.. _git_support:

===========

GIT support

===========

Although There are some limitations on git usage.

- large pushes needs http server with chunked encoding support.

if you plan to use git you need to run RhodeCode with some

http server that supports chunked encoding which git http protocol uses, 

i recommend using waitress_ or gunicorn_ (linux only) for `paste` wsgi app 

replacement.

    use = egg:Paste#http

    use = egg:waitress#main

And comment out bellow options::

    threadpool_workers = 

    threadpool_max_requests = 

    use_threadpool = 

You can simply run `paster serve` as usual.

You can always disable git/hg support by editing a 

file **rhodecode/__init__.py** and commenting out backends

.. code-block:: python

   BACKENDS = {

       'hg': 'Mercurial repository',

       #'git': 'Git repository',

   }

.. _waitress: http://pypi.python.org/pypi/waitress

.. _gunicorn: http://pypi.python.org/pypi/gunicorn

# -*- coding: utf-8 -*-

"""

    vcs.backends.git

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

    Git backend implementation.

    :created_on: Apr 8, 2010

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

"""

import os

import re

import time

import posixpath

from dulwich.repo import Repo, NotGitRepository

#from dulwich.config import ConfigFile

from string import Template

from subprocess import Popen, PIPE

from rhodecode.lib.vcs.backends.base import BaseRepository

from rhodecode.lib.vcs.exceptions import BranchDoesNotExistError

from rhodecode.lib.vcs.exceptions import ChangesetDoesNotExistError

from rhodecode.lib.vcs.exceptions import EmptyRepositoryError

from rhodecode.lib.vcs.exceptions import RepositoryError

from rhodecode.lib.vcs.exceptions import TagAlreadyExistError

from rhodecode.lib.vcs.exceptions import TagDoesNotExistError

from rhodecode.lib.vcs.utils import safe_unicode, makedate, date_fromtimestamp

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

from rhodecode.lib.vcs.utils.ordered_dict import OrderedDict

from rhodecode.lib.vcs.utils.paths import abspath

from rhodecode.lib.vcs.utils.paths import get_user_home

from .workdir import GitWorkdir

from .changeset import GitChangeset

from .inmemory import GitInMemoryChangeset

from .config import ConfigFile

class GitRepository(BaseRepository):

    """

    Git repository backend.

    """

    DEFAULT_BRANCH_NAME = 'master'

    scm = 'git'

    def __init__(self, repo_path, create=False, src_url=None,

                 update_after_clone=False, bare=False):

        self.path = abspath(repo_path)

        self._repo = self._get_repo(create, src_url, update_after_clone, bare)

        #temporary set that to now at later we will move it to constructor

        baseui = None

        if baseui is None:

            from mercurial.ui import ui

            baseui = ui()

        # patch the instance of GitRepo with an "FAKE" ui object to add

        # compatibility layer with Mercurial

        setattr(self._repo, 'ui', baseui)

        try:

            self.head = self._repo.head()

        except KeyError:

            self.head = None

        self._config_files = [

            bare and abspath(self.path, 'config') or abspath(self.path, '.git',

                'config'),

            abspath(get_user_home(), '.gitconfig'),

        ]

        self.bare = self._repo.bare

    @LazyProperty

    def revisions(self):

        """

        Returns list of revisions' ids, in ascending order.  Being lazy

        attribute allows external tools to inject shas from cache.

        """

        return self._get_all_revisions()

    def run_git_command(self, cmd):

        """

        Runs given ``cmd`` as git command and returns tuple

        (returncode, stdout, stderr).

        .. note::

           This method exists only until log/blame functionality is implemented

           at Dulwich (see https://bugs.launchpad.net/bugs/645142). Parsing

           os command's output is road to hell...

        :param cmd: git command to be executed

        """

        _copts = ['-c', 'core.quotepath=false', ]

        _str_cmd = False

        if isinstance(cmd, basestring):

            cmd = [cmd]

            _str_cmd = True

        if _str_cmd:

            cmd = ' '.join(cmd)

        try:

            opts = dict(

                shell=isinstance(cmd, basestring),

                stdout=PIPE,

            if os.path.isdir(self.path):

                opts['cwd'] = self.path

            p = Popen(cmd, **opts)

        except OSError, err:

            raise RepositoryError("Couldn't run git command (%s).\n"

                "Original error was:%s" % (cmd, err))

        so, se = p.communicate()

        if not se.startswith("fatal: bad default revision 'HEAD'") and \

            p.returncode != 0:

            raise RepositoryError("Couldn't run git command (%s).\n"

                "stderr:\n%s" % (cmd, se))

        return so, se

    def _check_url(self, url):

        """

        Functon will check given url and try to verify if it's a valid

        link. Sometimes it may happened that mercurial will issue basic

        auth request that can cause whole API to hang when used from python

        or other external calls.

        On failures it'll raise urllib2.HTTPError

        """

        #TODO: implement this

        pass

    def _get_repo(self, create, src_url=None, update_after_clone=False,

            bare=False):

        if create and os.path.exists(self.path):

            raise RepositoryError("Location already exist")

        if src_url and not create:

            raise RepositoryError("Create should be set to True if src_url is "

                                  "given (clone operation creates repository)")

        try:

            if create and src_url:

                self._check_url(src_url)

                self.clone(src_url, update_after_clone, bare)

                return Repo(self.path)

            elif create:

                os.mkdir(self.path)

                if bare:

                    return Repo.init_bare(self.path)

                else:

                    return Repo.init(self.path)

            else:

                return Repo(self.path)

        except (NotGitRepository, OSError), err:

            raise RepositoryError(err)

    def _get_all_revisions(self):

        cmd = 'rev-list --all --date-order'

        try:

            so, se = self.run_git_command(cmd)

        except RepositoryError:

            # Can be raised for empty repositories

            return []

        revisions = so.splitlines()

        revisions.reverse()

        return revisions

    def _get_revision(self, revision):

        """

        For git backend we always return integer here. This way we ensure

        that changset's revision attribute would become integer.

        """

        pattern = re.compile(r'^[[0-9a-fA-F]{12}|[0-9a-fA-F]{40}]$')

        is_bstr = lambda o: isinstance(o, (str, unicode))

        is_null = lambda o: len(o) == revision.count('0')

        if len(self.revisions) == 0:

            raise EmptyRepositoryError("There are no changesets yet")

        if revision in (None, '', 'tip', 'HEAD', 'head', -1):

            revision = self.revisions[-1]

        if ((is_bstr(revision) and revision.isdigit() and len(revision) < 12)

            or isinstance(revision, int) or is_null(revision)):

            try:

                revision = self.revisions[int(revision)]

            except:

                raise ChangesetDoesNotExistError("Revision %r does not exist "

                    "for this repository %s" % (revision, self))

        elif is_bstr(revision):

            if not pattern.match(revision) or revision not in self.revisions:

                raise ChangesetDoesNotExistError("Revision %r does not exist "

                    "for this repository %s" % (revision, self))

        # Ensure we return full id

        if not pattern.match(str(revision)):

            raise ChangesetDoesNotExistError("Given revision %r not recognized"

                % revision)

        return revision

    def _get_archives(self, archive_name='tip'):

        for i in [('zip', '.zip'), ('gz', '.tar.gz'), ('bz2', '.tar.bz2')]:

                yield {"type": i[0], "extension": i[1], "node": archive_name}

    def _get_url(self, url):

        """

        Returns normalized url. If schema is not given, would fall to

        filesystem (``file:///``) schema.

        """

        url = str(url)

        if url != 'default' and not '://' in url:

            url = ':///'.join(('file', url))

        return url

    @LazyProperty

    def name(self):

        return os.path.basename(self.path)

    @LazyProperty

    def last_change(self):

        """

        Returns last change made on this repository as datetime object

        """

        return date_fromtimestamp(self._get_mtime(), makedate()[1])

    def _get_mtime(self):

        try:

            return time.mktime(self.get_changeset().date.timetuple())

        except RepositoryError:

            idx_loc = '' if self.bare else '.git'

            # fallback to filesystem

            in_path = os.path.join(self.path, idx_loc, "index")

            he_path = os.path.join(self.path, idx_loc, "HEAD")

            if os.path.exists(in_path):

                return os.stat(in_path).st_mtime

            else:

                return os.stat(he_path).st_mtime

    @LazyProperty

    def description(self):

        idx_loc = '' if self.bare else '.git'

        undefined_description = u'unknown'

        description_path = os.path.join(self.path, idx_loc, 'description')

        if os.path.isfile(description_path):

            return safe_unicode(open(description_path).read())

        else:

            return undefined_description

    @LazyProperty

    def contact(self):

        undefined_contact = u'Unknown'

        return undefined_contact

    @property

    def branches(self):

        if not self.revisions:

            return {}

        refs = self._repo.refs.as_dict()

        sortkey = lambda ctx: ctx[0]

        _branches = [('/'.join(ref.split('/')[2:]), head)

            for ref, head in refs.items()

            if ref.startswith('refs/heads/') and not ref.endswith('/HEAD')]

        return OrderedDict(sorted(_branches, key=sortkey, reverse=False))

    def _heads(self, reverse=False):

        refs = self._repo.get_refs()

        heads = {}

        for key, val in refs.items():

            for ref_key in ['refs/heads/', 'refs/remotes/origin/']:

                if key.startswith(ref_key):

                    n = key[len(ref_key):]

                    if n not in ['HEAD']:

                        heads[n] = val

        return heads if reverse else dict((y,x) for x,y in heads.iteritems())

    def _get_tags(self):

        if not self.revisions:

            return {}

        sortkey = lambda ctx: ctx[0]

        _tags = [('/'.join(ref.split('/')[2:]), head) for ref, head in

            self._repo.get_refs().items() if ref.startswith('refs/tags/')]

        return OrderedDict(sorted(_tags, key=sortkey, reverse=True))

    @LazyProperty

    def tags(self):

        return self._get_tags()

    def tag(self, name, user, revision=None, message=None, date=None,

            **kwargs):

        """

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

        :param name: name for new tag

        :param user: full username, i.e.: "Joe Doe <joe.doe@example.com>"

        :param revision: changeset id for which new tag would be created