'''

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.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

        b = s.read(cs)

        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)

        _shell = kwargs.get('shell') or True

        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)

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

        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()

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

"""

    rhodecode.lib.utils

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

    Utilities library for RhodeCode

    :created_on: Apr 18, 2010

    :author: marcink

    :copyright: (C) 2010-2012 Marcin Kuzminski <marcin@python-works.com>

    :license: GPLv3, see COPYING for more details.

"""

# This program is free software: you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation, either version 3 of the License, or

# (at your option) any later version.

#

# This program 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 General Public License for more details.

#

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

# along with this program.  If not, see <http://www.gnu.org/licenses/>.

import os

import re

import logging

import datetime

import traceback

import paste

import beaker

import tarfile

import shutil

import decorator

import warnings

from os.path import abspath

from os.path import dirname as dn, join as jn

from paste.script.command import Command, BadCommand

from mercurial import ui, config

from webhelpers.text import collapse, remove_formatting, strip_tags

from rhodecode.lib.vcs import get_backend

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

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

from rhodecode.lib.vcs.utils.helpers import get_scm

from rhodecode.lib.vcs.exceptions import VCSError

from rhodecode.lib.caching_query import FromCache

from rhodecode.model import meta

from rhodecode.model.db import Repository, User, RhodeCodeUi, \

    UserLog, RepoGroup, RhodeCodeSetting, CacheInvalidation

from rhodecode.model.meta import Session

from rhodecode.model.repos_group import ReposGroupModel

from rhodecode.lib.utils2 import safe_str, safe_unicode

from rhodecode.lib.vcs.utils.fakemod import create_module

log = logging.getLogger(__name__)

REMOVED_REPO_PAT = re.compile(r'rm__\d{8}_\d{6}_\d{6}__.*')

def recursive_replace(str_, replace=' '):

    """

    Recursive replace of given sign to just one instance

    :param str_: given string

    :param replace: char to find and replace multiple instances

    Examples::

    >>> recursive_replace("Mighty---Mighty-Bo--sstones",'-')

    'Mighty-Mighty-Bo-sstones'

    """

    if str_.find(replace * 2) == -1:

        return str_

    else:

        str_ = str_.replace(replace * 2, replace)

        return recursive_replace(str_, replace)

def repo_name_slug(value):

    """

    Return slug of name of repository

    This function is called on each creation/modification

    of repository to prevent bad names in repo

    """

    slug = remove_formatting(value)

    slug = strip_tags(slug)

    for c in """`?=[]\;'"<>,/~!@#$%^&*()+{}|: """:

        slug = slug.replace(c, '-')

    slug = recursive_replace(slug, '-')

    slug = collapse(slug, '-')

    return slug

def get_repo_slug(request):

    _repo = request.environ['pylons.routes_dict'].get('repo_name')

    if _repo:

        _repo = _repo.rstrip('/')

    return _repo

def get_repos_group_slug(request):

    _group = request.environ['pylons.routes_dict'].get('group_name')

    if _group:

        _group = _group.rstrip('/')

    return _group

def action_logger(user, action, repo, ipaddr='', sa=None, commit=False):

    """

    Action logger for various actions made by users

    :param user: user that made this action, can be a unique username string or

        object containing user_id attribute

    :param action: action to log, should be on of predefined unique actions for

        easy translations

    :param repo: string name of repository or object containing repo_id,

        that action was made on

    :param ipaddr: optional ip address from what the action was made

    :param sa: optional sqlalchemy session

    """

    if not sa:

        sa = meta.Session()

    try:

        if hasattr(user, 'user_id'):

            user_obj = User.get(user.user_id)

        elif isinstance(user, basestring):

            user_obj = User.get_by_username(user)

        else:

            raise Exception('You have to provide a user object or a username')

        if hasattr(repo, 'repo_id'):

            repo_obj = Repository.get(repo.repo_id)

            repo_name = repo_obj.repo_name

        elif  isinstance(repo, basestring):

            repo_name = repo.lstrip('/')

            repo_obj = Repository.get_by_repo_name(repo_name)

        else:

            repo_obj = None

            repo_name = ''

        user_log = UserLog()

        user_log.user_id = user_obj.user_id

        user_log.username = user_obj.username

        user_log.action = safe_unicode(action)

        user_log.repository = repo_obj

        user_log.repository_name = repo_name

        user_log.action_date = datetime.datetime.now()

        user_log.user_ip = ipaddr

        sa.add(user_log)

        log.info('Logging action %s on %s by %s' %

                 (action, safe_unicode(repo), user_obj))

        if commit:

            sa.commit()

    except:

        log.error(traceback.format_exc())

        raise

def get_repos(path, recursive=False, skip_removed_repos=True):

    """

    Scans given path for repos and return (name,(type,path)) tuple

    :param path: path to scan for repositories

    :param recursive: recursive search and return names with subdirs in front

    """

    # remove ending slash for better results

    path = path.rstrip(os.sep)

    log.debug('now scanning in %s location recursive:%s...' % (path, recursive))

    def _get_repos(p):

        if not os.access(p, os.W_OK):

            return

        for dirpath in os.listdir(p):

            if os.path.isfile(os.path.join(p, dirpath)):

                continue

            cur_path = os.path.join(p, dirpath)

            # skip removed repos

            if skip_removed_repos and REMOVED_REPO_PAT.match(dirpath):

                continue

            #skip .<somethin> dirs

            if dirpath.startswith('.'):

                continue

            try:

                scm_info = get_scm(cur_path)

                yield scm_info[1].split(path, 1)[-1].lstrip(os.sep), scm_info

            except VCSError:

                if not recursive:

                    continue

                #check if this dir containts other repos for recursive scan

                rec_path = os.path.join(p, dirpath)

                if os.path.isdir(rec_path):

                    for inner_scm in _get_repos(rec_path):

                        yield inner_scm

    return _get_repos(path)

#alias for backward compat

get_filesystem_repos = get_repos

def is_valid_repo(repo_name, base_path, scm=None):

    """

    Returns True if given path is a valid repository False otherwise.

    If scm param is given also compare if given scm is the same as expected

    from scm parameter

    :param repo_name:

    :param base_path:

    :param scm:

    :return True: if given path is a valid repository

    """

    full_path = os.path.join(safe_str(base_path), safe_str(repo_name))

    try:

        scm_ = get_scm(full_path)

        if scm:

            return scm_[0] == scm

        return True

    except VCSError:

        return False

def is_valid_repos_group(repos_group_name, base_path):

    """

    Returns True if given path is a repos group False otherwise

    :param repo_name:

    :param base_path:

    """

    full_path = os.path.join(safe_str(base_path), safe_str(repos_group_name))

    # check if it's not a repo

    if is_valid_repo(repos_group_name, base_path):

        return False

    try:

        # we need to check bare git repos at higher level

        # since we might match branches/hooks/info/objects or possible

        # other things inside bare git repo

        get_scm(os.path.dirname(full_path))

        return False

    except VCSError:

        pass

    # check if it's a valid path

    if os.path.isdir(full_path):

        return True

    return False

def ask_ok(prompt, retries=4, complaint='Yes or no please!'):

    while True:

        ok = raw_input(prompt)

        if ok in ('y', 'ye', 'yes'):

            return True

        if ok in ('n', 'no', 'nop', 'nope'):

            return False

        retries = retries - 1

        if retries < 0:

            raise IOError

        print complaint

#propagated from mercurial documentation

ui_sections = ['alias', 'auth',

                'decode/encode', 'defaults',

                'diff', 'email',

                'extensions', 'format',

                'merge-patterns', 'merge-tools',

                'hooks', 'http_proxy',

                'smtp', 'patch',

                'paths', 'profiling',

                'server', 'trusted',

                'ui', 'web', ]

def make_ui(read_from='file', path=None, checkpaths=True, clear_session=True):

    """

    A function that will read python rc files or database

    and make an mercurial ui object from read options

    :param path: path to mercurial config file

    :param checkpaths: check the path

    :param read_from: read from 'file' or 'db'

    """

    baseui = ui.ui()

    # clean the baseui object

    baseui._ocfg = config.config()

    baseui._ucfg = config.config()

    baseui._tcfg = config.config()

    if read_from == 'file':

        if not os.path.isfile(path):

            log.debug('hgrc file is not present at %s, skipping...' % path)

            return False

        log.debug('reading hgrc from %s' % path)

        cfg = config.config()

        cfg.read(path)

        for section in ui_sections:

            for k, v in cfg.items(section):

                log.debug('settings ui from file: [%s] %s=%s' % (section, k, v))

                baseui.setconfig(safe_str(section), safe_str(k), safe_str(v))

    elif read_from == 'db':

        sa = meta.Session()

        ret = sa.query(RhodeCodeUi)\

            .options(FromCache("sql_cache_short", "get_hg_ui_settings"))\

            .all()

        hg_ui = ret

        for ui_ in hg_ui:

            if ui_.ui_active:

                log.debug('settings ui from db: [%s] %s=%s', ui_.ui_section,

                          ui_.ui_key, ui_.ui_value)

                baseui.setconfig(safe_str(ui_.ui_section), safe_str(ui_.ui_key),

                                 safe_str(ui_.ui_value))

            if ui_.ui_key == 'push_ssl':

                # force set push_ssl requirement to False, rhodecode

                # handles that

                baseui.setconfig(safe_str(ui_.ui_section), safe_str(ui_.ui_key),

                                 False)

        if clear_session:

            meta.Session.remove()

    return baseui

def set_rhodecode_config(config):

    """

    Updates pylons config with new settings from database

    :param config:

    """

    hgsettings = RhodeCodeSetting.get_app_settings()

    for k, v in hgsettings.items():

        config[k] = v

def invalidate_cache(cache_key, *args):

    """

    Puts cache invalidation task into db for

    further global cache invalidation

    """

    from rhodecode.model.scm import ScmModel

    if cache_key.startswith('get_repo_cached_'):

        name = cache_key.split('get_repo_cached_')[-1]

        ScmModel().mark_for_invalidation(name)

def map_groups(path):

    """

    Given a full path to a repository, create all nested groups that this

    repo is inside. This function creates parent-child relationships between

    groups and creates default perms for all new groups.

    :param paths: full path to repository

    """

    sa = meta.Session()

    groups = path.split(Repository.url_sep())

    parent = None

    group = None

    # last element is repo in nested groups structure

    groups = groups[:-1]

    rgm = ReposGroupModel(sa)

    for lvl, group_name in enumerate(groups):

        group_name = '/'.join(groups[:lvl] + [group_name])

        group = RepoGroup.get_by_group_name(group_name)

        desc = '%s group' % group_name

        # skip folders that are now removed repos

        if REMOVED_REPO_PAT.match(group_name):

            break

        if group is None:

            log.debug('creating group level: %s group_name: %s' % (lvl,

                                                                   group_name))

            group = RepoGroup(group_name, parent)

            group.group_description = desc

            sa.add(group)

            rgm._create_default_perms(group)

            sa.flush()

        parent = group

    return group

def repo2db_mapper(initial_repo_list, remove_obsolete=False,

                   install_git_hook=False):

    """

    maps all repos given in initial_repo_list, non existing repositories

    are created, if remove_obsolete is True it also check for db entries

    that are not in initial_repo_list and removes them.

    :param initial_repo_list: list of repositories found by scanning methods

    :param remove_obsolete: check for obsolete entries in database

    :param install_git_hook: if this is True, also check and install githook

        for a repo if missing

    """

    from rhodecode.model.repo import RepoModel

    from rhodecode.model.scm import ScmModel

    sa = meta.Session()

    rm = RepoModel()

    user = sa.query(User).filter(User.admin == True).first()

    if user is None:

        raise Exception('Missing administrative account!')

    added = []

#    # clear cache keys

#    log.debug("Clearing cache keys now...")

#    CacheInvalidation.clear_cache()

#    sa.commit()

    ##creation defaults

    defs = RhodeCodeSetting.get_default_repo_settings(strip_prefix=True)

    enable_statistics = defs.get('repo_enable_statistics')

    enable_locking = defs.get('repo_enable_locking')

    enable_downloads = defs.get('repo_enable_downloads')

    private = defs.get('repo_private')

    for name, repo in initial_repo_list.items():

        group = map_groups(name)

        db_repo = rm.get_by_repo_name(name)

        # found repo that is on filesystem not in RhodeCode database

        if not db_repo:

            log.info('repository %s not found, creating now' % name)

            added.append(name)

            desc = (repo.description

                    if repo.description != 'unknown'

                    else '%s repository' % name)

            new_repo = rm.create_repo(

                repo_name=name,

                repo_type=repo.alias,