RE: support for aligned iterations

You can achieve this in XPath 2.0 using constructs such as:

every $i in 1 to count(/msg/a) satisfies
   /msg/a[$i]/a1 = /msg/b[$i]/b1


Regards,

Michael Kay


> -----Original Message-----
> From: winkowski@mitre.org [mailto:winkowski@mitre.org] 
> Sent: 30 May 2003 16:53
> To: public-qt-comments@w3.org
> Subject: support for aligned iterations
> 
> 
> 
> I have a document such as:
> 
> <msg>
>  <a>
>   <a1>xxx</a1>
>  </a>
> <a>
>   <a1>xxx</a1>
>  </a>
>  <a>
>   <a1>xxx</a1>
>  </a>
> <b>
>   <b1>xxx</b1>
>  </b>
> <b>
>   <b1>xxx</b1>
>  </b>
>  <b>
>   <b1>xxx</b1>
>  </b>
> <c>
>   <c1>xxx</c1>
>  </c>
> <c>
>   <c1>xxx</c1>
>  </c>
>  <c>
>   <c1>xxx</c1>
>  </c>
> </msg>
> 
> I need to apply a test/operator over /msg/a/a1 and /msg/b/b1 
> such that the evaluation is aligned over corresponding pairs 
> of a1,b1 in document order. That is, test the first 
> corresponding pair /msg/a[1]/a1 and /msg/b[1]/b1, the second 
> corresponding pair /msg/a[2]/a1 and /msg/b[2]/b1, etc...
> 
> I am deriving the XPath expression directly from a constraint 
> language and would like to stick as close to XPath as 
> possible. These XPath expressions could then be compiled into 
> a language such as XSL or Java for evaluation. 
> 
> I found I needed to introduce a new variable notation ($$) 
> into XPath to indicate this. Say the test was equality then 
> an example is: /msg/a[position=$$A]/a1 = /msg/b[position=$$A]/b1
> 
> As a result I no longer produce XPath 1.0 but something else. 
> These new variables need to be addressed in the compilation 
> stage when building XSL or Java code. My question is will 
> XPath 2.0 support document ordered iteration over multiple paths? 
> 
> The constraint language consists of a statement and 
> condition. So a more complex rule stating that if 
> corresponding c1's are greater than the corresponding b1's 
> then the corresponding a1's must equal the corresponding
> b1's:
> (/msg/a[position=$$A]/a1 = /msg/b[position=$$A]/b1) and 
> (/msg/c[position=$$A]/c1 > /msg/b[position=$$A]/b1) 
> 
> Complex rules may result in multiple corresponding variables 
> as well ($$A, $$B, ...).
> 
> - Dan
> 
> 
>

Received on Monday, 2 June 2003 10:59:06 UTC