild_index(full_index=False) # only for newly added repos

    @classmethod

    def teardown_class(cls):

        # delete in reversed order, to delete fork destination at first

        for reponame, init_or_fork, groupname in reversed(repos):

            RepoModel().delete(repoids[reponame])

        for reponame, init_or_fork, groupname in reversed(repos):

            if groupname in groupids:

                RepoGroupModel().delete(groupids.pop(groupname),

                                        force_delete=True)

        Session().commit()

        Session.remove()

        rebuild_index(full_index=True) # rebuild fully for subsequent tests

    @parametrize('reponame', [

        (u'indexing_test'),

        (u'indexing_test-fork'),

        (u'group/indexing_test'),

        (u'this-is-it'),

        (u'*-fork'),

        (u'group/*'),

    ])

    @parametrize('searchtype,query,hit', [

        ('content', 'this_should_be_unique_content', 1),

        ('commit', 'this_should_be_unique_commit_log', 1),

        ('path', 'this_should_be_unique_filename.txt', 1),

    ])

    def test_repository_tokenization(self, reponame, searchtype, query, hit):

        self.log_user()

        q = 'repository:%s %s' % (reponame, query)

        response = self.app.get(url(controller='search', action='index'),

                                {'q': q, 'type': searchtype})

        response.mustcontain('>%d results' % hit)

    @parametrize('searchtype,query,hit', [

        ('content', 'this_should_be_unique_content', 1),

        ('commit', 'this_should_be_unique_commit_log', 1),

        ('path', 'this_should_be_unique_filename.txt', 1),

    ])

    def test_repository_case_sensitivity(self, searchtype, query, hit):

        self.log_user()

        lname = u'indexing_test-foo'

        uname = u'indexing_test-FOO'

        # (1) "repository:REPONAME" condition should match against

        # repositories case-insensitively

        q = 'repository:%s %s' % (lname, query)

        response = self.app.get(url(controller='search', action='index'),

                                {'q': q, 'type': searchtype})

        response.mustcontain('>%d results' % (hit * 2))

        # (2) on the other hand, searching under the specific

        # repository should return results only for that repository,

        # even if specified name matches against another repository

        # case-insensitively.

        response = self.app.get(url(controller='search', action='index',

                                    repo_name=uname),

                                {'q': query, 'type': searchtype})

        response.mustcontain('>%d results' % hit)

        # confirm that there is no matching against lower name repository

        assert uname in response

        assert lname not in response

    @parametrize('searchtype,query,hit', [

        ('content', 'path:this/is/it def test', 1),

        ('commit', 'added:this/is/it bother to ask where', 1),

        # this condition matches against files below, because

        # "path:" condition is also applied on "repository path".

        # - "this/is/it" in "stopword_test" repo

        # - "this_should_be_unique_filename.txt" in "this-is-it" repo

        ('path', 'this/is/it', 2),

        ('content', 'extension:us', 1),

        ('path', 'extension:us', 1),

    ])

    def test_filename_stopword(self, searchtype, query, hit):

        response = self.app.get(url(controller='search', action='index'),

                                {'q': query, 'type': searchtype})

        response.mustcontain('>%d results' % hit)

    @parametrize('searchtype,query,hit', [

        # matching against both 2 files

        ('content', 'owner:"this is it"', 0),

        ('content', 'owner:this-is-it', 0),

        ('path', 'owner:"this is it"', 0),

        ('path', 'owner:this-is-it', 0),

        # matching against both 2 revisions

        ('commit', 'owner:"this is it"', 0),

        ('commit', 'owner:"this-is-it"', 0),

        # matching against only 1 revision

        ('commit', 'author:"this is it"', 1),

        ('commit', 'author:"this-is-it"', 1),

    ])

    def test_mailaddr_stopword(self, searchtype, query, hit):

        response = self.app.get(url(controller='search', action='index'),

                                {'q': query, 'type': searchtype})

        response.mustcontain('>%d results' % hit)