view sat/plugins/plugin_comp_file_sharing.py @ 3370:2157880ba3b4

core: plugins can now be loaded from a directory: A plugin can now be a directory with a name starting by `plugin_`. It will then be loaded as a module. The `__init__.py` file must contain the `PLUGIN_INFO` dict with metadata.
author Goffi <goffi@goffi.org>
date Sun, 20 Sep 2020 14:05:40 +0200
parents 90aee1f2d07c
children be6d91572633
line wrap: on
line source

#!/usr/bin/env python3

# SAT plugin for parrot mode (experimental)
# Copyright (C) 2009-2020 Jérôme Poisson (goffi@goffi.org)

# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero 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 Affero General Public License for more details.

# You should have received a copy of the GNU Affero General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.

import os
import os.path
import mimetypes
import tempfile
from functools import partial
import shortuuid
import unicodedata
from urllib.parse import urljoin, urlparse, quote, unquote
from pathlib import Path
from sat.core.i18n import _
from sat.core.constants import Const as C
from sat.core import exceptions
from sat.core.log import getLogger
from sat.tools import stream
from sat.tools import video
from sat.tools.common import regex
from sat.tools.common import uri
from sat.tools.common import files_utils
from sat.tools.common import tls
from twisted.internet import defer, reactor
from twisted.words.protocols.jabber import error
from twisted.web import server, resource, static, http
from wokkel import pubsub
from wokkel import generic


log = getLogger(__name__)


PLUGIN_INFO = {
    C.PI_NAME: "File sharing component",
    C.PI_IMPORT_NAME: "file_sharing",
    C.PI_MODES: [C.PLUG_MODE_COMPONENT],
    C.PI_TYPE: C.PLUG_TYPE_ENTRY_POINT,
    C.PI_PROTOCOLS: [],
    C.PI_DEPENDENCIES: [
        "FILE",
        "FILE_SHARING_MANAGEMENT",
        "XEP-0106",
        "XEP-0234",
        "XEP-0260",
        "XEP-0261",
        "XEP-0264",
        "XEP-0329",
        "XEP-0363",
    ],
    C.PI_RECOMMENDATIONS: [],
    C.PI_MAIN: "FileSharing",
    C.PI_HANDLER: C.BOOL_TRUE,
    C.PI_DESCRIPTION: _("""Component hosting and sharing files"""),
}

HASH_ALGO = "sha-256"
NS_COMMENTS = "org.salut-a-toi.comments"
NS_FS_AFFILIATION = "org.salut-a-toi.file-sharing-affiliation"
COMMENT_NODE_PREFIX = "org.salut-a-toi.file_comments/"
# Directory used to buffer request body (i.e. file in case of PUT) we use more than one @
# there, to be sure than it's not conflicting with a JID
TMP_BUFFER_DIR = "@@tmp@@"

server.version = unicodedata.normalize(
    'NFKD',
    f"{C.APP_NAME} file sharing {C.APP_VERSION}"
).encode('ascii','ignore')


class HTTPFileServer(resource.Resource):
    isLeaf = True

    def errorPage(self, request, code):
        request.setResponseCode(code)
        if code == http.BAD_REQUEST:
            brief = 'Bad Request'
            details = "Your request is invalid"
        elif code == http.FORBIDDEN:
            brief = 'Forbidden'
            details = "You're not allowed to use this resource"
        elif code == http.NOT_FOUND:
            brief = 'Not Found'
            details = "No resource found at this URL"
        else:
            brief = 'Error'
            details = "This resource can't be used"
            log.error(f"Unexpected return code used: {code}")
        log.warning(
            f'Error returned while trying to access url {request.uri.decode()}: '
            f'"{brief}" ({code}): {details}'
        )

        return resource.ErrorPage(code, brief, details).render(request)

    def getDispositionType(self, media_type, media_subtype):
        if media_type in ('image', 'video'):
            return 'inline'
        elif media_type == 'application' and media_subtype == 'pdf':
            return 'inline'
        else:
            return 'attachment'

    def render(self, request):
        request.setHeader("Access-Control-Allow-Origin", "*")
        request.setHeader("Access-Control-Allow-Methods", "OPTIONS, HEAD, GET, PUT")
        request.setHeader(
            "Access-Control-Allow-Headers",
            "Content-Type, Range, Xmpp-File-Path, Xmpp-File-No-Http")
        request.setHeader("Access-Control-Allow-Credentials", "true")
        request.setHeader("Accept-Ranges", "bytes")

        request.setHeader(
            "Access-Control-Expose-Headers",
            "Date, Content-Length, Content-Range")
        return super().render(request)

    def render_OPTIONS(self, request):
        request.setResponseCode(http.OK)
        return b""

    def render_GET(self, request):
        try:
            request.upload_data
        except exceptions.DataError:
            return self.errorPage(request, http.NOT_FOUND)

        defer.ensureDeferred(self.renderGet(request))
        return server.NOT_DONE_YET

    async def renderGet(self, request):
        try:
            upload_id, filename = request.upload_data
        except exceptions.DataError:
            request.write(self.errorPage(request, http.FORBIDDEN))
            request.finish()
            return
        found_files = await request.file_sharing.host.memory.getFiles(
            client=None, peer_jid=None, perms_to_check=None, public_id=upload_id)
        if not found_files:
            request.write(self.errorPage(request, http.NOT_FOUND))
            request.finish()
            return
        if len(found_files) > 1:
            log.error(f"more that one files found for public id {upload_id!r}")

        found_file = found_files[0]
        file_path = request.file_sharing.files_path/found_file['file_hash']
        file_res = static.File(file_path)
        file_res.type = f'{found_file["media_type"]}/{found_file["media_subtype"]}'
        file_res.encoding = file_res.contentEncodings.get(Path(found_file['name']).suffix)
        disp_type = self.getDispositionType(
            found_file['media_type'], found_file['media_subtype'])
        # the URL is percent encoded, and not all browsers/tools unquote the file name,
        # thus we add a content disposition header
        request.setHeader(
            'Content-Disposition',
            f"{disp_type}; filename*=UTF-8''{quote(found_file['name'])}"
        )
        # cf. https://xmpp.org/extensions/xep-0363.html#server
        request.setHeader(
            'Content-Security-Policy',
            "default-src 'none'; frame-ancestors 'none';"
        )
        ret = file_res.render(request)
        if ret != server.NOT_DONE_YET:
            # HEAD returns directly the result (while GET use a produced)
            request.write(ret)
            request.finish()

    def render_PUT(self, request):
        defer.ensureDeferred(self.renderPut(request))
        return server.NOT_DONE_YET

    async def renderPut(self, request):
        try:
            client, upload_request = request.upload_request_data
            upload_id, filename = request.upload_data
        except AttributeError:
            request.write(self.errorPage(request, http.BAD_REQUEST))
            request.finish()
            return

        # at this point request is checked and file is buffered, we can store it
        # we close the content here, before registering the file
        request.content.close()
        tmp_file_path = Path(request.content.name)
        request.content = None

        # the 2 following headers are not standard, but useful in the context of file
        # sharing with HTTP Upload: first one allow uploaded to specify the path
        # and second one will disable public exposure of the file through HTTP
        path = request.getHeader("Xmpp-File-Path")
        if path:
            path = unquote(path)
        else:
            path =  "/uploads"
        if request.getHeader("Xmpp-File-No-Http") is not None:
            public_id = None
        else:
            public_id = upload_id

        file_data = {
            "name": unquote(upload_request.filename),
            "mime_type": upload_request.content_type,
            "size": upload_request.size,
            "path": path
        }

        await request.file_sharing.registerReceivedFile(
            client, upload_request.from_, file_data, tmp_file_path,
            public_id=public_id,
        )

        request.setResponseCode(http.CREATED)
        request.finish()


class FileSharingRequest(server.Request):

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self._upload_data = None

    @property
    def upload_data(self):
        """A tuple with upload_id and filename retrieve from requested path"""
        if self._upload_data is not None:
            return self._upload_data

        # self.path is not available if we are easly in the request (e.g. when gotLength
        # is called), in which case channel._path must be used. On the other hand, when
        # render_[VERB] is called, only self.path is available
        path = self.channel._path if self.path is None else self.path
        # we normalise the path
        path = urlparse(path.decode()).path
        try:
            __, upload_id, filename = path.split('/')
        except ValueError:
            raise exceptions.DataError("no enought path elements")
        if len(upload_id) < 10:
            raise exceptions.DataError(f"invalid upload ID received for a PUT: {upload_id!r}")

        self._upload_data = (upload_id, filename)
        return self._upload_data

    @property
    def file_sharing(self):
        return self.channel.site.file_sharing

    @property
    def file_tmp_dir(self):
        return self.channel.site.file_tmp_dir

    def refuseRequest(self):
        if self.content is not None:
            self.content.close()
        self.content = open(os.devnull, 'w+b')
        self.channel._respondToBadRequestAndDisconnect()

    def gotLength(self, length):
        if self.channel._command.decode().upper() == 'PUT':
            # for PUT we check early if upload_id is fine, to avoid buffering a file we'll refuse
            # we buffer the file in component's TMP_BUFFER_DIR, so we just have to rename it at the end
            try:
                upload_id, filename = self.upload_data
            except exceptions.DataError as e:
                log.warning(f"Invalid PUT request, we stop here: {e}")
                return self.refuseRequest()
            try:
                client, upload_request, timer = self.file_sharing.expected_uploads.pop(upload_id)
            except KeyError:
                log.warning(f"unknown (expired?) upload ID received for a PUT: {upload_id!r}")
                return self.refuseRequest()

            if not timer.active:
                log.warning(f"upload id {upload_id!r} used for a PUT, but it is expired")
                return self.refuseRequest()

            timer.cancel()

            if upload_request.filename != filename:
                log.warning(
                    f"invalid filename for PUT (upload id: {upload_id!r}, URL: {self.channel._path.decode()}). Original "
                    f"{upload_request.filename!r} doesn't match {filename!r}"
                )
                return self.refuseRequest()

            self.upload_request_data = (client, upload_request)

            file_tmp_path = files_utils.get_unique_name(
                self.file_tmp_dir/upload_id)

            self.content = open(file_tmp_path, 'w+b')
        else:
            return super().gotLength(length)


class FileSharingSite(server.Site):
    requestFactory = FileSharingRequest

    def __init__(self, file_sharing):
        self.file_sharing = file_sharing
        self.file_tmp_dir = file_sharing.host.getLocalPath(
            None, C.FILES_TMP_DIR, TMP_BUFFER_DIR, component=True, profile=False
        )
        for old_file in self.file_tmp_dir.iterdir():
            log.debug(f"purging old buffer file at {old_file}")
            old_file.unlink()
        super().__init__(HTTPFileServer())

    def getContentFile(self, length):
        file_tmp_path = self.file_tmp_dir/shortuuid.uuid()
        return open(file_tmp_path, 'w+b')


class FileSharing:

    def __init__(self, host):
        self.host = host
        self.initialised = False

    def init(self):
        # we init once on first component connection,
        # there is not need to init this plugin if not component use it
        # TODO: this plugin should not be loaded at all if no component uses it
        #   and should be loaded dynamically as soon as a suitable profile is created
        if self.initialised:
            return
        self.initialised = True
        log.info(_("File Sharing initialization"))
        self._f = self.host.plugins["FILE"]
        self._jf = self.host.plugins["XEP-0234"]
        self._h = self.host.plugins["XEP-0300"]
        self._t = self.host.plugins["XEP-0264"]
        self.host.plugins["XEP-0363"].registerHandler(self._onHTTPUpload)
        self.host.trigger.add("FILE_getDestDir", self._getDestDirTrigger)
        self.host.trigger.add(
            "XEP-0234_fileSendingRequest", self._fileSendingRequestTrigger, priority=1000
        )
        self.host.trigger.add("XEP-0234_buildFileElement", self._addFileMetadataElts)
        self.host.trigger.add("XEP-0234_parseFileElement", self._getFileMetadataElts)
        self.host.trigger.add("XEP-0329_compGetFilesFromNode", self._addFileMetadata)
        self.host.trigger.add(
            "XEP-0329_compGetFilesFromNode_build_directory",
            self._addDirectoryMetadataElts)
        self.host.trigger.add(
            "XEP-0329_parseResult_directory",
            self._getDirectoryMetadataElts)
        self.files_path = self.host.getLocalPath(None, C.FILES_DIR, profile=False)
        self.http_port = self.host.memory.getConfig(
            'component file_sharing', 'http_upload_port', 8888)
        connection_type = self.host.memory.getConfig(
            'component file_sharing', 'http_upload_connection_type', 'https')
        if connection_type not in ('http', 'https'):
            raise exceptions.ConfigError(
                'bad http_upload_connection_type, you must use one of "http" or "https"'
            )
        self.server = FileSharingSite(self)
        self.expected_uploads = {}
        if connection_type == 'http':
            reactor.listenTCP(self.http_port, self.server)
        else:
            options = tls.getOptionsFromConfig(
                self.host.memory.config, "component file_sharing")
            tls.TLSOptionsCheck(options)
            context_factory = tls.getTLSContextFactory(options)
            reactor.listenSSL(self.http_port, self.server, context_factory)

    def getHandler(self, client):
        return Comments_handler(self)

    def profileConnecting(self, client):
        self.init()
        public_base_url = self.host.memory.getConfig(
            'component file_sharing', 'http_upload_public_facing_url')
        if public_base_url is None:
            client._file_sharing_base_url = f"https://{client.host}:{self.http_port}"
        else:
            client._file_sharing_base_url = public_base_url
        path = client.file_tmp_dir = os.path.join(
            self.host.memory.getConfig("", "local_dir"),
            C.FILES_TMP_DIR,
            regex.pathEscape(client.profile),
        )
        if not os.path.exists(path):
            os.makedirs(path)

    async def generate_thumbnails(self, extra: dict, image_path: Path):
        thumbnails = extra.setdefault(C.KEY_THUMBNAILS, [])
        for max_thumb_size in self._t.SIZES:
            try:
                thumb_size, thumb_id = await self._t.generateThumbnail(
                    image_path,
                    max_thumb_size,
                    #  we keep thumbnails for 6 months
                    60 * 60 * 24 * 31 * 6,
                )
            except Exception as e:
                log.warning(_("Can't create thumbnail: {reason}").format(reason=e))
                break
            thumbnails.append({"id": thumb_id, "size": thumb_size})

    async def registerReceivedFile(
            self, client, peer_jid, file_data, file_path, public_id=None, extra=None):
        """Post file reception tasks

        once file is received, this method create hash/thumbnails if necessary
        move the file to the right location, and create metadata entry in database
        """
        name = file_data["name"]
        if extra is None:
            extra = {}

        mime_type = file_data.get("mime_type")
        if not mime_type or mime_type == "application/octet-stream":
            mime_type = mimetypes.guess_type(name)[0]

        is_image = mime_type is not None and mime_type.startswith("image")
        is_video = mime_type is not None and mime_type.startswith("video")

        if file_data.get("hash_algo") == HASH_ALGO:
            log.debug(_("Reusing already generated hash"))
            file_hash = file_data["hash_hasher"].hexdigest()
        else:
            hasher = self._h.getHasher(HASH_ALGO)
            with file_path.open('rb') as f:
                file_hash = await self._h.calculateHash(f, hasher)
        final_path = self.files_path/file_hash

        if final_path.is_file():
            log.debug(
                "file [{file_hash}] already exists, we can remove temporary one".format(
                    file_hash=file_hash
                )
            )
            file_path.unlink()
        else:
            file_path.rename(final_path)
            log.debug(
                "file [{file_hash}] moved to {files_path}".format(
                    file_hash=file_hash, files_path=self.files_path
                )
            )

        if is_image:
            await self.generate_thumbnails(extra, final_path)
        elif is_video:
            with tempfile.TemporaryDirectory() as tmp_dir:
                thumb_path = Path(tmp_dir) / "thumbnail.jpg"
                try:
                    await video.get_thumbnail(final_path, thumb_path)
                except Exception as e:
                    log.warning(_("Can't get thumbnail for {final_path}: {e}").format(
                        final_path=final_path, e=e))
                else:
                    await self.generate_thumbnails(extra, thumb_path)

        self.host.memory.setFile(
            client,
            name=name,
            version="",
            file_hash=file_hash,
            hash_algo=HASH_ALGO,
            size=file_data["size"],
            path=file_data.get("path"),
            namespace=file_data.get("namespace"),
            mime_type=mime_type,
            public_id=public_id,
            owner=peer_jid,
            extra=extra,
        )

    def _getDestDirTrigger(
        self, client, peer_jid, transfer_data, file_data, stream_object
    ):
        """This trigger accept file sending request, and store file locally"""
        if not client.is_component:
            return True, None
        assert stream_object
        assert "stream_object" not in transfer_data
        assert C.KEY_PROGRESS_ID in file_data
        filename = file_data["name"]
        assert filename and not "/" in filename
        file_tmp_dir = self.host.getLocalPath(
            None, C.FILES_TMP_DIR, peer_jid.userhost(), component=True, profile=False
        )
        file_tmp_path = file_data['file_path'] = files_utils.get_unique_name(
            file_tmp_dir/filename)

        transfer_data["finished_d"].addCallback(
            lambda __: defer.ensureDeferred(
                self.registerReceivedFile(client, peer_jid, file_data, file_tmp_path)
            )
        )

        self._f.openFileWrite(
            client, file_tmp_path, transfer_data, file_data, stream_object
        )
        return False, defer.succeed(True)

    @defer.inlineCallbacks
    def _retrieveFiles(
        self, client, session, content_data, content_name, file_data, file_elt
    ):
        """This method retrieve a file on request, and send if after checking permissions"""
        peer_jid = session["peer_jid"]
        owner = client.getOwnerFromJid(session['local_jid'])
        try:
            found_files = yield self.host.memory.getFiles(
                client,
                peer_jid=peer_jid,
                name=file_data.get("name"),
                file_hash=file_data.get("file_hash"),
                hash_algo=file_data.get("hash_algo"),
                path=file_data.get("path"),
                namespace=file_data.get("namespace"),
                owner=owner,
            )
        except exceptions.NotFound:
            found_files = None
        except exceptions.PermissionError:
            log.warning(
                _("{peer_jid} is trying to access an unauthorized file: {name}").format(
                    peer_jid=peer_jid, name=file_data.get("name")
                )
            )
            defer.returnValue(False)

        if not found_files:
            log.warning(
                _("no matching file found ({file_data})").format(file_data=file_data)
            )
            defer.returnValue(False)

        # we only use the first found file
        found_file = found_files[0]
        if found_file['type'] != C.FILE_TYPE_FILE:
            raise TypeError("a file was expected, type is {type_}".format(
                type_=found_file['type']))
        file_hash = found_file["file_hash"]
        file_path = self.files_path / file_hash
        file_data["hash_hasher"] = hasher = self._h.getHasher(found_file["hash_algo"])
        size = file_data["size"] = found_file["size"]
        file_data["file_hash"] = file_hash
        file_data["hash_algo"] = found_file["hash_algo"]

        # we complete file_elt so peer can have some details on the file
        if "name" not in file_data:
            file_elt.addElement("name", content=found_file["name"])
        file_elt.addElement("size", content=str(size))
        content_data["stream_object"] = stream.FileStreamObject(
            self.host,
            client,
            file_path,
            uid=self._jf.getProgressId(session, content_name),
            size=size,
            data_cb=lambda data: hasher.update(data),
        )
        defer.returnValue(True)

    def _fileSendingRequestTrigger(
        self, client, session, content_data, content_name, file_data, file_elt
    ):
        if not client.is_component:
            return True, None
        else:
            return (
                False,
                self._retrieveFiles(
                    client, session, content_data, content_name, file_data, file_elt
                ),
            )

    ## HTTP Upload ##

    def _purge_slot(self, upload_id):
        try:
            del self.expected_uploads[upload_id]
        except KeyError:
            log.error(f"trying to purge an inexisting upload slot ({upload_id})")

    def _onHTTPUpload(self, client, request):
        # filename should be already cleaned, but it's better to double check
        assert '/' not in request.filename
        # client._file_sharing_allowed_hosts is set in plugin XEP-0329
        if request.from_.host not in client._file_sharing_allowed_hosts:
            raise error.StanzaError("forbidden")

        upload_id = shortuuid.ShortUUID().random(length=30)
        assert '/' not in upload_id
        timer = reactor.callLater(30, self._purge_slot, upload_id)
        self.expected_uploads[upload_id] = (client, request, timer)
        url = urljoin(client._file_sharing_base_url, f"{upload_id}/{request.filename}")
        slot = self.host.plugins["XEP-0363"].Slot(
            put=url,
            get=url,
            headers=[],
        )
        return slot

    ## metadata triggers ##

    def _addFileMetadataElts(self, client, file_elt, extra_args):
        # affiliation
        affiliation = extra_args.get('affiliation')
        if affiliation is not None:
            file_elt.addElement((NS_FS_AFFILIATION, "affiliation"), content=affiliation)

        # comments
        try:
            comments_url = extra_args.pop("comments_url")
        except KeyError:
            return

        comment_elt = file_elt.addElement((NS_COMMENTS, "comments"), content=comments_url)

        try:
            count = len(extra_args["extra"]["comments"])
        except KeyError:
            count = 0

        comment_elt["count"] = str(count)
        return True

    def _getFileMetadataElts(self, client, file_elt, file_data):
        # affiliation
        try:
            affiliation_elt = next(file_elt.elements(NS_FS_AFFILIATION, "affiliation"))
        except StopIteration:
            pass
        else:
            file_data["affiliation"] = str(affiliation_elt)

        # comments
        try:
            comments_elt = next(file_elt.elements(NS_COMMENTS, "comments"))
        except StopIteration:
            pass
        else:
            file_data["comments_url"] = str(comments_elt)
            file_data["comments_count"] = comments_elt["count"]
        return True

    def _addFileMetadata(
            self, client, iq_elt, iq_result_elt, owner, node_path, files_data):
        for file_data in files_data:
            file_data["comments_url"] = uri.buildXMPPUri(
                "pubsub",
                path=client.jid.full(),
                node=COMMENT_NODE_PREFIX + file_data["id"],
            )
        return True

    def _addDirectoryMetadataElts(
            self, client, file_data, directory_elt, owner, node_path):
        affiliation = file_data.get('affiliation')
        if affiliation is not None:
            directory_elt.addElement(
                (NS_FS_AFFILIATION, "affiliation"),
                content=affiliation
            )

    def _getDirectoryMetadataElts(
            self, client, elt, file_data):
        try:
            affiliation_elt = next(elt.elements((NS_FS_AFFILIATION, "affiliation")))
        except StopIteration:
            pass
        else:
            file_data['affiliation'] = str(affiliation_elt)


class Comments_handler(pubsub.PubSubService):
    """This class is a minimal Pubsub service handling virtual nodes for comments"""

    def __init__(self, plugin_parent):
        super(Comments_handler, self).__init__()
        self.host = plugin_parent.host
        self.plugin_parent = plugin_parent
        self.discoIdentity = {
            "category": "pubsub",
            "type": "virtual",  # FIXME: non standard, here to avoid this service being considered as main pubsub one
            "name": "files commenting service",
        }

    def _getFileId(self, nodeIdentifier):
        if not nodeIdentifier.startswith(COMMENT_NODE_PREFIX):
            raise error.StanzaError("item-not-found")
        file_id = nodeIdentifier[len(COMMENT_NODE_PREFIX) :]
        if not file_id:
            raise error.StanzaError("item-not-found")
        return file_id

    @defer.inlineCallbacks
    def getFileData(self, requestor, nodeIdentifier):
        file_id = self._getFileId(nodeIdentifier)
        try:
            files = yield self.host.memory.getFiles(self.parent, requestor, file_id)
        except (exceptions.NotFound, exceptions.PermissionError):
            # we don't differenciate between NotFound and PermissionError
            # to avoid leaking information on existing files
            raise error.StanzaError("item-not-found")
        if not files:
            raise error.StanzaError("item-not-found")
        if len(files) > 1:
            raise error.InternalError("there should be only one file")
        defer.returnValue(files[0])

    def commentsUpdate(self, extra, new_comments, peer_jid):
        """update comments (replace or insert new_comments)

        @param extra(dict): extra data to update
        @param new_comments(list[tuple(unicode, unicode, unicode)]): comments to update or insert
        @param peer_jid(unicode, None): bare jid of the requestor, or None if request is done by owner
        """
        current_comments = extra.setdefault("comments", [])
        new_comments_by_id = {c[0]: c for c in new_comments}
        updated = []
        # we now check every current comment, to see if one id in new ones
        # exist, in which case we must update
        for idx, comment in enumerate(current_comments):
            comment_id = comment[0]
            if comment_id in new_comments_by_id:
                # a new comment has an existing id, update is requested
                if peer_jid and comment[1] != peer_jid:
                    # requestor has not the right to modify the comment
                    raise exceptions.PermissionError
                # we replace old_comment with updated one
                new_comment = new_comments_by_id[comment_id]
                current_comments[idx] = new_comment
                updated.append(new_comment)

        # we now remove every updated comments, to only keep
        # the ones to insert
        for comment in updated:
            new_comments.remove(comment)

        current_comments.extend(new_comments)

    def commentsDelete(self, extra, comments):
        try:
            comments_dict = extra["comments"]
        except KeyError:
            return
        for comment in comments:
            try:
                comments_dict.remove(comment)
            except ValueError:
                continue

    def _getFrom(self, item_elt):
        """retrieve published of an item

        @param item_elt(domish.element): <item> element
        @return (unicode): full jid as string
        """
        iq_elt = item_elt
        while iq_elt.parent != None:
            iq_elt = iq_elt.parent
        return iq_elt["from"]

    @defer.inlineCallbacks
    def publish(self, requestor, service, nodeIdentifier, items):
        #  we retrieve file a first time to check authorisations
        file_data = yield self.getFileData(requestor, nodeIdentifier)
        file_id = file_data["id"]
        comments = [(item["id"], self._getFrom(item), item.toXml()) for item in items]
        if requestor.userhostJID() == file_data["owner"]:
            peer_jid = None
        else:
            peer_jid = requestor.userhost()
        update_cb = partial(self.commentsUpdate, new_comments=comments, peer_jid=peer_jid)
        try:
            yield self.host.memory.fileUpdate(file_id, "extra", update_cb)
        except exceptions.PermissionError:
            raise error.StanzaError("not-authorized")

    @defer.inlineCallbacks
    def items(self, requestor, service, nodeIdentifier, maxItems, itemIdentifiers):
        file_data = yield self.getFileData(requestor, nodeIdentifier)
        comments = file_data["extra"].get("comments", [])
        if itemIdentifiers:
            defer.returnValue(
                [generic.parseXml(c[2]) for c in comments if c[0] in itemIdentifiers]
            )
        else:
            defer.returnValue([generic.parseXml(c[2]) for c in comments])

    @defer.inlineCallbacks
    def retract(self, requestor, service, nodeIdentifier, itemIdentifiers):
        file_data = yield self.getFileData(requestor, nodeIdentifier)
        file_id = file_data["id"]
        try:
            comments = file_data["extra"]["comments"]
        except KeyError:
            raise error.StanzaError("item-not-found")

        to_remove = []
        for comment in comments:
            comment_id = comment[0]
            if comment_id in itemIdentifiers:
                to_remove.append(comment)
                itemIdentifiers.remove(comment_id)
                if not itemIdentifiers:
                    break

        if itemIdentifiers:
            # not all items have been to_remove, we can't continue
            raise error.StanzaError("item-not-found")

        if requestor.userhostJID() != file_data["owner"]:
            if not all([c[1] == requestor.userhost() for c in to_remove]):
                raise error.StanzaError("not-authorized")

        remove_cb = partial(self.commentsDelete, comments=to_remove)
        yield self.host.memory.fileUpdate(file_id, "extra", remove_cb)