# -*- 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])

    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 {}

        sortkey = lambda ctx: ctx[0]

        _branches = [(x[0], x[1][0])

                     for x in self._parsed_refs.iteritems() if x[1][1] == 'H']

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

    @LazyProperty

    def tags(self):

        return self._get_tags()

    def _get_tags(self):

        if not self.revisions:

            return {}

        sortkey = lambda ctx: ctx[0]

        _tags = [(x[0], x[1][0])

                 for x in self._parsed_refs.iteritems() if x[1][1] == 'T']

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

    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

        :param message: message of the tag's commit

        :param date: date of tag's commit

        :raises TagAlreadyExistError: if tag with same name already exists

        """

        if name in self.tags:

            raise TagAlreadyExistError("Tag %s already exists" % name)

        changeset = self.get_changeset(revision)

        message = message or "Added tag %s for commit %s" % (name,

            changeset.raw_id)

        self._repo.refs["refs/tags/%s" % name] = changeset._commit.id

        self._parsed_refs = self._get_parsed_refs()

        self.tags = self._get_tags()

        return changeset

    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)

        tagpath = posixpath.join(self._repo.refs.path, 'refs', 'tags', name)

        try:

            os.remove(tagpath)

            self._parsed_refs = self._get_parsed_refs()

            self.tags = self._get_tags()

        except OSError, e:

            raise RepositoryError(e.strerror)

    @LazyProperty

    def _parsed_refs(self):

        return self._get_parsed_refs()

    def _get_parsed_refs(self):

        # cache the property

        _repo = self._repo

        refs = _repo.get_refs()

        keys = [('refs/heads/', 'H'),

                ('refs/remotes/origin/', 'RH'),

                ('refs/tags/', 'T')]

        _refs = {}

        for ref, sha in refs.iteritems():

            for k, type_ in keys:

                if ref.startswith(k):

                    _key = ref[len(k):]

                    if type_ == 'T':

                        obj = _repo.get_object(sha)

                        if isinstance(obj, Tag):

                            sha = _repo.get_object(sha).object[1]

                    _refs[_key] = [sha, type_]

                    break

        return _refs

    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_changeset(self, revision=None):

        """

        Returns ``GitChangeset`` object representing commit from git repository

        at the given revision or head (most recent commit) if None given.

        """

        if isinstance(revision, GitChangeset):

            return revision

        revision = self._get_revision(revision)

        changeset = GitChangeset(repository=self, revision=revision)

        return changeset

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

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

        """

        Returns iterator of ``GitChangeset`` objects from start to end (both

        are inclusive), in ascending date order (unless ``reverse`` is set).

        :param start: changeset ID, as str; first returned changeset

        :param end: changeset ID, as str; last returned changeset

        :param start_date: if specified, changesets with commit date less than

          ``start_date`` would be filtered out from returned set

        :param end_date: if specified, changesets with commit date greater than

          ``end_date`` would be filtered out from returned set

        :param branch_name: if specified, changesets not reachable from given

          branch would be filtered out from returned set

        :param reverse: if ``True``, returned generator would be reversed

          (meaning that returned changesets would have descending date order)

        :raise BranchDoesNotExistError: If given ``branch_name`` does not

            exist.

        :raise ChangesetDoesNotExistError: If changeset for given ``start`` or

          ``end`` could not be found.

        """

        if branch_name and branch_name not in self.branches:

            raise BranchDoesNotExistError("Branch '%s' not found" \

                                          % branch_name)

        # %H at format means (full) commit hash, initial hashes are retrieved

        # in ascending date order

        cmd_template = 'log --date-order --reverse --pretty=format:"%H"'

        cmd_params = {}

        if start_date:

            cmd_template += ' --since "$since"'

            cmd_params['since'] = start_date.strftime('%m/%d/%y %H:%M:%S')

        if end_date:

            cmd_template += ' --until "$until"'

            cmd_params['until'] = end_date.strftime('%m/%d/%y %H:%M:%S')

        if branch_name:

            cmd_template += ' $branch_name'

            cmd_params['branch_name'] = branch_name

        else:

            rev_filter = _git_path = settings.GIT_REV_FILTER

            cmd_template += ' %s' % (rev_filter)

        cmd = string.Template(cmd_template).safe_substitute(**cmd_params)

        revs = self.run_git_command(cmd)[0].splitlines()

        start_pos = 0

        end_pos = len(revs)

        if start:

'''

Module provides a class allowing to wrap communication over subprocess.Popen

input, output, error streams into a meaningfull, non-blocking, concurrent

stream processor exposing the output data as an iterator fitting to be a

return value passed by a WSGI applicaiton to a WSGI server per PEP 3333.

Copyright (c) 2011  Daniel Dotsenko <dotsa@hotmail.com>

This file is part of git_http_backend.py Project.

git_http_backend.py Project is free software: you can redistribute it and/or

modify it under the terms of the GNU Lesser General Public License as

published by the Free Software Foundation, either version 2.1 of the License,

or (at your option) any later version.

git_http_backend.py Project is distributed in the hope that it will be useful,

but WITHOUT ANY WARRANTY; without even the implied warranty of

MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License

along with git_http_backend.py Project.

If not, see <http://www.gnu.org/licenses/>.

'''

import os

import subprocess

from rhodecode.lib.vcs.utils.compat import deque, Event, Thread, _bytes, _bytearray

class StreamFeeder(Thread):

    """

    Normal writing into pipe-like is blocking once the buffer is filled.

    This thread allows a thread to seep data from a file-like into a pipe

    without blocking the main thread.

    We close inpipe once the end of the source stream is reached.

    """

    def __init__(self, source):

        super(StreamFeeder, self).__init__()

        self.daemon = True

        filelike = False

        self.bytes = _bytes()

        if type(source) in (type(''), _bytes, _bytearray):  # string-like

            self.bytes = _bytes(source)

        else:  # can be either file pointer or file-like

            if type(source) in (int, long):  # file pointer it is

                ## converting file descriptor (int) stdin into file-like

                try:

                    source = os.fdopen(source, 'rb', 16384)

                except Exception:

                    pass

            # let's see if source is file-like by now

            try:

                filelike = source.read

            except Exception:

                pass

        if not filelike and not self.bytes:

            raise TypeError("StreamFeeder's source object must be a readable "

                            "file-like, a file descriptor, or a string-like.")

        self.source = source

        self.readiface, self.writeiface = os.pipe()

    def run(self):

        t = self.writeiface

        if self.bytes:

            os.write(t, self.bytes)

        else:

            s = self.source

            b = s.read(4096)

            while b:

                os.write(t, b)

                b = s.read(4096)

        os.close(t)

    @property

    def output(self):

        return self.readiface

class InputStreamChunker(Thread):

    def __init__(self, source, target, buffer_size, chunk_size):

        super(InputStreamChunker, self).__init__()

        self.daemon = True  # die die die.

        self.source = source

        self.target = target

        self.chunk_count_max = int(buffer_size / chunk_size) + 1

        self.chunk_size = chunk_size

        self.data_added = Event()

        self.data_added.clear()

        self.keep_reading = Event()

        self.keep_reading.set()

        self.EOF = Event()

        self.EOF.clear()

        self.go = Event()

        self.go.set()

    def stop(self):

        self.go.clear()

        self.EOF.set()

        try:

            # this is not proper, but is done to force the reader thread let

            # go of the input because, if successful, .close() will send EOF

            # down the pipe.

            self.source.close()

        except:

            pass

    def run(self):

        s = self.source

        t = self.target

        cs = self.chunk_size

        ccm = self.chunk_count_max

        kr = self.keep_reading

        da = self.data_added

        go = self.go

        try:

            b = s.read(cs)

        except ValueError:

            b = ''

        while b and go.is_set():

            if len(t) > ccm:

                kr.clear()

                kr.wait(2)

#                # this only works on 2.7.x and up

#                if not kr.wait(10):

#                    raise Exception("Timed out while waiting for input to be read.")

                # instead we'll use this

                if len(t) > ccm + 3:

                    raise IOError("Timed out while waiting for input from subprocess.")

            t.append(b)

            da.set()

            b = s.read(cs)

        self.EOF.set()

        da.set()  # for cases when done but there was no input.

class BufferedGenerator():

    '''

    Class behaves as a non-blocking, buffered pipe reader.

    Reads chunks of data (through a thread)

    from a blocking pipe, and attaches these to an array (Deque) of chunks.

    Reading is halted in the thread when max chunks is internally buffered.

    The .next() may operate in blocking or non-blocking fashion by yielding

    '' if no data is ready

    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

            )