Source code for girder.utility.acl_mixin

#!/usr/bin/env python
# -*- coding: utf-8 -*-

###############################################################################
#  Copyright 2015 Kitware Inc.
#
#  Licensed under the Apache License, Version 2.0 ( the "License" );
#  you may not use this file except in compliance with the License.
#  You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
#  Unless required by applicable law or agreed to in writing, software
#  distributed under the License is distributed on an "AS IS" BASIS,
#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#  See the License for the specific language governing permissions and
#  limitations under the License.
###############################################################################

from bson.py3compat import abc
import collections
import itertools
import six

from ..models.model_base import Model, AccessControlledModel, _permissionClauses
from ..exceptions import AccessException
from ..constants import AccessType, TEXT_SCORE_SORT_MAX


[docs]class AccessControlMixin(object): """ This mixin is intended to be used for resources which aren't access controlled by default, but resolve their access controls through other resources. As such, the overridden methods retain the same parameters and only alter functionality related to access control resolution. resourceColl corresponds to the resource collection that needs to be used for resolution, for example the Item model has a resourceColl of folder. resourceParent corresponds to the field in which the parent resource belongs, so for an item it would be the folderId. """ resourceColl = None resourceParent = None
[docs] def load(self, id, level=AccessType.ADMIN, user=None, objectId=True, force=False, fields=None, exc=False): """ Calls Model.load on the current item, and then attempts to load the resourceParent which the user must have access to in order to load this model. Takes the same parameters as :py:func:`girder.models.model_base.AccessControlledModel.load`. """ loadFields = fields if not force: extraFields = {'attachedToId', 'attachedToType'} if self.resourceParent: extraFields.add(self.resourceParent) loadFields = self._supplementFields(fields, extraFields) doc = Model.load(self, id=id, objectId=objectId, fields=loadFields, exc=exc) if not force and doc is not None: if doc.get(self.resourceParent): loadType = self.resourceColl loadId = doc[self.resourceParent] else: loadType = doc.get('attachedToType') loadId = doc.get('attachedToId') if isinstance(loadType, six.string_types): self.model(loadType).load(loadId, level=level, user=user, exc=exc) elif isinstance(loadType, list) and len(loadType) == 2: self.model(*loadType).load(loadId, level=level, user=user, exc=exc) else: raise Exception('Invalid model type: %s' % str(loadType)) self._removeSupplementalFields(doc, fields) return doc
[docs] def hasAccess(self, resource, user=None, level=AccessType.READ): """ Determines if a user has access to a resource based on their access to the resourceParent. Takes the same parameters as :py:func:`girder.models.model_base.AccessControlledModel.hasAccess`. """ resource = self.model(self.resourceColl) \ .load(resource[self.resourceParent], force=True) return self.model(self.resourceColl).hasAccess( resource, user=user, level=level)
[docs] def hasAccessFlags(self, doc, user=None, flags=None): """ See the documentation of AccessControlledModel.hasAccessFlags, which this wraps. """ if not flags: return True resource = self.model(self.resourceColl).load(doc[self.resourceParent], force=True) return self.model(self.resourceColl).hasAccessFlags(resource, user, flags)
[docs] def requireAccess(self, doc, user=None, level=AccessType.READ): """ This wrapper just provides a standard way of throwing an access denied exception if the access check fails. """ if not self.hasAccess(doc, user, level): if level == AccessType.READ: perm = 'Read' elif level == AccessType.WRITE: perm = 'Write' elif level in (AccessType.ADMIN, AccessType.SITE_ADMIN): perm = 'Admin' else: perm = 'Unknown level' if user: userid = str(user.get('_id', '')) else: userid = None raise AccessException("%s access denied for %s %s (user %s)." % (perm, self.name, doc.get('_id', 'unknown'), userid))
[docs] def requireAccessFlags(self, doc, user=None, flags=None): """ See the documentation of AccessControlledModel.requireAccessFlags, which this wraps. """ if not flags: return resource = self.model(self.resourceColl).load(doc[self.resourceParent], force=True) return self.model(self.resourceColl).requireAccessFlags(resource, user, flags)
[docs] def filterResultsByPermission(self, cursor, user, level, limit=0, offset=0, removeKeys=(), flags=None): """ Yields filtered results from the cursor based on the access control existing for the resourceParent. Takes the same parameters as :py:func:`girder.models.model_base.AccessControlledModel.filterResultsByPermission`. """ # Cache mapping resourceIds -> access granted (bool) resourceAccessCache = {} def hasAccess(_result): resourceId = _result[self.resourceParent] # return cached check if it exists if resourceId in resourceAccessCache: return resourceAccessCache[resourceId] # if the resourceId is not cached, check for permission "level" # and set the cache resource = self.model(self.resourceColl).load(resourceId, force=True) val = self.model(self.resourceColl).hasAccess( resource, user=user, level=level) if flags: val = val and self.model(self.resourceColl).hasAccessFlags( resource, user=user, flags=flags) resourceAccessCache[resourceId] = val return val endIndex = offset + limit if limit else None filteredCursor = six.moves.filter(hasAccess, cursor) for result in itertools.islice(filteredCursor, offset, endIndex): for key in removeKeys: if key in result: del result[key] yield result
def textSearch(self, query, user=None, filters=None, limit=0, offset=0, sort=None, fields=None, level=AccessType.READ): filters, fields = self._textSearchFilters(query, filters, fields) defaultSort = [('_textScore', {'$meta': 'textScore'})] cursor = self.findWithPermissions( filters, offset=offset, limit=limit, sort=sort, fields=fields, user=user, level=level, aggregateSort=defaultSort) if (sort is None and not getattr(cursor, 'fromAggregate', False) and callable(getattr(cursor, 'count', None)) and cursor.count() < TEXT_SCORE_SORT_MAX): cursor = self.findWithPermissions( filters, offset=offset, limit=limit, sort=defaultSort, fields=fields, user=user, level=level) return cursor
[docs] def prefixSearch(self, query, user=None, filters=None, limit=0, offset=0, sort=None, fields=None, level=AccessType.READ, prefixSearchFields=None): """ Custom override of Model.prefixSearch to also force permission-based filtering. The parameters are the same as Model.prefixSearch. :param user: The user to apply permission filtering for. :type user: dict or None :param level: The access level to require. :type level: girder.constants.AccessType """ filters = self._prefixSearchFilters(query, filters, prefixSearchFields) return self.findWithPermissions( filters, offset=offset, limit=limit, sort=sort, fields=fields, user=user, level=level)
def permissionClauses(self, user=None, level=None, prefix=''): return _permissionClauses(user, level, prefix) def _findWithPermissionsFallback(self, query, offset, limit, timeout, fields, sort, user, level, **kwargs): """ See findWithPermissions. This is called if Mongo doesn't support appropriate aggregations or find is used on a model that uses an acl-mixin model as its parent and therefore needs multiple steps to reach the owner of the access control list. See findWithPermissions for parameters and return. """ removeKeys = () if (fields and any(fields[key] is True for key in fields) and not fields.get(self.resourceParent)): fields = fields.copy() fields[self.resourceParent] = True removeKeys = (self.resourceParent, ) cursor = self.find(query, timeout=timeout, fields=fields, sort=sort, **kwargs) result = self.filterResultsByPermission( cursor=cursor, user=user, level=level, limit=limit, offset=offset, removeKeys=removeKeys) if not hasattr(result, 'count'): origResult, origSelf = result, self class resultWithCount(object): def count(self): cursor = origSelf.find( query, timeout=timeout, fields=fields, sort=sort, **kwargs) result = origSelf.filterResultsByPermission( cursor=cursor, user=user, level=level, limit=limit, offset=offset, removeKeys=removeKeys) return len(list(result)) def __iter__(self): return self def __next__(self): return six.next(origResult) next = __next__ result = resultWithCount() return result
[docs] def findWithPermissions(self, query=None, offset=0, limit=0, timeout=None, fields=None, sort=None, user=None, level=AccessType.READ, aggregateSort=None, **kwargs): """ Search the collection by a set of parameters, only returning results that the combined user and level have permission to access. Passes any extra kwargs through to the underlying pymongo.collection.find function. :param query: The search query (see general MongoDB docs for "find()") :type query: dict :param offset: The offset into the results :type offset: int :param limit: Maximum number of documents to return :type limit: int :param timeout: Cursor timeout in ms. Default is no timeout. :type timeout: int :param fields: A mask for filtering result documents by key, or None to return the full document, passed to MongoDB find() as the `projection` param. :type fields: `str, list of strings or tuple of strings for fields to be included from the document, or dict for an inclusion or exclusion projection`. :param sort: The sort order. :type sort: List of (key, order) tuples. :param user: The user to check policies against. :type user: dict or None :param level: The access level. Explicitly passing None skips doing permissions checks. :type level: AccessType :param aggregateSort: A sort order to use if `sort` is None and an aggregation is used. :type aggregateSort: List of (key, order) tuples. :returns: A pymongo Cursor, CommandCursor, or an iterable. If a CommandCursor, it has been augmented with a count function. """ if level is not None and (not user or not user['admin']): # If the resourceColl isn't an access controlled model that we # know how to reach, fall back to performing the ordinary query and # then filtering it by permission. For instance, if a model uses # the acl mixin to get acl from a model that itself uses the acl # mixin, this will return the correct results, but without the # utility of being able perform count(). # Note, this also handles models which use attachedToType and # attachedToId, since self.model(None) will not be an access # controlled model. # This is also the fall-back for Mongo < 3.4, as those versions do # not support the aggregation steps that are used. if (not isinstance(self.model(self.resourceColl), AccessControlledModel) or not getattr(self, '_dbserver_version', None) or getattr(self, '_dbserver_version', None) < (3, 4)): return self._findWithPermissionsFallback( query, offset, limit, timeout, fields, sort, user, level, **kwargs) query = query or {} initialPipeline = [ {'$match': query}, {'$lookup': { 'from': self.resourceColl, 'localField': self.resourceParent, 'foreignField': '_id', 'as': '__parent' }}, {'$match': self.permissionClauses(user, level, '__parent.')}, ] countPipeline = initialPipeline + [ {'$count': 'count'}, ] fullPipeline = initialPipeline + [ {'$project': {'__parent': False}}, ] if sort is not None or aggregateSort is not None: fullPipeline.append({'$sort': collections.OrderedDict(sort or aggregateSort)}) # limit should immediately follow sort for efficiency if limit: fullPipeline.append({'$limit': limit + (offset or 0)}) if offset: fullPipeline.append({'$skip': offset}) if fields is not None: # fields can be a Sequence, Set, or Mapping. If a Mapping, the # values are typically booleans or themselves a mapping (such # as from text search to add a field like _textScore: {$meta: # 'textScore'}). Convert sequences and sets to mappings (as # done in pymongo), then use values that aren't themselves # mappings as a projection and those that are mappings as # added fields. if isinstance(fields, (abc.Sequence, abc.Set)): fields = dict.fromkeys(fields, 1) if any(not isinstance(v, abc.Mapping) for v in fields.values()): fullPipeline.append({'$project': { k: v for k, v in six.iteritems(fields) if not isinstance(v, abc.Mapping)}}) if any(isinstance(v, abc.Mapping) for v in fields.values()): fullPipeline.append({'$addFields': { k: v for k, v in six.iteritems(fields) if isinstance(v, abc.Mapping)}}) options = { # By allowing disk use, large sorted queries will work. If # disallowed, they will fail. Although this is slower than # memory sorting, actual experiemnts show it to be acceptable 'allowDiskUse': True, # Start with a 0-sized batch. This avoids fetching data from # the Mongo server if the query is never polled and starts # streaming data faster than a fixed batch size. 'cursor': {'batchSize': 0} } if timeout: options['maxTimeMS'] = timeout result = self.collection.aggregate(fullPipeline, **options) def count(): try: return next(iter(self.collection.aggregate(countPipeline, **options)))['count'] except StopIteration: # If there are no values, this won't return the count, in # which case it is zero. return 0 result.count = count # Mark that this result came from an aggregate. If an aggregate # is used, the results could be sorted via the aggregateSort # parameter. This informs the consumer of the result. result.fromAggregate = True return result return self.find(query, offset, limit, timeout, fields, sort, **kwargs)