Ha, nice MarkLogic trick. :-)


Other XSLT parsers usually expect a well-formed input document. But then
again, with XML-ER, an empty sequence might just be well-formed..



:-DD



*Van:* [email protected] [mailto:
[email protected]] *Namens *Evan Lenz
*Verzonden:* donderdag 23 februari 2012 21:02
*Aan:* MarkLogic Developer Discussion
*Onderwerp:* Re: [MarkLogic Dev General] weird xpath behaviour



Norm Walsh added this quite right correction (forwarding 'cause he's having
email issues):



Evan Lenz <[email protected]> writes:

In XSLT in MarkLogic, "/" is never short for "collection()/", but then a
context item is

usually defined in XSLT (everywhere, I think, except inside a user-defined
function).



I think you can also get an undefined context item by explicitly passing an

empty sequence as the input and starting the stylesheet at a named template:



  xdmp:xslt-invoke("style.xsl", (), (),

       <options xmlns="xdmp:eval"><template>{fn:QName("",
"start-here")}</template></options>)



                                        Be seeing you,

                                          norm





*From: *Evan Lenz <[email protected]>
*Reply-To: *MarkLogic Developer Discussion <[email protected]>
*Date: *Thu, 23 Feb 2012 11:20:32 -0800
*To: *MarkLogic Developer Discussion <[email protected]>
*Subject: *Re: [MarkLogic Dev General] weird xpath behaviour



Yeah, when there's a context item in 1.0-ml (and always in 1.0), "/" is
short for "root(.) treat as document-node()/", which means it throws an
error when the context item isn't a node (e.g. 42[/x]). In 1.0-ml, if
there's no context item (such as at the top level of a query), then "/" is
short for "collection()/". In 1.0, it would be an error.



In practice, this means that leading "/" can mean two things in 1.0-ml:

   - At the top level of a query: "all documents in the database"
   - Inside a predicate: "root of the current document"

In XSLT in MarkLogic, "/" is never short for "collection()/", but then a
context item is usually defined in XSLT (everywhere, I think, except inside
a user-defined function).



Evan Lenz

Software Developer, Community
MarkLogic Corporation
community.marklogic.com

*From: *Geert Josten <[email protected]>
*Reply-To: *MarkLogic Developer Discussion <[email protected]>
*Date: *Thu, 23 Feb 2012 06:55:06 -0800
*To: *MarkLogic Developer Discussion <[email protected]>
*Subject: *Re: [MarkLogic Dev General] weird xpath behaviour



The fact that MarkLogic expands a / at the beginning of an XPath is just a
convenient extension. That a bare / as start within a predicated works like
root()/ is conform spec to my knowledge.



Kind regards,

Geert



*Van:* [email protected] [mailto:
[email protected]] *Namens *Szabolcs Illes
*Verzonden:* donderdag 23 februari 2012 15:50
*Aan:* MarkLogic Developer Discussion
*Onderwerp:* Re: [MarkLogic Dev General] weird xpath behaviour



Hi Geert,



That explains it, is it a bug  or a feature?  ( The same thing works as
expected in eXists … )



Cheers,

Szabolcs



*From:* [email protected]
[mailto:[email protected]] *On Behalf Of *Geert Josten
*Sent:* 23 February 2012 14:21
*To:* MarkLogic Developer Discussion
*Subject:* Re: [MarkLogic Dev General] weird xpath behaviour



Hi Szabolcs,



Evan pointed out the following to me recently. / at the start of an XPath
expression is interpreted within MarkLogic as ‘collection()/’. A / within a
predicate,  however, is not. It is effectively interpreted as ‘root()/ ‘.



Kind regards,

Geert







*Van:* [email protected] [mailto:
[email protected]] *Namens *Szabolcs Illes
*Verzonden:* donderdag 23 februari 2012 13:54
*Aan:* [email protected]
*Onderwerp:* [MarkLogic Dev General] weird xpath behaviour



Hi,



I have seen this before, but came across again, so I thought ask this time.
Same/similar xquery but different results:



*Version 1:*



let $ids := /HorseRacingCard/Meeting[@country eq "England" and
AdvancedGoing]/@id

return (/HorseRacing/Meeting[@id = $ids] )[1]



returns expected results.



*Version 2*: Same as 1 but ids vatriable is just inlined



(/HorseRacing/Meeting[@id = /HorseRacingCard/Meeting[@country eq "England"
and AdvancedGoing]/@id ] )[1]



returns nothing [ slow, but that’s expected: ]



Other than Version 2 is NOT optimal since the expression inside the
predicate is evaluated for each sequence item, I guess it should return the
same results.



*Version 3:*

(/HorseRacing/Meeting[@id =* fn:collection()/*HorseRacingCard/Meeting[@country
eq "England" and AdvancedGoing]/@id ] )[1]



Returns expected results and it’s fast, expression is only evaluated once



Can someone shed some light why Version 2 is not returning any results?
Could it be a bug in the optimizer?



Thanks,

Szabolcs
_______________________________________________
General mailing list
[email protected]
http://developer.marklogic.com/mailman/listinfo/general

Reply via email to