[jira] [Created] (JENA-1895) Incorrect results on join of two subqueries with HAVING

["Dave Griffith (Jira)" <ji...@apache.org>]

Dave Griffith created JENA-1895:
-----------------------------------

             Summary: Incorrect results on join of two subqueries with HAVING
                 Key: JENA-1895
                 URL: https://issues.apache.org/jira/browse/JENA-1895
             Project: Apache Jena
          Issue Type: Bug
            Reporter: Dave Griffith


This query
PREFIX p0: <https://ddwmetrics.linked.data.world/d/zendesk-data/>

SELECT?v_16?v_17?v_3?v_4?v_15
WHERE {
{
SELECT (?v_5AS?v_3) (?v_6AS?v_4) (COUNT(*) AS?v_16)
WHERE {
SELECT?v_5?v_6
WHERE {
{
SELECT?v_7
WHERE {
?v_9 a p0:tbl-tickets .
?v_9 p0:col-tickets-status ?v_8 .
?v_9 p0:col-tickets-created ?v_7 .
FILTER (?v_8!="deleted")
}
}
BIND (<http://data.world/function/functions#datePart>("year", ?v_7) AS?v_5)
BIND (<http://data.world/function/functions#datePart>("month", ?v_7) AS?v_6)
}
}
GROUP BY ?v_5 ?v_6
}
{
SELECT (?v_10AS?v_3) (?v_11AS?v_4) (COUNT(*) AS?v_17)
WHERE {
SELECT?v_10?v_11
WHERE {
{
SELECT?v_12
WHERE {
?v_14 a p0:tbl-tickets .
?v_14 p0:col-tickets-status ?v_13 .
OPTIONAL {
?v_14 p0:col-tickets-solved ?v_12 .
}
FILTER (?v_13!="deleted")
}
}
BIND (<http://data.world/function/functions#datePart>("year", ?v_12) AS?v_10)
BIND (<http://data.world/function/functions#datePart>("month", ?v_12) AS?v_11)
}
}
GROUPBY ?v_10 ?v_11
}
FILTER (?v_3 > 2017||?v_3=2017&&?v_4 > 7)
BIND (CONCAT(<http://www.w3.org/2001/XMLSchema#string>(?v_3), "-", <http://www.w3.org/2001/XMLSchema#string>(?v_4)) AS?v_15)
}
Results, correctly, in 34 rows when executed against our graph.
 
This similar query, with the final FILTER replaced by equivalent HAVINGs on subqueries
 
PREFIX p0: <https://ddwmetrics.linked.data.world/d/zendesk-data/>

SELECT?v_16?v_17?v_3?v_4?v_15
WHERE {
{
SELECT (?v_5AS?v_3) (?v_6AS?v_4) (COUNT(*) AS?v_16)
WHERE {
SELECT?v_5?v_6
WHERE {
{
SELECT?v_7
WHERE {
?v_9 a p0:tbl-tickets .
?v_9 p0:col-tickets-status ?v_8 .
?v_9 p0:col-tickets-created ?v_7 .
FILTER (?v_8!="deleted")
}
}
BIND (<http://data.world/function/functions#datePart>("year", ?v_7) AS?v_5)
BIND (<http://data.world/function/functions#datePart>("month", ?v_7) AS?v_6)
}
}
GROUPBY?v_5?v_6
HAVING (?v_5 > 2017||?v_5=2017&&?v_6 > 7)
}
{
SELECT (?v_10AS?v_3) (?v_11AS?v_4) (COUNT(*) AS?v_17)
WHERE {
SELECT?v_10?v_11
WHERE {
{
SELECT?v_12
WHERE {
?v_14 a p0:tbl-tickets .
?v_14 p0:col-tickets-status ?v_13 .
OPTIONAL {
?v_14 p0:col-tickets-solved ?v_12 .
}
FILTER (?v_13!="deleted")
}
}
BIND (<http://data.world/function/functions#datePart>("year", ?v_12) AS?v_10)
BIND (<http://data.world/function/functions#datePart>("month", ?v_12) AS?v_11)
}
}
GROUPBY?v_10?v_11
HAVING (?v_10 > 2017||?v_10=2017&&?v_11 > 7)
}
BIND (CONCAT(<http://www.w3.org/2001/XMLSchema#string>(?v_3), "-", <http://www.w3.org/2001/XMLSchema#string>(?v_4)) AS?v_15)
 
results in 1156 rows when executed against our graph, in ways that suggest that a cartesian cross is being done instead of a join between the two subqueries.  
 
Additionally, our logs show many instances of errors like 
 
merge: Mismatch : "2018"^^http://www.w3.org/2001/XMLSchema#integer != "2019"^^http://www.w3.org/2001/XMLSchema#integer level=WARN logger=org.apache.jena.sparql.engine.iterator.QueryIterProjectMerge 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)