[xquery-talk] Error in unexecuted(?) code

[Michael Kay]

XQuery processors are in fact given a surprising amount of latitude to
rewrite expressions in ways that can cause errors to be reported that would
not be reported if they were evaluated as written. However, there's an
explicit guarantee for "if" and "typeswitch": "Conditional and typeswitch
expressions raise only dynamic errors that occur in the branch that is
actually selected." (Section 2.3.4). So I think DB2 is wrong here.

Michael Kay
http://www.saxonica.com/


> -----Original Message-----
> From: talk-bounces at x-query.com 
> [mailto:talk-bounces at x-query.com] On Behalf Of Ronald Bourret
> Sent: 12 March 2007 05:22
> To: xquery-talk
> Subject: [xquery-talk] Error in unexecuted(?) code
> 
> Hello,
> 
> I have a data structure where a <timeinfo> element can have 
> either a <sngdate> (single date) child or a <rngdate> (range 
> of dates) child. The following query tests for which child is 
> present and constructs an xs:date appropriately.
> 
> for $timeinfo in
> <timeinfo><rngdates><begdate>19550101</begdate><enddate>195501
> 01</enddate></rngdates></timeinfo>
> return
> <date>
> {
>     fn:min(if ($timeinfo/sngdate)
>            then let $fgdcDate:=$timeinfo/sngdate/caldate
>                 return xs:date(
>                         fn:concat(
>                           fn:substring($fgdcDate, 1, 4),
>                           "-",
>                           fn:substring($fgdcDate, 5, 2),
>                           "-",
>                           fn:substring($fgdcDate, 7, 2)))
>            else let $fgdcDate:=$timeinfo/rngdates/begdate
>                 return xs:date(
>                         fn:concat(
>                           fn:substring($fgdcDate, 1, 4),
>                           "-",
>                           fn:substring($fgdcDate, 5, 2),
>                           "-",
>                           fn:substring($fgdcDate, 7, 2)))
>            )
> }
> </date>
> 
> This executes correctly in Saxon 8, but fails in DB2 v9 with 
> the following error:
> 
>     The value "--" cannot be constructed as, or cast (using
>     an implicit or explicit cast) to the data type "xs:date"
> 
> The error goes away if I remove the xs:date constructor from 
> the "then" 
> clause. For example:
> 
> for $timeinfo in
> <timeinfo><rngdates><begdate>19550101</begdate><enddate>195501
> 01</enddate></rngdates></timeinfo>
> return
> <date>
> {
>     fn:min(if ($timeinfo/sngdate)
>            then $timeinfo/sngdate/caldate
>            else let $fgdcDate:=$timeinfo/rngdates/begdate
>                 return xs:date(
>                         fn:concat(
>                           fn:substring($fgdcDate, 1, 4),
>                           "-",
>                           fn:substring($fgdcDate, 5, 2),
>                           "-",
>                           fn:substring($fgdcDate, 7, 2)))
>            )
> }
> </date>
> 
> This makes me believe that the then clause is being executed 
> (even though sngdate is not present) with an empty sequence. 
> An empty sequence would cause concat to return "--" and 
> xs:date() to return the aforementioned error.
> 
> Is a DB2 bug (my guess) or am I missing something subtle? For 
> example, perhaps XQuery processors can execute any code they 
> like and only take the results that apply? This would 
> contradict the statement in 3.10 of the XQuery spec that:
> 
> "Similarly, if the effective value of the test expression is 
> false, the conditional expression ignores any dynamic errors 
> encountered in the then-expression, and the then-expression 
> need not be evaluated."
> 
> -- Ron
> 
> _______________________________________________
> talk at x-query.com
> http://x-query.com/mailman/listinfo/talk