2010/1/18 François Cassistat <[email protected]>:
> I can't make a minimal example with the exactly the same problem, but there
> seems to have something wrong with this "is null"/"is not null" operation :
>
> Here's a simple repository :
> /a
> /a/@jcr:primaryType=nt:unstructured
> /a/c
> /a/c/@test=value
> /a/c/@jcr:primaryType=nt:unstructured
> /a/c/d
> /a/c/d/@jcr:primaryType=nt:unstructured
>
> workspace.getQueryManager().createQuery("//*...@test='value' and not(d/e)]",
> Query.XPATH).execute().getNodes().nextNode();
>> returns /a/c (which is normal)
>
> workspace.getQueryManager().createQuery("//*...@test='value' and
> not(d/e/@f)]", Query.XPATH).execute().getNodes().nextNode();
>> fails, no results (which is wrong, i think)
The existence/non-existence check only works for properties (same for
NULL/IS NULL in sql). The spec does not require support for relative
paths to properties in those predicates.
Using a subpath like "d/e/@f" will not work reliably, only direct
check like "@f". I think it would be nice to fix that so that it works
consistently, but I guess it's quite difficult to scale that, as the
lucene index used only indexes properties, but not paths. But I am not
so sure about that.
> workspace.getQueryManager().createQuery("//*...@test='value' and d/e]",
> Query.XPATH).execute().getNodes().nextNode();
>> returns /a/c (which is wrong too, i think)
Existence/non-existence does not work for nodes at all.
Regards,
Alex
--
Alexander Klimetschek
[email protected]
