Re: [Exist-development] help debugging "no supertype for empty()" warning when importing functx

[Adam R. <ad...@ex...>]

> However whilst experimenting to find a short reproducible test-case, I
> came upon this:
>
> xs:string(
>   if(current-date() eq current-date() - xs:dayTimeDuration("P1D"))then
>     ()
>   else
>     "thing"
> )
>
> Now this code should do exactly the same as the above code, as both ()
> and fn:error(...) are evaluated to an empty sequence. However they
> dont, and this code does not cause the warning you were seeing. In
> fact if I debug it, it seems that () is being evaluated not to the
> type "empty()" but instead to the type "node()".
>
> This definitely stikes me as a bug in eXist, as () is certainly an
> empty sequence and not a node...
>
> Wolf any thoughts please?

Digging a little deeper into this one, it looks to me like
PathExpr.returnsType() should be re-written from:

public int returnsType() {
        if (steps.size() == 0)
        //Not so simple. ITEM should be re-tuned in some circumstances
that have to be determined
            return Type.NODE;
        return steps.get(steps.size() - 1).returnsType();
    }

to:

@Override
    public int returnsType() {
        if(steps.isEmpty()) {
            return Type.EMPTY;
        } else {
            return steps.get(steps.size() - 1).returnsType();
        }
    }


I havent run the test suite yet, but a quick play with eXist and a few
demo apps and all appears to work - I will run the test suite
shortly...

--
Adam Retter

eXist Developer
{ United Kingdom }
ad...@ex...
irc://irc.freenode.net/existdb