# copyright 2003-2011 LOGILAB S.A. (Paris, FRANCE), all rights reserved.

# contact http://www.logilab.fr/ -- mailto:contact@logilab.fr

#

# This file is part of CubicWeb.

#

# CubicWeb 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.

#

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

"""twisted server for CubicWeb web instances"""

from __future__ import with_statement

__docformat__ = "restructuredtext en"

import sys

import os

import os.path as osp

import select

import traceback

import threading

import re

import hashlib

from os.path import join

from time import mktime

from datetime import date, timedelta

from urlparse import urlsplit, urlunsplit

from cgi import FieldStorage, parse_header

from cStringIO import StringIO

from twisted.internet import reactor, task, threads

from twisted.internet.defer import maybeDeferred

from twisted.web import http, server

from twisted.web import static, resource

from twisted.web.server import NOT_DONE_YET

from logilab.common.decorators import monkeypatch

from cubicweb import (AuthenticationError, ConfigurationError,

                      CW_EVENT_MANAGER, CubicWebException)

from cubicweb.utils import json_dumps

from cubicweb.web import Redirect, DirectResponse, StatusResponse, LogOut

from cubicweb.web.application import CubicWebPublisher

from cubicweb.web.http_headers import generateDateTime

from cubicweb.etwist.request import CubicWebTwistedRequestAdapter

from cubicweb.etwist.http import HTTPResponse

def start_task(interval, func):

    lc = task.LoopingCall(func)

    # wait until interval has expired to actually start the task, else we have

    # to wait all task to be finished for the server to be actually started

    lc.start(interval, now=False)

def host_prefixed_baseurl(baseurl, host):

    scheme, netloc, url, query, fragment = urlsplit(baseurl)

    netloc_domain = '.' + '.'.join(netloc.split('.')[-2:])

    if host.endswith(netloc_domain):

        netloc = host

    baseurl = urlunsplit((scheme, netloc, url, query, fragment))

    return baseurl

class ForbiddenDirectoryLister(resource.Resource):

    def render(self, request):

        return HTTPResponse(twisted_request=request,

                            code=http.FORBIDDEN,

                            stream='Access forbidden')

class NoListingFile(static.File):

    def set_expires(self, request):

        if not self.config.debugmode:

            # XXX: Don't provide additional resource information to error responses

            #

            # the HTTP RFC recommands not going further than 1 year ahead

            expires = date.today() + timedelta(days=6*30)

            request.setHeader('Expires', generateDateTime(mktime(expires.timetuple())))

    def directoryListing(self):

        return ForbiddenDirectoryLister()

class DataLookupDirectory(NoListingFile):

    def __init__(self, config, path):

        self.md5_version = config.instance_md5_version()

        NoListingFile.__init__(self, path)

        self.config = config

        self.here = path

        self._defineChildResources()

        if self.config.debugmode:

            self.data_modconcat_basepath = '/data/??'

        else:

            self.data_modconcat_basepath = '/data/%s/??' % self.md5_version

    def _defineChildResources(self):

        self.putChild(self.md5_version, self)

    def getChild(self, path, request):

        if not path:

            uri = request.uri

            if uri.startswith('/https/'):

                uri = uri[6:]

            if uri.startswith(self.data_modconcat_basepath):

                resource_relpath = uri[len(self.data_modconcat_basepath):]

                if resource_relpath:

                    paths = resource_relpath.split(',')

                    try:

                        self.set_expires(request)

                        return ConcatFiles(self.config, paths)

                    except ConcatFileNotFoundError:

                        return self.childNotFound

            return self.directoryListing()

        childpath = join(self.here, path)

        dirpath, rid = self.config.locate_resource(childpath)

        if dirpath is None:

            # resource not found

            return self.childNotFound

        filepath = os.path.join(dirpath, rid)

        if os.path.isdir(filepath):

            resource = DataLookupDirectory(self.config, childpath)

            # cache resource for this segment path to avoid recomputing

            # directory lookup

            self.putChild(path, resource)

            return resource

        else:

            self.set_expires(request)

            return NoListingFile(filepath)

class FCKEditorResource(NoListingFile):

    def __init__(self, config, path):

        NoListingFile.__init__(self, path)

        self.config = config

    def getChild(self, path, request):

        pre_path = request.path.split('/')[1:]

        if pre_path[0] == 'https':

            pre_path.pop(0)

            uiprops = self.config.https_uiprops

        else:

            uiprops = self.config.uiprops

        return static.File(osp.join(uiprops['FCKEDITOR_PATH'], path))

class LongTimeExpiringFile(DataLookupDirectory):

    """overrides static.File and sets a far future ``Expires`` date

    on the resouce.

    versions handling is done by serving static files by different

    URLs for each version. For instance::

      http://localhost:8080/data-2.48.2/cubicweb.css

      http://localhost:8080/data-2.49.0/cubicweb.css

      etc.

    """

    def _defineChildResources(self):

        pass

class ConcatFileNotFoundError(CubicWebException):

    pass

class ConcatFiles(LongTimeExpiringFile):

    def __init__(self, config, paths):

        _, ext = osp.splitext(paths[0])

        self._resources = {}

        # create a unique / predictable filename. We don't consider cubes

        # version since uicache is cleared at server startup, and file's dates

        # are checked in debug mode

        fname = 'cache_concat_' + hashlib.md5(';'.join(paths)).hexdigest() + ext

        filepath = osp.join(config.appdatahome, 'uicache', fname)

        LongTimeExpiringFile.__init__(self, config, filepath)

        self._concat_cached_filepath(filepath, paths)

    def _resource(self, path):

        try:

            return self._resources[path]

        except KeyError:

            self._resources[path] = self.config.locate_resource(path)

            return self._resources[path]

    def _concat_cached_filepath(self, filepath, paths):

        if not self._up_to_date(filepath, paths):

            with open(filepath, 'wb') as f:

                for path in paths:

                    dirpath, rid = self._resource(path)

                    if rid is None:

                        # In production mode log an error, do not return a 404

                        # XXX the erroneous content is cached anyway

                        LOGGER.error('concatenated data url error: %r file '

                                     'does not exist', path)

                        if self.config.debugmode:

                            raise ConcatFileNotFoundError(path)

                    else:

                        for line in open(osp.join(dirpath, rid)):

                            f.write(line)

                        f.write('\n')

    def _up_to_date(self, filepath, paths):

        """

        The concat-file is considered up-to-date if it exists.

        In debug mode, an additional check is performed to make sure that

        concat-file is more recent than all concatenated files

        """

        if not osp.isfile(filepath):

            return False

        if self.config.debugmode:

            concat_lastmod = os.stat(filepath).st_mtime

            for path in paths:

                dirpath, rid = self._resource(path)

                if rid is None:

                    raise ConcatFileNotFoundError(path)

                path = osp.join(dirpath, rid)

                if os.stat(path).st_mtime > concat_lastmod:

                    return False

        return True

class CubicWebRootResource(resource.Resource):

    def __init__(self, config, vreg=None):

        resource.Resource.__init__(self)

        self.config = config

        # instantiate publisher here and not in init_publisher to get some

        # checks done before daemonization (eg versions consistency)

        self.appli = CubicWebPublisher(config, vreg=vreg)

        self.base_url = config['base-url']

        self.https_url = config['https-url']

        global MAX_POST_LENGTH

        MAX_POST_LENGTH = config['max-post-length']

        self.putChild('static', NoListingFile(config.static_directory))

        self.putChild('fckeditor', FCKEditorResource(self.config, ''))

        self.putChild('data', DataLookupDirectory(self.config, ''))

    def init_publisher(self):

        config = self.config

        # when we have an in-memory repository, clean unused sessions every XX

        # seconds and properly shutdown the server

        if config.repo_method == 'inmemory':

            if config.pyro_enabled():

                # if pyro is enabled, we have to register to the pyro name

                # server, create a pyro daemon, and create a task to handle pyro

                # requests

                self.pyro_daemon = self.appli.repo.pyro_register()

                self.pyro_listen_timeout = 0.02

                self.appli.repo.looping_task(1, self.pyro_loop_event)

            if config.mode != 'test':

                reactor.addSystemEventTrigger('before', 'shutdown',

                                              self.shutdown_event)

                self.appli.repo.start_looping_tasks()

        self.set_url_rewriter()

        CW_EVENT_MANAGER.bind('after-registry-reload', self.set_url_rewriter)

    def start_service(self):

        start_task(self.appli.session_handler.clean_sessions_interval,

                   self.appli.session_handler.clean_sessions)

    def set_url_rewriter(self):

        self.url_rewriter = self.appli.vreg['components'].select_or_none('urlrewriter')

    def shutdown_event(self):

        """callback fired when the server is shutting down to properly

        clean opened sessions

        """

        self.appli.repo.shutdown()

    def pyro_loop_event(self):

        """listen for pyro events"""

        try:

            self.pyro_daemon.handleRequests(self.pyro_listen_timeout)

        except select.error:

            return

    def getChild(self, path, request):

        """Indicate which resource to use to process down the URL's path"""

        return self

    def render(self, request):

        """Render a page from the root resource"""

        # reload modified files in debug mode

        if self.config.debugmode:

            self.config.uiprops.reload_if_needed()

            if self.https_url:

                self.config.https_uiprops.reload_if_needed()

            self.appli.vreg.reload_if_needed()

        if self.config['profile']: # default profiler don't trace threads

            return self.render_request(request)

        else:

            deferred = threads.deferToThread(self.render_request, request)

            return NOT_DONE_YET

    def render_request(self, request):

        try:

            # processing HUGE files (hundred of megabytes) in http.processReceived

            # blocks other HTTP requests processing

            # due to the clumsy & slow parsing algorithm of cgi.FieldStorage

            # so we deferred that part to the cubicweb thread

            request.process_multipart()

            return self._render_request(request)

        except Exception:

            errorstream = StringIO()

            traceback.print_exc(file=errorstream)

            return HTTPResponse(stream='<pre>%s</pre>' % errorstream.getvalue(),

                                code=500, twisted_request=request)

    def _render_request(self, request):

        origpath = request.path

        host = request.host

        # dual http/https access handling: expect a rewrite rule to prepend

        # 'https' to the path to detect https access

        if origpath.split('/', 2)[1] == 'https':

            origpath = origpath[6:]

            request.uri = request.uri[6:]

            https = True

            baseurl = self.https_url or self.base_url

        else:

            https = False

            baseurl = self.base_url

        if self.config['use-request-subdomain']:

            baseurl = host_prefixed_baseurl(baseurl, host)

            self.warning('used baseurl is %s for this request', baseurl)

        req = CubicWebTwistedRequestAdapter(request, self.appli.vreg, https, baseurl)

        if req.authmode == 'http':

            # activate realm-based auth

            realm = self.config['realm']

            req.set_header('WWW-Authenticate', [('Basic', {'realm' : realm })], raw=False)

        try:

            self.appli.connect(req)

        except Redirect, ex:

            return self.redirect(request=req, location=ex.location)

        if https and req.session.anonymous_session:

            # don't allow anonymous on https connection

            return self.request_auth(request=req)

        if self.url_rewriter is not None:

            # XXX should occur before authentication?

            try:

                path = self.url_rewriter.rewrite(host, origpath, req)

            except Redirect, ex:

                return self.redirect(req, ex.location)

            request.uri.replace(origpath, path, 1)

        else:

            path = origpath

        if not path or path == "/":

            path = 'view'

        try:

            result = self.appli.publish(path, req)

        except DirectResponse, ex:

            return ex.response

        except StatusResponse, ex:

            return HTTPResponse(stream=ex.content, code=ex.status,

                                twisted_request=req._twreq,

                                headers=req.headers_out)

        except AuthenticationError:

            return self.request_auth(request=req)

        except LogOut, ex:

            if self.config['auth-mode'] == 'cookie' and ex.url:

                return self.redirect(request=req, location=ex.url)

            # in http we have to request auth to flush current http auth

            # information

            return self.request_auth(request=req, loggedout=True)

        except Redirect, ex:

            return self.redirect(request=req, location=ex.location)

        # request may be referenced by "onetime callback", so clear its entity

        # cache to avoid memory usage