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

"""

    vcs.backends.git.repository

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

    Git repository implementation.

    :created_on: Apr 8, 2010

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

"""

import os

import re

import time

import urllib

import urllib2

import logging

import posixpath

import string

from dulwich.objects import Tag

from dulwich.repo import Repo, NotGitRepository

from rhodecode.lib.vcs import subprocessio

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

from rhodecode.lib.vcs.conf import settings

from rhodecode.lib.vcs.exceptions import (

    BranchDoesNotExistError, ChangesetDoesNotExistError, EmptyRepositoryError,

    RepositoryError, TagAlreadyExistError, 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, get_user_home

from rhodecode.lib.vcs.utils.hgcompat import (

    hg_url, httpbasicauthhandler, httpdigestauthhandler

)

from .changeset import GitChangeset

from .config import ConfigFile

from .inmemory import GitInMemoryChangeset

from .workdir import GitWorkdir

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

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)

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

        self.bare = repo.bare

    @property

    def _config_files(self):

        return [

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

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

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

         ]

    @property

    def _repo(self):

        return Repo(self.path)

    @property

    def head(self):

        try:

            return self._repo.head()

        except KeyError:

            return None

    @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()

    @classmethod

    def _run_git_command(cls, cmd, **opts):

        """

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

        (stdout, stderr).

        :param cmd: git command to be executed

        :param opts: env options to pass into Subprocess command

        """

        if '_bare' in opts:

            _copts = []

            del opts['_bare']

        else:

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

        safe_call = False

        if '_safe' in opts:

            #no exc on failure

            del opts['_safe']

            safe_call = True

        _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'

        _git_path = settings.GIT_EXECUTABLE_PATH

        cmd = [_git_path] + _copts + cmd

        if _str_cmd:

            cmd = ' '.join(cmd)

        try:

            _opts = dict(

                env=gitenv,

            )

            _opts.update(opts)

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

        except (EnvironmentError, OSError), err:

            tb_err = ("Couldn't run git command (%s).\n"

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

            log.error(tb_err)

            if safe_call:

                return '', err

            else:

                raise RepositoryError(tb_err)

        return ''.join(p.output), ''.join(p.error)

    def run_git_command(self, cmd):

        opts = {}

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

            opts['cwd'] = self.path

        return self._run_git_command(cmd, **opts)

    @classmethod

    def _check_url(cls, 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

        """

        # 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('://')]):

            url = url[url.find('+') + 1:]

        handlers = []

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

        if not test_uri.endswith('info/refs'):

            test_uri = test_uri.rstrip('/') + '/info/refs'

        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 = [('User-Agent', 'git/1.7.8.0')]  # fake some git

        q = {"service": 'git-upload-pack'}

        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("[%s] %s" % (url, e))

    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:

                GitRepository._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 self._repo

        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 []

        rev_filter = _git_path = settings.GIT_REV_FILTER

        cmd = 'rev-list %s --reverse --date-order' % (rev_filter)

        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.

        """

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

        try:

            self.revisions[0]

        except (KeyError, IndexError):

            raise EmptyRepositoryError("There are no changesets yet")

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

            return self.revisions[-1]

        is_bstr = isinstance(revision, (str, unicode))

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

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

            try:

                revision = self.revisions[int(revision)]

            except Exception:

                raise ChangesetDoesNotExistError("Revision %s does not exist "

                    "for this repository" % (revision))

        elif is_bstr:

            # get by branch/tag name

            _ref_revision = self._parsed_refs.get(revision)

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

                return _ref_revision[0]

            _tags_shas = self.tags.values()

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

            if revision in _tags_shas:

                return _tags_shas[_tags_shas.index(revision)]

            elif not SHA_PATTERN.match(revision) or revision not in self.revisions:

                raise ChangesetDoesNotExistError("Revision %s does not exist "

                    "for this repository" % (revision))

        # Ensure we return full id

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

            raise ChangesetDoesNotExistError("Given revision %s 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

    def get_hook_location(self):

        """

        returns absolute path to location where hooks are stored

        """

        loc = os.path.join(self.path, 'hooks')

        if not self.bare:

            loc = os.path.join(self.path, '.git', 'hooks')

        return loc

    @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])

    to be sent or by not returning until there is some data to send

    When we get EOF from underlying source pipe we raise the marker to raise

    StopIteration after the last chunk of data is yielded.

    '''

    def __init__(self, source, buffer_size=65536, chunk_size=4096,

                 starting_values=[], bottomless=False):

        if bottomless:

            maxlen = int(buffer_size / chunk_size)

        else:

            maxlen = None

        self.data = deque(starting_values, maxlen)

        self.worker = InputStreamChunker(source, self.data, buffer_size,

                                         chunk_size)

        if starting_values:

            self.worker.data_added.set()

        self.worker.start()

    ####################

    # Generator's methods

    ####################

    def __iter__(self):

        return self

    def next(self):

        while not len(self.data) and not self.worker.EOF.is_set():

            self.worker.data_added.clear()

            self.worker.data_added.wait(0.2)

        if len(self.data):

            self.worker.keep_reading.set()

            return _bytes(self.data.popleft())

        elif self.worker.EOF.is_set():

            raise StopIteration

    def throw(self, type, value=None, traceback=None):

        if not self.worker.EOF.is_set():

            raise type(value)

    def start(self):

        self.worker.start()

    def stop(self):

        self.worker.stop()

    def close(self):

        try:

            self.worker.stop()

            self.throw(GeneratorExit)

        except (GeneratorExit, StopIteration):

            pass

    def __del__(self):

        self.close()

    ####################

    # Threaded reader's infrastructure.

    ####################

    @property

    def input(self):

        return self.worker.w

    @property

    def data_added_event(self):

        return self.worker.data_added

    @property

    def data_added(self):

        return self.worker.data_added.is_set()

    @property

    def reading_paused(self):

        return not self.worker.keep_reading.is_set()

    @property

    def done_reading_event(self):

        '''

        Done_reding does not mean that the iterator's buffer is empty.

        Iterator might have done reading from underlying source, but the read

        chunks might still be available for serving through .next() method.

        @return An Event class instance.

        '''

        return self.worker.EOF

    @property

    def done_reading(self):

        '''

        Done_reding does not mean that the iterator's buffer is empty.

        Iterator might have done reading from underlying source, but the read

        chunks might still be available for serving through .next() method.

        @return An Bool value.

        '''

        return self.worker.EOF.is_set()

    @property

    def length(self):

        '''

        returns int.

        This is the lenght of the que of chunks, not the length of

        the combined contents in those chunks.

        __len__() cannot be meaningfully implemented because this

        reader is just flying throuh a bottomless pit content and

        can only know the lenght of what it already saw.

        If __len__() on WSGI server per PEP 3333 returns a value,

        the responce's length will be set to that. In order not to

        confuse WSGI PEP3333 servers, we will not implement __len__

        at all.

        '''

        return len(self.data)

    def prepend(self, x):

        self.data.appendleft(x)

    def append(self, x):

        self.data.append(x)

    def extend(self, o):

        self.data.extend(o)

    def __getitem__(self, i):

        return self.data[i]

class SubprocessIOChunker(object):

    '''

    Processor class wrapping handling of subprocess IO.

    In a way, this is a "communicate()" replacement with a twist.

    - We are multithreaded. Writing in and reading out, err are all sep threads.

    - We support concurrent (in and out) stream processing.

    - The output is not a stream. It's a queue of read string (bytes, not unicode)

      chunks. The object behaves as an iterable. You can "for chunk in obj:" us.

    - We are non-blocking in more respects than communicate()

      (reading from subprocess out pauses when internal buffer is full, but

       does not block the parent calling code. On the flip side, reading from

       slow-yielding subprocess may block the iteration until data shows up. This

       does not block the parallel inpipe reading occurring parallel thread.)

    The purpose of the object is to allow us to wrap subprocess interactions into

    and interable that can be passed to a WSGI server as the application's return

    value. Because of stream-processing-ability, WSGI does not have to read ALL

    of the subprocess's output and buffer it, before handing it to WSGI server for

    HTTP response. Instead, the class initializer reads just a bit of the stream

    to figure out if error ocurred or likely to occur and if not, just hands the

    further iteration over subprocess output to the server for completion of HTTP

    response.

    The real or perceived subprocess error is trapped and raised as one of

    EnvironmentError family of exceptions

    Example usage:

    #    try:

    #        answer = SubprocessIOChunker(

    #            cmd,

    #            input,

    #            buffer_size = 65536,

    #            chunk_size = 4096

    #            )

    #    except (EnvironmentError) as e:

    #        print str(e)

    #        raise e

    #

    #    return answer

    '''

    def __init__(self, cmd, inputstream=None, buffer_size=65536,

                 chunk_size=4096, starting_values=[], **kwargs):

        '''

        Initializes SubprocessIOChunker

        :param cmd: A Subprocess.Popen style "cmd". Can be string or array of strings

        :param inputstream: (Default: None) A file-like, string, or file pointer.

        :param buffer_size: (Default: 65536) A size of total buffer per stream in bytes.

        :param chunk_size: (Default: 4096) A max size of a chunk. Actual chunk may be smaller.

        :param starting_values: (Default: []) An array of strings to put in front of output que.

        '''

        if inputstream:

            input_streamer = StreamFeeder(inputstream)

            input_streamer.start()

            inputstream = input_streamer.output

        if isinstance(cmd, (list, tuple)):

            cmd = ' '.join(cmd)

        kwargs['shell'] = _shell

        _p = subprocess.Popen(cmd,

            bufsize=-1,

            stdin=inputstream,

            stdout=subprocess.PIPE,

            stderr=subprocess.PIPE,

            **kwargs

            )

        bg_out = BufferedGenerator(_p.stdout, buffer_size, chunk_size, starting_values)

        bg_err = BufferedGenerator(_p.stderr, 16000, 1, bottomless=True)

        while not bg_out.done_reading and not bg_out.reading_paused and not bg_err.length:

            # doing this until we reach either end of file, or end of buffer.

            bg_out.data_added_event.wait(1)

            bg_out.data_added_event.clear()

        # at this point it's still ambiguous if we are done reading or just full buffer.

        # Either way, if error (returned by ended process, or implied based on

        # presence of stuff in stderr output) we error out.

        # Else, we are happy.

        _returncode = _p.poll()

        if _returncode or (_returncode == None and bg_err.length):

            try:

                _p.terminate()

            except:

                pass

            bg_out.stop()

            bg_err.stop()

            err = '%s' % ''.join(bg_err)

            if err:

                raise EnvironmentError("Subprocess exited due to an error:\n" + err)

            raise EnvironmentError("Subprocess exited with non 0 ret code:%s" % _returncode)

        self.process = _p

        self.output = bg_out

        self.error = bg_err

    def __iter__(self):

        return self

    def next(self):

        if self.process.poll():

            err = '%s' % ''.join(self.error)

            raise EnvironmentError("Subprocess exited due to an error:\n" + err)

        return self.output.next()

    def throw(self, type, value=None, traceback=None):

        if self.output.length or not self.output.done_reading:

            raise type(value)

    def close(self):

        try:

            self.process.terminate()

        except:

            pass

        try:

            self.output.close()

        except:

            pass

        try:

            self.error.close()

        except:

            pass

    def __del__(self):

        self.close()