# -*- 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

import logging

import traceback

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

from rhodecode.lib import subprocessio

log = logging.getLogger(__name__)

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

        gitenv = os.environ

        # need to clean fix GIT_DIR !

        if 'GIT_DIR' in gitenv:

            del gitenv['GIT_DIR']

        gitenv['GIT_CONFIG_NOGLOBAL'] = '1'

        cmd = ['git'] + _copts + cmd

        if _str_cmd:

            cmd = ' '.join(cmd)

        try:

            opts = dict(

                env=gitenv,

            )

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

                opts['cwd'] = self.path

            p = subprocessio.SubprocessIOChunker(cmd, **opts)

        except (EnvironmentError, OSError), err:

            log.error(traceback.format_exc())

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

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

        so = ''.join(p)

        se = None

        return so, se

        """

        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

        """

    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.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):

        # we must check if this repo is not empty, since later command

        # fails if it is. And it's cheaper to ask than throw the subprocess

        # errors

        try:

            self._repo.head()

        except KeyError:

            return []

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

        try:

            so, se = self.run_git_command(cmd)

        except RepositoryError:

            # Can be raised for empty repositories

            return []

        return so.splitlines()

    def _get_all_revisions2(self):

        #alternate implementation using dulwich

        includes = [x[1][0] for x in self._parsed_refs.iteritems()

                    if x[1][1] != 'T']

        return [c.commit.id for c in self._repo.get_walker(include=includes)]

    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):

            # get by branch/tag name

            _ref_revision = self._parsed_refs.get(revision)

            _tags_shas = self.tags.values()

            if _ref_revision:  # and _ref_revision[1] in ['H', 'RH', 'T']:

                return _ref_revision[0]

            # maybe it's a tag ? we don't have them in self.revisions

            elif revision in _tags_shas:

                return _tags_shas[_tags_shas.index(revision)]

            elif 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):

            message = "Added tag %s for changeset %s" % (name,

                changeset.short_id)

        if date is None:

            date = datetime.datetime.now().ctime()

        try:

            self._repo.tag(name, changeset._ctx.node(), message, local, user,

                date)

        except Abort, e:

            raise RepositoryError(e.message)

        # Reinitialize tags

        self.tags = self._get_tags()

        tag_id = self.tags[name]

        return self.get_changeset(revision=tag_id)

    def remove_tag(self, name, user, message=None, date=None):

        """

        Removes tag with the given ``name``.

        :param name: name of the tag to be removed

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

        :param message: message of the tag's removal commit

        :param date: date of tag's removal commit

        :raises TagDoesNotExistError: if tag with given name does not exists

        """

        if name not in self.tags:

            raise TagDoesNotExistError("Tag %s does not exist" % name)

        if message is None:

            message = "Removed tag %s" % name

        if date is None:

            date = datetime.datetime.now().ctime()

        local = False

        try:

            self._repo.tag(name, nullid, message, local, user, date)

            self.tags = self._get_tags()

        except Abort, e:

            raise RepositoryError(e.message)

    @LazyProperty

    def bookmarks(self):

        """

        Get's bookmarks for this repository

        """

        return self._get_bookmarks()

    def _get_bookmarks(self):

        if self._empty:

            return {}

        sortkey = lambda ctx: ctx[0]  # sort by name

        _bookmarks = [(safe_unicode(n), hex(h),) for n, h in

                 self._repo._bookmarks.items()]

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

    def _get_all_revisions(self):

        return map(lambda x: hex(x[7]), self._repo.changelog.index)[:-1]

    def get_diff(self, rev1, rev2, path='', ignore_whitespace=False,

                  context=3):

        """

        Returns (git like) *diff*, as plain text. Shows changes introduced by

        ``rev2`` since ``rev1``.

        :param rev1: Entry point from which diff is shown. Can be

          ``self.EMPTY_CHANGESET`` - in this case, patch showing all

          the changes since empty state of the repository until ``rev2``

        :param rev2: Until which revision changes should be shown.

        :param ignore_whitespace: If set to ``True``, would not show whitespace

          changes. Defaults to ``False``.

        :param context: How many lines before/after changed lines should be

          shown. Defaults to ``3``.

        """

        if hasattr(rev1, 'raw_id'):

            rev1 = getattr(rev1, 'raw_id')

        if hasattr(rev2, 'raw_id'):

            rev2 = getattr(rev2, 'raw_id')

        # Check if given revisions are present at repository (may raise

        # ChangesetDoesNotExistError)

        if rev1 != self.EMPTY_CHANGESET:

            self.get_changeset(rev1)

        self.get_changeset(rev2)

        file_filter = match(self.path, '', [path])

        return ''.join(patch.diff(self._repo, rev1, rev2, match=file_filter,

                          opts=diffopts(git=True,

                                        ignorews=ignore_whitespace,

                                        context=context)))

        """

        Function 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, return code 200 if url

        is valid or True if it's a local path

        """

        from mercurial.util import url as Url

        # those authnadlers are patched for python 2.6.5 bug an

        # infinit looping when given invalid resources

        from mercurial.url import httpbasicauthhandler, httpdigestauthhandler

        # check first if it's not an local url

        if os.path.isdir(url) or url.startswith('file:'):

            return True

        if('+' in url[:url.find('://')]):

        handlers = []

        test_uri, authinfo = Url(url).authinfo()

        if authinfo:

            #create a password manager

            passmgr = urllib2.HTTPPasswordMgrWithDefaultRealm()

            passmgr.add_password(*authinfo)

            handlers.extend((httpbasicauthhandler(passmgr),

                             httpdigestauthhandler(passmgr)))

        o = urllib2.build_opener(*handlers)

        o.addheaders = [('Content-Type', 'application/mercurial-0.1'),

                        ('Accept', 'application/mercurial-0.1')]

        q = {"cmd": 'between'}

        q.update({'pairs': "%s-%s" % ('0' * 40, '0' * 40)})

        qs = '?%s' % urllib.urlencode(q)

        cu = "%s%s" % (test_uri, qs)

        req = urllib2.Request(cu, None, {})

        try:

            resp = o.open(req)

            return resp.code == 200

        except Exception, e:

            # means it cannot be cloned

            raise urllib2.URLError(e)

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

        """

        Function will check for mercurial repository in given path and return

        a localrepo object. If there is no repository in that path it will

        raise an exception unless ``create`` parameter is set to True - in

        that case repository would be created and returned.

        If ``src_url`` is given, would try to clone repository from the

        location at given clone_point. Additionally it'll make update to

        working copy accordingly to ``update_after_clone`` flag

        """

        try:

            if src_url:

                url = str(self._get_url(src_url))

                opts = {}

                if not update_after_clone:

                    opts.update({'noupdate': True})

                try:

                    clone(self.baseui, url, self.path, **opts)

#                except urllib2.URLError:

#                    raise Abort("Got HTTP 404 error")

                except Exception:

                    raise

                # Don't try to create if we've already cloned repo

                create = False

            return localrepository(self.baseui, self.path, create=create)

        except (Abort, RepoError), err:

            if create:

                msg = "Cannot create repository at %s. Original error was %s"\

                    % (self.path, err)

            else:

                msg = "Not valid repository at %s. Original error was %s"\

                    % (self.path, err)

            raise RepositoryError(msg)

    @LazyProperty

    def in_memory_changeset(self):

        return MercurialInMemoryChangeset(self)

    @LazyProperty

    def description(self):

        undefined_description = u'unknown'

        return safe_unicode(self._repo.ui.config('web', 'description',

                                   undefined_description, untrusted=True))

    @LazyProperty

    def contact(self):

        undefined_contact = u'Unknown'

        return safe_unicode(get_contact(self._repo.ui.config)

                            or undefined_contact)

    @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:

            #fallback to filesystem

            cl_path = os.path.join(self.path, '.hg', "00changelog.i")

            st_path = os.path.join(self.path, '.hg', "store")

            if os.path.exists(cl_path):

                return os.stat(cl_path).st_mtime

            else:

                return os.stat(st_path).st_mtime

    def _get_hidden(self):

        return self._repo.ui.configbool("web", "hidden", untrusted=True)

    def _get_revision(self, revision):

        """

        Get's an ID revision given as str. This will always return a fill

        40 char revision number

        :param revision: str or int or None

        """

        if self._empty:

            raise EmptyRepositoryError("There are no changesets yet")

        if revision in [-1, 'tip', None]:

            revision = 'tip'

        try:

            revision = hex(self._repo.lookup(revision))

        except (IndexError, ValueError, RepoLookupError, TypeError):

            raise ChangesetDoesNotExistError("Revision %r does not "

                                    "exist for this repository %s" \

                                    % (revision, self))

        return revision

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

        allowed = self.baseui.configlist("web", "allow_archive",

                                         untrusted=True)

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

            if i[0] in allowed or self._repo.ui.configbool("web",

                                                           "allow" + i[0],

                                                           untrusted=True):

                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 = "file:" + urllib.pathname2url(url)

        return url