Path query chaining issue

[bgyori]

Consider the following RDF file example.rdf

<http://example.org/A> <http://example.org/relations/isa> <http://example.org/X> .
<http://example.org/A> <http://example.org/relations/isa> <http://example.org/Y> .
<http://example.org/B> <http://example.org/relations/isa> <http://example.org/X> .

With rdflib 4.2.2 I do

import rdflib
g = rdflib.Graph()
g.parse('example.rdf', format='nt')
g.query('SELECT ?child ?parent WHERE {?child <http://example.org/relations/isa> ?parent .}')

and print the results, and I get the correct relationships:

http://example.org/A http://example.org/Y
http://example.org/B http://example.org/X
http://example.org/A http://example.org/X

However, when changing the query to

g.query('SELECT ?child ?parent WHERE {?child <http://example.org/relations/isa>+ ?parent .}')

I get

http://example.org/A http://example.org/Y
http://example.org/A http://example.org/X
http://example.org/B http://example.org/X
http://example.org/B http://example.org/Y

where the relationship B isa+ Y was incorrectly inferred. Is this a bug or am I misunderstanding the meaning of + in path chaining?

[joernhees]

reproducible in 5.0.0-dev:

In [1]: from rdflib import URIRef, Graph
INFO:rdflib:RDFLib Version: 5.0.0-dev

In [2]: g = Graph()

In [3]: a, b, x, y = [URIRef(s) for s in "abxy"]

In [4]: isa = URIRef('isa')

In [5]: g.add((a, isa, x))

In [6]: g.add((a, isa, y))

In [7]: g.add((b, isa, x))

In [8]: list(g)
Out[8]:
[(rdflib.term.URIRef(u'a'),
  rdflib.term.URIRef(u'isa'),
  rdflib.term.URIRef(u'y')),
 (rdflib.term.URIRef(u'b'),
  rdflib.term.URIRef(u'isa'),
  rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'a'),
  rdflib.term.URIRef(u'isa'),
  rdflib.term.URIRef(u'x'))]

In [9]: list(g.query('SELECT ?child ?parent WHERE {?child <isa> ?parent .}'))
Out[9]:
[(rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'y')),
 (rdflib.term.URIRef(u'b'), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'x'))]

In [10]: list(g.query('SELECT ?child ?parent WHERE {?child <isa>+ ?parent .}'))
Out[10]:
[(rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'y')),
 (rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'b'), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'b'), rdflib.term.URIRef(u'y'))]

In [11]: list(g.query('SELECT ?child ?parent WHERE {?child <isa>* ?parent .}'))
Out[11]:
[(rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'a')),
 (rdflib.term.URIRef(u'y'), rdflib.term.URIRef(u'y')),
 (rdflib.term.URIRef(u'b'), rdflib.term.URIRef(u'b')),
 (rdflib.term.URIRef(u'x'), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'y')),
 (rdflib.term.URIRef(u'a'), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'b'), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'b'), rdflib.term.URIRef(u'y'))]

[gromgull]

This is unrelated to SPARQL, if I follow on from @joernhees's ipython sesssion:

In [67]: list(g.triples( ( None, isa*'+', None ) ) )
Out[67]:
[(rdflib.term.URIRef(u'a'), Path(isa+), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'a'), Path(isa+), rdflib.term.URIRef(u'y')),
 (rdflib.term.URIRef(u'b'), Path(isa+), rdflib.term.URIRef(u'x')),
 (rdflib.term.URIRef(u'b'), Path(isa+), rdflib.term.URIRef(u'y'))]

[gromgull]

This means the bug is in here somewhere: https://github.com/RDFLib/rdflib/blob/master/rdflib/paths.py#L380-L388

[joernhees]

fixed by #822