[jira] [Commented] (JENA-1254) GROUP_CONCAT returns nothing when concatenating unbound and bound values.

["Andy Seaborne (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/JENA-1254?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15621670#comment-15621670 ] 

Andy Seaborne commented on JENA-1254:
-------------------------------------

If evaluation of _{{expr}}_ in {{GROUP_CONCAT(expr,...)}} causes and eval error, then the value of {{GROUP_CONCAT}} is an error and the AS does not bind a value. This is defined in the SPARQL 1.1 spec, section 18.5.1.7. {{Flatten(M)}} contains error values, and so where it uses the string {{CONCAT}} to build the string, we get an eval error on {{CONCAT}}.

To replace unbound (which is a form of error when asked for its value) with a value, use {{COALESCE}} (there other ways such as {{IF}} and {{BOUND}}).

{{GROUP_CONCAT(COALESCE(expr,"SomeDefault")  ...)}}

which is more easily separated out as 

{noformat}
SELECT GROUP_CONCAT(?X  ...)}}
WHERE {
    ...
    BIND(COALESCE(expr,"SomeDefault") AS ?X)
}
{noformat}

> GROUP_CONCAT returns nothing when concatenating unbound and bound values.
> -------------------------------------------------------------------------
>
>                 Key: JENA-1254
>                 URL: https://issues.apache.org/jira/browse/JENA-1254
>             Project: Apache Jena
>          Issue Type: Bug
>          Components: ARQ
>    Affects Versions: Jena 3.1.0
>            Reporter: Yev Bronshteyn
>         Attachments: c_spdxtest_5530.rdf
>
>
> The following two queries have the same triples in their {{where}} clauses in different order. While it is expected that the first query will return only one solution and the second query - multiple (per https://wiki.blazegraph.com/wiki/index.php/SPARQL_Order_Matters), it is *not* expected that the optional {{copyrightTexts}} value will appear in the result of the first query, but not in the corresponding solution in the result of the second query.
> {code}
> prefix spdx: <http://spdx.org/rdf/terms#>
> prefix rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
> prefix doap:<http://usefulinc.com/ns/doap#>
>  
> select
> 	(group_concat(distinct(?licText);separator="\n\n") as ?licTexts)
>     (group_concat(?copyrightText;separator="\n\n") as ?copyrightTexts)
> 	?name
> where {
>         optional  {
>             ?elem spdx:copyrightText ?copyrightText .
>             filter (isLiteral(?copyrightText))
>         } .
>         {
>             # And the name
>             {?elem spdx:name ?name}
>             union
>             {
>                 ?elem spdx:artifactOf ?artifact .
>                 ?artifact doap:name ?name
>             }
>         }.
> }
> group by ?name
> {code}
> And
> {code}
> prefix spdx: <http://spdx.org/rdf/terms#>
> prefix rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
> prefix doap:<http://usefulinc.com/ns/doap#>
>  
> select
> 	(group_concat(distinct(?licText);separator="\n\n") as ?licTexts)
>         (group_concat(?copyrightText;separator="\n\n") as ?copyrightTexts)
> 	?name
> where {
>         {
>             # And the name
>             {?elem spdx:name ?name}
>             union
>             {
>                 ?elem spdx:artifactOf ?artifact .
>                 ?artifact doap:name ?name
>             }
>         }.
>         optional  {
>             ?elem spdx:copyrightText ?copyrightText .
>             filter (isLiteral(?copyrightText))
>         } .
> }
> group by ?name
> {code}
> The issue seems to be that with the second query, GROUP_CONCAT has to concatenate the bound value of ?copyrightText with unbound values. Changing the second query to unconditionally bind a value to ?copyrightText eliminates the behavior.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)