import logging from collections import defaultdict from itertools import chain from os import path from string import Template from urllib.parse import urldefrag import arrow from rdflib import Dataset, Graph, Literal, URIRef, plugin from rdflib.namespace import RDF from rdflib.query import ResultException from rdflib.resource import Resource from rdflib.store import Store from lakesuperior import basedir, env from lakesuperior.dictionaries.namespaces import ns_collection as nsc from lakesuperior.dictionaries.namespaces import ns_mgr as nsm from lakesuperior.dictionaries.srv_mgd_terms import srv_mgd_subjects, \ srv_mgd_predicates, srv_mgd_types from lakesuperior.exceptions import (InvalidResourceError, ResourceNotExistsError, TombstoneError, PathSegmentError) from lakesuperior.env import env from lakesuperior.store.ldp_rs.lmdb_store import TxnManager META_GR_URI = nsc['fcsystem']['meta'] HIST_GR_URI = nsc['fcsystem']['histmeta'] PTREE_GR_URI = nsc['fcsystem']['pairtree'] VERS_CONT_LABEL = 'fcr:versions' Lmdb = plugin.register('Lmdb', Store, 'lakesuperior.store.ldp_rs.lmdb_store', 'LmdbStore') logger = logging.getLogger(__name__) class RsrcCentricLayout: """ This class exposes an interface to build graph store layouts. It also provides the basics of the triplestore connection. Some store layouts are provided. New ones aimed at specific uses and optimizations of the repository may be developed by extending this class and implementing all its abstract methods. A layout is implemented via application configuration. However, once contents are ingested in a repository, changing a layout will most likely require a migration. The custom layout must be in the lakesuperior.store.rdf package and the class implementing the layout must be called `StoreLayout`. The module name is the one defined in the app configuration. E.g. if the configuration indicates `simple_layout` the application will look for `lakesuperior.store.rdf.simple_layout.SimpleLayout`. """ _graph_uids = ('fcadmin', 'fcmain', 'fcstruct') # @TODO Move to a config file? attr_map = { nsc['fcadmin']: { # List of server-managed predicates. Triples bearing one of these # predicates will go in the metadata graph. 'p': { nsc['ebucore'].hasMimeType, nsc['fcrepo'].created, nsc['fcrepo'].createdBy, nsc['fcrepo'].hasParent, nsc['fcrepo'].hasVersion, nsc['fcrepo'].lastModified, nsc['fcrepo'].lastModifiedBy, nsc['fcsystem'].tombstone, # The following 3 are set by the user but still in this group # for convenience. nsc['ldp'].membershipResource, nsc['ldp'].hasMemberRelation, nsc['ldp'].insertedContentRelation, nsc['iana'].describedBy, nsc['premis'].hasMessageDigest, nsc['premis'].hasSize, }, # List of metadata RDF types. Triples bearing one of these types in # the object will go in the metadata graph. 't': { nsc['fcrepo'].Binary, nsc['fcrepo'].Container, nsc['fcrepo'].Pairtree, nsc['fcrepo'].Resource, nsc['fcsystem'].Tombstone, nsc['ldp'].BasicContainer, nsc['ldp'].Container, nsc['ldp'].DirectContainer, nsc['ldp'].IndirectContainer, nsc['ldp'].NonRDFSource, nsc['ldp'].RDFSource, nsc['ldp'].Resource, }, }, nsc['fcstruct']: { # These are placed in a separate graph for optimization purposes. 'p': { nsc['ldp'].contains, nsc['pcdm'].hasMember, } }, } # RDF types of graphs by prefix. graph_ns_types = { nsc['fcadmin']: nsc['fcsystem'].AdminGraph, nsc['fcmain']: nsc['fcsystem'].UserProvidedGraph, nsc['fcstruct']: nsc['fcsystem'].StructureGraph, } ## MAGIC METHODS ## def __init__(self, config): """Initialize the graph store and a layout. NOTE: `rdflib.Dataset` requires a RDF 1.1 compliant store with support for Graph Store HTTP protocol (https://www.w3.org/TR/sparql11-http-rdf-update/). Blazegraph supports this only in the (currently unreleased) 2.2 branch. It works with Jena, which is currently the reference implementation. """ self.config = config self.store = plugin.get('Lmdb', Store)(config['location']) self.ds = Dataset(self.store, default_union=True) self.ds.namespace_manager = nsm @property def attr_routes(self): """ This is a map that allows specific triples to go to certain graphs. It is a machine-friendly version of the static attribute `attr_map` which is formatted for human readability and to avoid repetition. The attributes not mapped here (usually user-provided triples with no special meaning to the application) go to the `fcmain:` graph. The output of this is a dict with a similar structure:: { 'p': { : , : , : , : , [...] }, 't': { : , : , [...] } } """ if not hasattr(self, '_attr_routes'): self._attr_routes = {'p': {}, 't': {}} for dest in self.attr_map.keys(): for term_k, terms in self.attr_map[dest].items(): self._attr_routes[term_k].update( {term: dest for term in terms}) return self._attr_routes def bootstrap(self): """ Delete all graphs and insert the basic triples. """ logger.info('Deleting all data from the graph store.') store = self.ds.store if getattr(store, 'is_txn_open', False): store.rollback() store.destroy(store.path) logger.info('Initializing the graph store with system data.') store.open() fname = path.join( basedir, 'data', 'bootstrap', 'rsrc_centric_layout.sparql') with TxnManager(store, True): with open(fname, 'r') as f: data = Template(f.read()) self.ds.update(data.substitute(timestamp=arrow.utcnow())) def get_raw(self, uri, ctx=None): """ Get a raw graph of a non-LDP resource. The graph is queried across all contexts or within a specific one. :param rdflib.term.URIRef s: URI of the subject. :param rdflib.term.URIRef ctx: URI of the optional context. If None, all named graphs are queried. :rtype: rdflib.Graph """ return self.store.triples((nsc['fcres'][uid], None, None), ctx) def count_rsrc(self): """ Return a count of first-class resources, subdivided in "live" and historic snapshots. """ with TxnManager(self.ds.store) as txn: main = set( self.ds.graph(META_GR_URI)[ : nsc['foaf'].primaryTopic : ]) hist = set( self.ds.graph(HIST_GR_URI)[ : nsc['foaf'].primaryTopic : ]) return {'main': len(main), 'hist': len(hist)} def raw_query(self, qry_str): """ Perform a straight query to the graph store. """ return self.ds.query(qry_str) def get_imr( self, uid, ver_uid=None, strict=True, incl_inbound=False, incl_children=True, embed_children=False, **kwargs): """ See base_rdf_layout.get_imr. """ if ver_uid: uid = self.snapshot_uid(uid, ver_uid) graphs = {pfx[uid] for pfx in self.graph_ns_types.keys()} # Exclude children: remove containment graphs. if not incl_children: graphs.remove(nsc['fcstruct'][uid]) rsrc_graphs = [ self.ds.graph(gr) for gr in graphs] resultset = set(chain.from_iterable(rsrc_graphs)) imr = Graph(identifier=nsc['fcres'][uid]) imr += resultset # Include inbound relationships. if incl_inbound and len(imr): imr += self.get_inbound_rel(nsc['fcres'][uid]) #logger.debug('Found resource: {}'.format( # imr.serialize(format='turtle').decode('utf-8'))) if strict: self._check_rsrc_status(imr) return imr def ask_rsrc_exists(self, uid): """ See base_rdf_layout.ask_rsrc_exists. """ logger.debug('Checking if resource exists: {}'.format(uid)) meta_gr = self.ds.graph(nsc['fcadmin'][uid]) return bool( meta_gr[nsc['fcres'][uid] : RDF.type : nsc['fcrepo'].Resource]) def get_metadata(self, uid, ver_uid=None, strict=True): """ This is an optimized query to get only the administrative metadata. """ logger.debug('Getting metadata for: {}'.format(uid)) if ver_uid: uid = self.snapshot_uid(uid, ver_uid) uri = nsc['fcres'][uid] gr = Graph(identifier=uri) gr += self.ds.graph(nsc['fcadmin'][uid]) if strict: self._check_rsrc_status(gr) return gr def get_user_data(self, uid): """ Get all the user-provided data. :param string uid: Resource UID. """ # *TODO* This only works as long as there is only one user-provided # graph. If multiple user-provided graphs will be supported, this # should use another query to get all of them. userdata_gr = Graph(identifier=nsc['fcres'][uid]) userdata_gr += self.ds.graph(nsc['fcmain'][uid]) return userdata_gr def get_version_info(self, uid, strict=True): """ Get all metadata about a resource's versions. """ # **Note:** This pretty much bends the ontology—it replaces the graph URI # with the subject URI. But the concepts of data and metadata in Fedora # are quite fluid anyways... # WIP—Is it worth to replace SPARQL here? #versions = self.ds.graph(nsc['fcadmin'][uid]).triples( # (nsc['fcres'][uid], nsc['fcrepo'].hasVersion, None)) #for version in versions: # version_meta = self.ds.graph(HIST_GRAPH_URI).triples( qry = """ CONSTRUCT { ?s fcrepo:hasVersion ?v . ?v ?p ?o . } { GRAPH ?ag { ?s fcrepo:hasVersion ?v . } GRAPH ?hg { ?vm foaf:primaryTopic ?v . ?vm ?p ?o . FILTER (?o != ?v) } }""" gr = self._parse_construct(qry, init_bindings={ 'ag': nsc['fcadmin'][uid], 'hg': HIST_GR_URI, 's': nsc['fcres'][uid]}) ver_info_gr = Graph(identifier=nsc['fcres'][uid]) ver_info_gr += gr if strict: self._check_rsrc_status(ver_info_gr) return ver_info_gr def get_inbound_rel(self, subj_uri, full_triple=True): """ Query inbound relationships for a subject. This can be a list of either complete triples, or of subjects referring to the given URI. It excludes historic version snapshots. :param rdflib.URIRef subj_uri: Subject URI. :param boolean full_triple: Whether to return the full triples found or only the subjects. By default, full triples are returned. :rtype: Iterator(tuple(rdflib.term.Identifier) or rdflib.URIRef) :return: Inbound triples or subjects. """ # Only return non-historic graphs. meta_gr = self.ds.graph(META_GR_URI) ptopic_uri = nsc['foaf'].primaryTopic yield from ( (match[:3] if full_triple else match[0]) for match in self.ds.quads((None, None, subj_uri, None)) if set(meta_gr[ : ptopic_uri : match[0]]) ) def get_descendants(self, uid, recurse=True): """ Get descendants (recursive children) of a resource. :param str uid: Resource UID. :rtype: Iterator(rdflib.URIRef) :return: Subjects of descendant resources. """ ds = self.ds subj_uri = nsc['fcres'][uid] ctx_uri = nsc['fcstruct'][uid] def _recurse(dset, s, p, c): new_dset = set(ds.graph(c)[s : p]) for ss in new_dset: dset.add(ss) cc = URIRef(ss.replace(nsc['fcres'], nsc['fcstruct'])) if set(ds.graph(cc)[ss : p]): _recurse(dset, ss, p, cc) return dset return ( _recurse(set(), subj_uri, nsc['ldp'].contains, ctx_uri) if recurse else ds.graph(ctx_uri)[subj_uri : nsc['ldp'].contains : ]) def patch_rsrc(self, uid, qry): """ Patch a resource with SPARQL-Update statements. The statement(s) is/are executed on the user-provided graph only to ensure that the scope is limited to the resource. :param str uid: UID of the resource to be patched. :param dict qry: Parsed and translated query, or query string. """ # Add meta graph for user-defined triples. This may not be used but # it's simple and harmless to add here. self.ds.graph(META_GR_URI).add( (nsc['fcmain'][uid], nsc['foaf'].primaryTopic, nsc['fcres'][uid])) gr = self.ds.graph(nsc['fcmain'][uid]) logger.debug('Updating graph {} with statements: {}'.format( nsc['fcmain'][uid], qry)) return gr.update(qry) def forget_rsrc(self, uid, inbound=True, children=True): """ Completely delete a resource and (optionally) its children and inbound references. NOTE: inbound references in historic versions are not affected. """ # Localize variables to be used in loops. uri = nsc['fcres'][uid] topic_uri = nsc['foaf'].primaryTopic uid_fn = self.uri_to_uid # remove children. if children: logger.debug('Purging children for /{}'.format(uid)) for rsrc_uri in self.get_descendants(uid, False): self.forget_rsrc(uid_fn(rsrc_uri), inbound, False) # Remove structure graph. self.ds.remove_graph(nsc['fcstruct'][uid]) # Remove inbound references. if inbound: for ibs in self.get_inbound_rel(uri): self.ds.remove(ibs) # Remove versions. for ver_uri in self.ds.graph(nsc['fcadmin'][uid])[ uri : nsc['fcrepo'].hasVersion : None]: self.delete_rsrc(uid_fn(ver_uri), True) # Remove resource itself. self.delete_rsrc(uid) def truncate_rsrc(self, uid): """ Remove all user-provided data from a resource and only leave admin and structure data. """ userdata = set(self.get_user_data(uid)) return self.modify_rsrc(uid, remove_trp=userdata) def modify_rsrc(self, uid, remove_trp=set(), add_trp=set()): """ Modify triples about a subject. This method adds and removes triple sets from specific graphs, indicated by the term router. It also adds metadata about the changed graphs. """ remove_routes = defaultdict(set) add_routes = defaultdict(set) historic = VERS_CONT_LABEL in uid graph_types = set() # Graphs that need RDF type metadata added. # Create add and remove sets for each graph. for t in remove_trp: map_graph = self._map_graph_uri(t, uid) target_gr_uri = map_graph[0] remove_routes[target_gr_uri].add(t) graph_types.add(map_graph) for t in add_trp: map_graph = self._map_graph_uri(t, uid) target_gr_uri = map_graph[0] add_routes[target_gr_uri].add(t) graph_types.add(map_graph) # Decide if metadata go into historic or current graph. meta_gr_uri = HIST_GR_URI if historic else META_GR_URI meta_gr = self.ds.graph(meta_gr_uri) # Remove and add triple sets from each graph. for gr_uri, trp in remove_routes.items(): gr = self.ds.graph(gr_uri) gr -= trp for gr_uri, trp in add_routes.items(): gr = self.ds.graph(gr_uri) gr += trp # Add metadata. meta_gr.set( (gr_uri, nsc['foaf'].primaryTopic, nsc['fcres'][uid])) ts = getattr(env, 'timestamp_term', Literal(arrow.utcnow())) meta_gr.set((gr_uri, nsc['fcrepo'].created, ts)) if historic: # @FIXME Ugly reverse engineering. ver_uid = uid.split(VERS_CONT_LABEL)[1].lstrip('/') meta_gr.set(( gr_uri, nsc['fcrepo'].hasVersionLabel, Literal(ver_uid))) # *TODO* More provenance metadata can be added here. # Add graph RDF types. for gr_uri, gr_type in graph_types: meta_gr.add((gr_uri, RDF.type, gr_type)) def delete_rsrc(self, uid, historic=False): """ Delete all aspect graphs of an individual resource. :param uid: Resource UID. :param bool historic: Whether the UID is of a historic version. """ meta_gr_uri = HIST_GR_URI if historic else META_GR_URI for gr_uri in self.ds.graph(meta_gr_uri)[ : nsc['foaf'].primaryTopic : nsc['fcres'][uid]]: self.ds.remove_context(gr_uri) self.ds.graph(meta_gr_uri).remove((gr_uri, None, None)) def snapshot_uid(self, uid, ver_uid): """ Create a versioned UID string from a main UID and a version UID. """ if VERS_CONT_LABEL in uid: raise InvalidResourceError(uid, 'Resource \'{}\' is already a version.') return '{}/{}/{}'.format(uid, VERS_CONT_LABEL, ver_uid) def uri_to_uid(self, uri): """ Convert an internal URI to a UID. """ return str(uri).replace(nsc['fcres'], '') def find_refint_violations(self): """ Find all referential integrity violations. This method looks for dangling relationships within a repository by checking the objects of each triple; if the object is an in-repo resource reference, and no resource with that URI results to be in the repo, that triple is reported. :rtype: set :return: Triples referencing a repository URI that is not a resource. """ for i, obj in enumerate(self.store.all_terms('o'), start=1): if ( isinstance(obj, URIRef) and obj.startswith(nsc['fcres']) and not obj.endswith('fcr:fixity') and not obj.endswith('fcr:versions') and not self.ask_rsrc_exists(self.uri_to_uid( urldefrag(obj).url))): logger.warn('Object not found: {}'.format(obj)) for trp in self.store.triples((None, None, obj)): yield trp if i % 100 == 0: logger.info('{} terms processed.'.format(i)) ## PROTECTED MEMBERS ## def _check_rsrc_status(self, gr): """ Check if a resource is not existing or if it is a tombstone. """ uid = self.uri_to_uid(gr.identifier) if not len(gr): raise ResourceNotExistsError(uid) # Check if resource is a tombstone. if gr[gr.identifier : RDF.type : nsc['fcsystem'].Tombstone]: raise TombstoneError( uid, gr.value(gr.identifier, nsc['fcrepo'].created)) elif gr.value(gr.identifier, nsc['fcsystem'].tombstone): raise TombstoneError( self.uri_to_uid( gr.value(gr.identifier, nsc['fcsystem'].tombstone)), gr.value(gr.identifier, nsc['fcrepo'].created)) def _parse_construct(self, qry, init_bindings={}): """ Parse a CONSTRUCT query. :rtype: rdflib.Graph """ try: qres = self.ds.query(qry, initBindings=init_bindings) except ResultException: # RDFlib bug: https://github.com/RDFLib/rdflib/issues/775 return Graph() else: return qres.graph def _map_graph_uri(self, t, uid): """ Map a triple to a namespace prefix corresponding to a graph. :rtype: tuple :return: 2-tuple with a graph URI and an associated RDF type. """ if t[1] in self.attr_routes['p'].keys(): pfx = self.attr_routes['p'][t[1]] elif t[1] == RDF.type and t[2] in self.attr_routes['t'].keys(): pfx = self.attr_routes['t'][t[2]] else: pfx = nsc['fcmain'] return (pfx[uid], self.graph_ns_types[pfx])