summaryrefslogtreecommitdiff
path: root/rdflib/plugins/stores/sparqlstore.py
blob: 4c93e213507e933b818ace3a2f20d5fa5ea3be1f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
# -*- coding: utf-8 -*-
#
"""
This is an RDFLib store around Ivan Herman et al.'s SPARQL service wrapper.
This was first done in layer-cake, and then ported to RDFLib

"""

# Defines some SPARQL keywords
LIMIT = 'LIMIT'
OFFSET = 'OFFSET'
ORDERBY = 'ORDER BY'

import re
import collections

from .sparqlconnector import SPARQLConnector

from rdflib.plugins.stores.regexmatching import NATIVE_REGEX

from rdflib.store import Store
from rdflib import Variable, BNode
from rdflib.graph import DATASET_DEFAULT_GRAPH_ID
from rdflib.term import Node

from six import string_types

BNODE_IDENT_PATTERN = re.compile('(?P<label>_\:[^\s]+)')


def _node_to_sparql(node):
    if isinstance(node, BNode):
        raise Exception(
            "SPARQLStore does not support BNodes! "
            "See http://www.w3.org/TR/sparql11-query/#BGPsparqlBNodes"
        )
    return node.n3()


class SPARQLStore(SPARQLConnector, Store):
    """An RDFLib store around a SPARQL endpoint

    This is context-aware and should work as expected
    when a context is specified.

    For ConjunctiveGraphs, reading is done from the "default graph". Exactly
    what this means depends on your endpoint, because SPARQL does not offer a
    simple way to query the union of all graphs as it would be expected for a
    ConjuntiveGraph. This is why we recommend using Dataset instead, which is
    motivated by the SPARQL 1.1.

    Fuseki/TDB has a flag for specifying that the default graph
    is the union of all graphs (``tdb:unionDefaultGraph`` in the Fuseki config).

    .. warning:: By default the SPARQL Store does not support blank-nodes!

                 As blank-nodes act as variables in SPARQL queries,
                 there is no way to query for a particular blank node without
                 using non-standard SPARQL extensions.

                 See http://www.w3.org/TR/sparql11-query/#BGPsparqlBNodes

    You can make use of such extensions through the ``node_to_sparql``
    argument. For example if you want to transform BNode('0001') into
    "<bnode:b0001>", you can use a function like this:

    >>> def my_bnode_ext(node):
    ...    if isinstance(node, BNode):
    ...        return '<bnode:b%s>' % node
    ...    return _node_to_sparql(node)
    >>> store = SPARQLStore('http://dbpedia.org/sparql',
    ...                     node_to_sparql=my_bnode_ext)

    You can request a particular result serialization with the
    ``returnFormat`` parameter. This is a string that must have a
    matching plugin registered. Built in is support for ``xml``,
    ``json``, ``csv``, ``tsv`` and ``application/rdf+xml``.

    The underlying SPARQLConnector builds in the requests library.
    Any extra kwargs passed to the SPARQLStore connector are passed to
    requests when doing HTTP calls. I.e. you have full control of
    cookies/auth/headers.

    Form example:

    >>> store = SPARQLStore('...my endpoint ...', auth=('user','pass'))

    will use HTTP basic auth.

    """
    formula_aware = False
    transaction_aware = False
    graph_aware = True
    regex_matching = NATIVE_REGEX

    def __init__(self,
                 endpoint=None,
                 sparql11=True, context_aware=True,
                 node_to_sparql=_node_to_sparql,
                 returnFormat='xml',
                 **sparqlconnector_kwargs):
        """
        """
        super(SPARQLStore, self).__init__(
            endpoint, returnFormat=returnFormat, **sparqlconnector_kwargs)

        self.node_to_sparql = node_to_sparql
        self.nsBindings = {}
        self.sparql11 = sparql11
        self.context_aware = context_aware
        self.graph_aware = context_aware
        self._queries = 0

    # Database Management Methods
    def create(self, configuration):
        raise TypeError('The SPARQL store is read only')

    def open(self, configuration, create=False):
        """
        sets the endpoint URL for this SPARQLStore
        if create==True an exception is thrown.
        """
        if create:
            raise Exception("Cannot create a SPARQL Endpoint")

        self.query_endpoint = configuration

    def destroy(self, configuration):
        raise TypeError('The SPARQL store is read only')

    # Transactional interfaces
    def commit(self):
        raise TypeError('The SPARQL store is read only')

    def rollback(self):
        raise TypeError('The SPARQL store is read only')

    def add(self, _, context=None, quoted=False):
        raise TypeError('The SPARQL store is read only')

    def addN(self, quads):
        raise TypeError('The SPARQL store is read only')

    def remove(self, _, context):
        raise TypeError('The SPARQL store is read only')

    def _query(self, *args, **kwargs):
        self._queries += 1

        return super(SPARQLStore, self).query(*args, **kwargs)

    def _inject_prefixes(self, query, extra_bindings):
        bindings = list(self.nsBindings.items()) + list(extra_bindings.items())
        if not bindings:
            return query
        return '\n'.join([
            '\n'.join(['PREFIX %s: <%s>' % (k, v) for k, v in bindings]),
            '',  # separate ns_bindings from query with an empty line
            query
        ])

    def _preprocess_query(self, query):
        return self._inject_prefixes(query)

    def query(self, query,
              initNs={},
              initBindings={},
              queryGraph=None,
              DEBUG=False):
        self.debug = DEBUG
        assert isinstance(query, string_types)

        query = self._inject_prefixes(query, initNs)

        if initBindings:
            if not self.sparql11:
                raise Exception(
                    "initBindings not supported for SPARQL 1.0 Endpoints.")
            v = list(initBindings)

            # VALUES was added to SPARQL 1.1 on 2012/07/24
            query += "\nVALUES ( %s )\n{ ( %s ) }\n"\
                % (" ".join("?" + str(x) for x in v),
                   " ".join(self.node_to_sparql(initBindings[x]) for x in v))

        return self._query(query,
                           default_graph=queryGraph if self._is_contextual(queryGraph) else None)

    def triples(self, spo, context=None):
        """
        - tuple **(s, o, p)**
            the triple used as filter for the SPARQL select.
            (None, None, None) means anything.
        - context **context**
            the graph effectively calling this method.

        Returns a tuple of triples executing essentially a SPARQL like
        SELECT ?subj ?pred ?obj WHERE { ?subj ?pred ?obj }

        **context** may include three parameter
        to refine the underlying query:
         * LIMIT: an integer to limit the number of results
         * OFFSET: an integer to enable paging of results
         * ORDERBY: an instance of Variable('s'), Variable('o') or Variable('p')
        or, by default, the first 'None' from the given triple

        .. warning::
        - Using LIMIT or OFFSET automatically include ORDERBY otherwise this is
        because the results are retrieved in a not deterministic way (depends on
        the walking path on the graph)
        - Using OFFSET without defining LIMIT will discard the first OFFSET - 1
        results

        ``
        a_graph.LIMIT = limit
        a_graph.OFFSET = offset
        triple_generator = a_graph.triples(mytriple):
            #do something
        #Removes LIMIT and OFFSET if not required for the next triple() calls
        del a_graph.LIMIT
        del a_graph.OFFSET
        ``
        """

        s, p, o = spo

        vars = []
        if not s:
            s = Variable('s')
            vars.append(s)

        if not p:
            p = Variable('p')
            vars.append(p)
        if not o:
            o = Variable('o')
            vars.append(o)

        if vars:
            v = ' '.join([term.n3() for term in vars])
            verb = 'SELECT %s ' % v
        else:
            verb = 'ASK'

        nts = self.node_to_sparql
        query = "%s { %s %s %s }" % (verb, nts(s), nts(p), nts(o))

        # The ORDER BY is necessary
        if hasattr(context, LIMIT) or hasattr(context, OFFSET) \
                or hasattr(context, ORDERBY):
            var = None
            if isinstance(s, Variable):
                var = s
            elif isinstance(p, Variable):
                var = p
            elif isinstance(o, Variable):
                var = o
            elif hasattr(context, ORDERBY) \
                    and isinstance(getattr(context, ORDERBY), Variable):
                var = getattr(context, ORDERBY)
            query = query + ' %s %s' % (ORDERBY, var.n3())

        try:
            query = query + ' LIMIT %s' % int(getattr(context, LIMIT))
        except (ValueError, TypeError, AttributeError):
            pass
        try:
            query = query + ' OFFSET %s' % int(getattr(context, OFFSET))
        except (ValueError, TypeError, AttributeError):
            pass

        result = self._query(query,
                             default_graph=context.identifier if self._is_contextual(context) else None)

        if vars:
            for row in result:
                yield (row.get(s, s),
                       row.get(p, p),
                       row.get(o, o)), None  # why is the context here not the passed in graph 'context'?
        else:
            if result.askAnswer:
                yield (s, p, o), None

    def triples_choices(self, _, context=None):
        """
        A variant of triples that can take a list of terms instead of a
        single term in any slot.  Stores can implement this to optimize
        the response time from the import default 'fallback' implementation,
        which will iterate over each term in the list and dispatch to
        triples.
        """
        raise NotImplementedError('Triples choices currently not supported')

    def __len__(self, context=None):
        if not self.sparql11:
            raise NotImplementedError(
                "For performance reasons, this is not" +
                "supported for sparql1.0 endpoints")
        else:
            q = "SELECT (count(*) as ?c) WHERE {?s ?p ?o .}"

            result = self._query(q,
                                 default_graph=context.identifier if self._is_contextual(context) else None)

            return int(next(iter(result)).c)

    def contexts(self, triple=None):
        """
        Iterates over results to "SELECT ?NAME { GRAPH ?NAME { ?s ?p ?o } }"
        or "SELECT ?NAME { GRAPH ?NAME {} }" if triple is `None`.

        Returns instances of this store with the SPARQL wrapper
        object updated via addNamedGraph(?NAME).

        This causes a named-graph-uri key / value  pair to be sent over
        the protocol.

        Please note that some SPARQL endpoints are not able to find empty named
        graphs.
        """

        if triple:
            nts = self.node_to_sparql
            s, p, o = triple
            params = (nts(s if s else Variable('s')),
                      nts(p if p else Variable('p')),
                      nts(o if o else Variable('o')))
            q = 'SELECT ?name WHERE { GRAPH ?name { %s %s %s }}' % params
        else:
            q = 'SELECT ?name WHERE { GRAPH ?name {} }'

        result = self._query(q)

        return (row.name for row in result)

    # Namespace persistence interface implementation
    def bind(self, prefix, namespace):
        self.nsBindings[prefix] = namespace

    def prefix(self, namespace):
        """ """
        return dict(
            [(v, k) for k, v in self.nsBindings.items()]
        ).get(namespace)

    def namespace(self, prefix):
        return self.nsBindings.get(prefix)

    def namespaces(self):
        for prefix, ns in self.nsBindings.items():
            yield prefix, ns

    def add_graph(self, graph):
        raise TypeError('The SPARQL store is read only')

    def remove_graph(self, graph):
        raise TypeError('The SPARQL store is read only')

    def _is_contextual(self, graph):
        """ Returns `True` if the "GRAPH" keyword must appear
        in the final SPARQL query sent to the endpoint.
        """
        if (not self.context_aware) or (graph is None):
            return False
        if isinstance(graph, string_types):
            return graph != '__UNION__'
        else:
            return graph.identifier != DATASET_DEFAULT_GRAPH_ID

    def close(self, commit_pending_transaction=None):
        SPARQLConnector.close(self)


class SPARQLUpdateStore(SPARQLStore):
    """A store using SPARQL queries for reading and SPARQL Update for changes.

    This can be context-aware, if so, any changes will be to the given named
    graph only.

    In favor of the SPARQL 1.1 motivated Dataset, we advise against using this
    with ConjunctiveGraphs, as it reads and writes from and to the
    "default graph". Exactly what this means depends on the endpoint and can
    result in confusion.

    For Graph objects, everything works as expected.

    See the :class:`SPARQLStore` base class for more information.

    """

    where_pattern = re.compile(r"""(?P<where>WHERE\s*\{)""", re.IGNORECASE)

    ##################################################################
    ### Regex for injecting GRAPH blocks into updates on a context ###
    ##################################################################

    # Observations on the SPARQL grammar (http://www.w3.org/TR/2013/REC-sparql11-query-20130321/):
    # 1. Only the terminals STRING_LITERAL1, STRING_LITERAL2,
    #    STRING_LITERAL_LONG1, STRING_LITERAL_LONG2, and comments can contain
    #    curly braces.
    # 2. The non-terminals introduce curly braces in pairs only.
    # 3. Unescaped " can occur only in strings and comments.
    # 3. Unescaped ' can occur only in strings, comments, and IRIRefs.
    # 4. \ always escapes the following character, especially \", \', and
    #    \\ denote literal ", ', and \ respectively.
    # 5. # always starts a comment outside of string and IRI
    # 6. A comment ends at the next newline
    # 7. IRIREFs need to be detected, as they may contain # without starting a comment
    # 8. PrefixedNames do not contain a #
    # As a consequence, it should be rather easy to detect strings and comments
    # in order to avoid unbalanced curly braces.

    # From the SPARQL grammar
    STRING_LITERAL1 = u"'([^'\\\\]|\\\\.)*'"
    STRING_LITERAL2 = u'"([^"\\\\]|\\\\.)*"'
    STRING_LITERAL_LONG1 = u"'''(('|'')?([^'\\\\]|\\\\.))*'''"
    STRING_LITERAL_LONG2 = u'"""(("|"")?([^"\\\\]|\\\\.))*"""'
    String = u'(%s)|(%s)|(%s)|(%s)' % (STRING_LITERAL1, STRING_LITERAL2,
                                       STRING_LITERAL_LONG1, STRING_LITERAL_LONG2)
    IRIREF = u'<([^<>"{}|^`\\]\\\\\[\\x00-\\x20])*>'
    COMMENT = u'#[^\\x0D\\x0A]*([\\x0D\\x0A]|\\Z)'

    # Simplified grammar to find { at beginning and } at end of blocks
    BLOCK_START = u'{'
    BLOCK_END = u'}'
    ESCAPED = u'\\\\.'

    # Match anything that doesn't start or end a block:
    BlockContent = u'(%s)|(%s)|(%s)|(%s)' % (String, IRIREF, COMMENT, ESCAPED)
    BlockFinding = u'(?P<block_start>%s)|(?P<block_end>%s)|(?P<block_content>%s)' % (BLOCK_START, BLOCK_END, BlockContent)
    BLOCK_FINDING_PATTERN = re.compile(BlockFinding)

    # Note that BLOCK_FINDING_PATTERN.finditer() will not cover the whole
    # string with matches. Everything that is not matched will have to be
    # part of the modified query as is.

    ##################################################################

    def __init__(self,
                 queryEndpoint=None, update_endpoint=None,
                 sparql11=True,
                 context_aware=True,
                 postAsEncoded=True,
                 autocommit=True,
                 dirty_reads=False,
                 **kwds
                 ):
        """
        :param autocommit if set, the store will commit after every
        writing operations. If False, we only make queries on the
        server once commit is called.

        :param dirty_reads if set, we do not commit before reading. So you
        cannot read what you wrote before manually calling commit.

        """

        SPARQLStore.__init__(
            self,
            queryEndpoint,
            sparql11,
            context_aware,
            update_endpoint=update_endpoint,
            **kwds
        )

        self.postAsEncoded = postAsEncoded
        self.autocommit = autocommit
        self.dirty_reads = dirty_reads
        self._edits = None
        self._updates = 0

    def query(self, *args, **kwargs):
        if not self.autocommit and not self.dirty_reads:
            self.commit()
        return SPARQLStore.query(self, *args, **kwargs)

    def triples(self, *args, **kwargs):
        if not self.autocommit and not self.dirty_reads:
            self.commit()
        return SPARQLStore.triples(self, *args, **kwargs)

    def contexts(self, *args, **kwargs):
        if not self.autocommit and not self.dirty_reads:
            self.commit()
        return SPARQLStore.contexts(self, *args, **kwargs)

    def __len__(self, *args, **kwargs):
        if not self.autocommit and not self.dirty_reads:
            self.commit()
        return SPARQLStore.__len__(self, *args, **kwargs)

    def open(self, configuration, create=False):
        """
        sets the endpoint URLs for this SPARQLStore
        :param configuration: either a tuple of (query_endpoint, update_endpoint),
            or a string with the endpoint which is configured as query and update endpoint
        :param create: if True an exception is thrown.
        """

        if create:
            raise Exception("Cannot create a SPARQL Endpoint")

        if isinstance(configuration, tuple):
            self.query_endpoint = configuration[0]
            if len(configuration) > 1:
                self.update_endpoint = configuration[1]
        else:
            self.query_endpoint = configuration
            self.update_endpoint = configuration

    def _transaction(self):
        if self._edits is None:
            self._edits = []
        return self._edits

    # Transactional interfaces
    def commit(self):
        """ add(), addN(), and remove() are transactional to reduce overhead of many small edits.
            Read and update() calls will automatically commit any outstanding edits.
            This should behave as expected most of the time, except that alternating writes
            and reads can degenerate to the original call-per-triple situation that originally existed.
        """
        if self._edits and len(self._edits) > 0:
            self._update('\n;\n'.join(self._edits))
            self._edits = None

    def rollback(self):
        self._edits = None

    def add(self, spo, context=None, quoted=False):
        """ Add a triple to the store of triples. """

        if not self.update_endpoint:
            raise Exception("UpdateEndpoint is not set - call 'open'")

        assert not quoted
        (subject, predicate, obj) = spo

        nts = self.node_to_sparql
        triple = "%s %s %s ." % (nts(subject), nts(predicate), nts(obj))
        if self._is_contextual(context):
            q = "INSERT DATA { GRAPH %s { %s } }" % (
                nts(context.identifier), triple)
        else:
            q = "INSERT DATA { %s }" % triple
        self._transaction().append(q)
        if self.autocommit:
            self.commit()

    def addN(self, quads):
        """ Add a list of quads to the store. """
        if not self.update_endpoint:
            raise Exception("UpdateEndpoint is not set - call 'open'")

        contexts = collections.defaultdict(list)
        for subject, predicate, obj, context in quads:
            contexts[context].append((subject, predicate, obj))
        data = []
        nts = self.node_to_sparql
        for context in contexts:
            triples = [
                "%s %s %s ." % (
                    nts(subject), nts(predicate), nts(obj)
                ) for subject, predicate, obj in contexts[context]
            ]
            data.append("INSERT DATA { GRAPH %s { %s } }\n" % (
                nts(context.identifier), '\n'.join(triples)))
        self._transaction().extend(data)
        if self.autocommit:
            self.commit()

    def remove(self, spo, context):
        """ Remove a triple from the store """
        if not self.update_endpoint:
            raise Exception("UpdateEndpoint is not set - call 'open'")

        (subject, predicate, obj) = spo
        if not subject:
            subject = Variable("S")
        if not predicate:
            predicate = Variable("P")
        if not obj:
            obj = Variable("O")

        nts = self.node_to_sparql
        triple = "%s %s %s ." % (nts(subject), nts(predicate), nts(obj))
        if self._is_contextual(context):
            cid = nts(context.identifier)
            q = "WITH %(graph)s DELETE { %(triple)s } WHERE { %(triple)s }" % { 'graph': cid, 'triple': triple }
        else:
            q = "DELETE { %s } WHERE { %s } " % (triple, triple)
        self._transaction().append(q)
        if self.autocommit:
            self.commit()

    def setTimeout(self, timeout):
        self._timeout = int(timeout)

    def _update(self, update):

        self._updates += 1

        SPARQLConnector.update(self, update)

    def update(self, query,
               initNs={},
               initBindings={},
               queryGraph=None,
               DEBUG=False):
        """
        Perform a SPARQL Update Query against the endpoint,
        INSERT, LOAD, DELETE etc.
        Setting initNs adds PREFIX declarations to the beginning of
        the update. Setting initBindings adds inline VALUEs to the
        beginning of every WHERE clause. By the SPARQL grammar, all
        operations that support variables (namely INSERT and DELETE)
        require a WHERE clause.
        Important: initBindings fails if the update contains the
        substring 'WHERE {' which does not denote a WHERE clause, e.g.
        if it is part of a literal.

        .. admonition:: Context-aware query rewriting

            - **When:**  If context-awareness is enabled and the graph is not the default graph of the store.
            - **Why:** To ensure consistency with the :class:`~rdflib.plugins.memory.IOMemory` store.
              The graph must except "local" SPARQL requests (requests with no GRAPH keyword)
              like if it was the default graph.
            - **What is done:** These "local" queries are rewritten by this store.
              The content of each block of a SPARQL Update operation is wrapped in a GRAPH block
              except if the block is empty.
              This basically causes INSERT, INSERT DATA, DELETE, DELETE DATA and WHERE to operate
              only on the context.
            - **Example:** `"INSERT DATA { <urn:michel> <urn:likes> <urn:pizza> }"` is converted into
              `"INSERT DATA { GRAPH <urn:graph> { <urn:michel> <urn:likes> <urn:pizza> } }"`.
            - **Warning:** Queries are presumed to be "local" but this assumption is **not checked**.
              For instance, if the query already contains GRAPH blocks, the latter will be wrapped in new GRAPH blocks.
            - **Warning:** A simplified grammar is used that should tolerate
              extensions of the SPARQL grammar. Still, the process may fail in
              uncommon situations and produce invalid output.

        """
        if not self.update_endpoint:
            raise Exception("UpdateEndpoint is not set - call 'open'")

        self.debug = DEBUG
        assert isinstance(query, string_types)
        query = self._inject_prefixes(query, initNs)

        if self._is_contextual(queryGraph):
            query = self._insert_named_graph(query, queryGraph)

        if initBindings:
            # For INSERT and DELETE the WHERE clause is obligatory
            # (http://www.w3.org/TR/2013/REC-sparql11-query-20130321/#rModify)
            # Other query types do not allow variables and don't
            # have a WHERE clause.  This also works for updates with
            # more than one INSERT/DELETE.
            v = list(initBindings)
            values = "\nVALUES ( %s )\n{ ( %s ) }\n"\
                % (" ".join("?" + str(x) for x in v),
                   " ".join(self.node_to_sparql(initBindings[x]) for x in v))

            query = self.where_pattern.sub("WHERE { " + values, query)

        self._transaction().append(query)
        if self.autocommit:
            self.commit()

    def _insert_named_graph(self, query, query_graph):
        """
            Inserts GRAPH <query_graph> {} into blocks of SPARQL Update operations

            For instance,  "INSERT DATA { <urn:michel> <urn:likes> <urn:pizza> }"
            is converted into
            "INSERT DATA { GRAPH <urn:graph> { <urn:michel> <urn:likes> <urn:pizza> } }"
        """
        if isinstance(query_graph, Node):
            query_graph = self.node_to_sparql(query_graph)
        else:
            query_graph = '<%s>' % query_graph
        graph_block_open = " GRAPH %s {" % query_graph
        graph_block_close = "} "

        # SPARQL Update supports the following operations:
        # LOAD, CLEAR, DROP, ADD, MOVE, COPY, CREATE, INSERT DATA, DELETE DATA, DELETE/INSERT, DELETE WHERE
        # LOAD, CLEAR, DROP, ADD, MOVE, COPY, CREATE do not make much sense in a context.
        # INSERT DATA, DELETE DATA, and DELETE WHERE require the contents of their block to be wrapped in a GRAPH <?> { }.
        # DELETE/INSERT supports the WITH keyword, which sets the graph to be
        # used for all following DELETE/INSERT instruction including the
        # non-optional WHERE block. Equivalently, a GRAPH block can be added to
        # all blocks.
        #
        # Strategy employed here: Wrap the contents of every top-level block into a `GRAPH <?> { }`.

        level = 0
        modified_query = []
        pos = 0
        for match in self.BLOCK_FINDING_PATTERN.finditer(query):
            if match.group('block_start') is not None:
                level += 1
                if level == 1:
                    modified_query.append(query[pos:match.end()])
                    modified_query.append(graph_block_open)
                    pos = match.end()
            elif match.group('block_end') is not None:
                if level == 1:
                    since_previous_pos = query[pos:match.start()]
                    if modified_query[-1] is graph_block_open and (since_previous_pos == "" or since_previous_pos.isspace()):
                        # In this case, adding graph_block_start and
                        # graph_block_end results in an empty GRAPH block. Some
                        # enpoints (e.g. TDB) can not handle this. Therefore
                        # remove the previously added block_start.
                        modified_query.pop()
                        modified_query.append(since_previous_pos)
                    else:
                        modified_query.append(since_previous_pos)
                        modified_query.append(graph_block_close)
                    pos = match.start()
                level -= 1
        modified_query.append(query[pos:])

        return "".join(modified_query)

    def add_graph(self, graph):
        if not self.graph_aware:
            Store.add_graph(self, graph)
        elif graph.identifier != DATASET_DEFAULT_GRAPH_ID:
            self.update(
                "CREATE GRAPH %s" % self.node_to_sparql(graph.identifier))

    def close(self, commit_pending_transaction=False):

        if commit_pending_transaction:
            self.commit()

        super(SPARQLStore, self).close()

    def remove_graph(self, graph):
        if not self.graph_aware:
            Store.remove_graph(self, graph)
        elif graph.identifier == DATASET_DEFAULT_GRAPH_ID:
            self.update("DROP DEFAULT")
        else:
            self.update(
                "DROP GRAPH %s" % self.node_to_sparql(graph.identifier))