Re: [Exist-open] Odd behaviour with fn:doc

[Martin H. <mh...@uv...>]

Hi there,

Pierrick Brihaye wrote:
> Hi,
> 
> [please follow RFC 1855 - Netiquette - rules if you want me to help you]

I was actually just reporting eXist's behaviour; the problem isn't there 
if I write good XQuery (with node() as the function return value).

> 
> Martin Holmes a écrit :
> 
>> This is interesting. Using your simplified code below, I do indeed
>> get an error popup. However, if I use my code (with a couple of
>> syntax errors fixe), then I get a freeze that has to be killed
>> through Task Manager. Here's the code:
>>
>> declare namespace f="http://exist-db.org/f-functions";
>>
>> declare function f:GetDoc() as element()* { let $fname :=
>> 'emls1-1.xml', $coll := '/db/emls/', $docpath := concat($coll,
>> $fname) return if (fn:doc-available($docpath)) then fn:doc($docpath) 
>> else <p>{$docpath} is not available.</p>
>>
>> };
>>
>> f:GetDoc()
> 
> This is mine, which *has to* be equivalent :

I guess that would depend on what's in your database, really.

> declare function local:GetDoc() as element()* {
>     fn:doc("/db/test.xml")
> };
> local:GetDoc()
> 
> I get :
> 
> An exception occurred during query execution: FORG0001: document(Table)
> is not a sub-type of element [at line 1, column 30]
> In call to function:
> 	local:GetDoc() [4:1]
> 
> ... which is normal IMHO.

That's what I would expect. Nevertheless, it's not what happens; I get a 
freeze.

>> Now, it only dies if the document exists in the db; otherwise it
>> returns the p tag with the "not available" message.
> 
> How big is your /db/emls/emls1-1.xml document ? Aren't you simply 
> flooding the result window ?

I don't think so; I can get the doc back using other methods (such as 
straight XPath).
> 
>> So I think what's happening is that eXist is retrieving the document,
>>  but at the point that it's delivering it as the result of the
>> function, the mismatch between the function return type and the
>> result of the query causes a fatal freeze.
> 
> My example seems to demonstrate that it is not the case, doesn't it ?
> 
> I may be missing something obvious but I really can't understand how 
> your query could freeze the client unless you have exhausted the GUI 
> capabilities. What about, for example, assigning more memory to the client ?

I'll try that after Christmas. But if you're sure it must be my fault 
and can't be a bug, then I'm happy to stop worrying about it and we can 
both get on with something else :-)

Cheers,
Martin