Re: [Exist-development] bug @fn:string

[Pieter D. <pie...@gm...>]

On Mon, Aug 9, 2010 at 3:39 PM, Dmitriy Shabanov <sha...@gm...>wrote:

>  On Mon, 2010-08-09 at 11:50 +0200, Pieter Deelen wrote:
>
> Let's forget about the string function for a moment and introduce function
> f:
>
> declare function local:f($elt as element()) {
>   "f"
> };
>
> Then if I evaluate the following expression
>
> (<a/>, <b/>)/local:f(.)
>
> I expect (and get from both eXist 1.4 and Saxon 9)
>
> ("f", "f")
>
> So the function f is called twice, the first time with <a/> as parameter,
> and the second time with <b/>.
>
> Similarly, (<a/>)/local:f(.) returns ("f"): the function f is called once,
> for <a/>.
>
> So if I evaluate
>
> ()/local:f(.)
>
> I would expect the function f to be called not at all and the empty
> sequence as result. Instead, eXist shows the following error:
>
> "The actual cardinality for parameter 1 does not match the cardinality
> declared in the function's signature: local:f($a as element()) item()*.
> Expected cardinality: exactly one, got 0."
>
> I.e., eXist tries to call the function f, but this fails because the
> function's signature does not allow it.
>
> Saxon gives me the result I expect.
>
>
> Change function definition & getting more serious, low level, engine bug
>
> ======================================================
> declare function local:f($elt) {
>   "f"
> };
>
> ()/local:f(.)
> ======================================================
>
> return ("f"), that is not correct.
>
> Can we split the bug into two:
>
> 1. xs:string( () ) should be empty string
>
There are a few subtle differences between xs:string (which is a type
constructor) and fn:string (which is a standard function). As far as I am
aware, xs:string(()) should yield () and fn:string(()) should yield "". I
can't find the exact spec right now, but both eXist and Saxon produce the
same results.


> 2. function call on empty sequence
>
That was the actual problem I was hinting at. I had hoped it would be
clearer if I used standard functions.


>
>   --
> Cheers,
>
> Dmitriy Shabanov
>
>