SPARQL: COUNT DISTINCT not working properly

[pchampin]

The combination of COUNT with DISTINCT is not working propetry in the RDFLib implementation of SPARQL.
Consider the following code:

from rdflib import *

g = Graph()
g.parse(format="turtle", publicID="http://example.org/", data="""
@prefix : <> .

<#a>
  :knows <#b>, <#c> ;
  :age 42 .

<#b>
  :knows <#a>, <#c> ;
  :age 36 .

<#c>
  :knows <#b>, <#c> ;
  :age 20 .

""")


print "Query 1: people knowing someone younger"
results = g.query("""
PREFIX : <http://example.org/>

SELECT DISTINCT ?x {
  ?x :age ?ax ; :knows [ :age ?ay ].
  FILTER( ?ax > ?ay )
}
""")
for i in results:
    print str(i[0])


print "\nQuery 2: count people knowing someone younger"
results = g.query("""
PREFIX : <http://example.org/>

SELECT (COUNT(DISTINCT ?x) as ?cx) {
  ?x :age ?ax ; :knows [ :age ?ay ].
  FILTER( ?ax > ?ay )
}
""")


for i in results:
    print str(i[0])

It produces the following output:

Query 1: people knowing someone younger
http://example.org/#a
http://example.org/#b

Query 2: count people knowing someone younger
3

while the result of Query 2 should obvioulsy be 2. It seems that the DISTINCT in Query 2 has no effect (indeed, <#a> matches the query twice, as she knows 2 people younger than herself).

[pchampin]

Actually, the same problem exists with SUM(DISTINCT ?x), AVG(DISTINCT ?x) or GROUP_CONCAT(DISTINCT ?x).

[gromgull]

This works in master now, presumably fixed by the refactoring of the aggregation code here: #678