<xml>
<table>
<rec id="1">
<para type="error" position="11"/>
<para type="warning" position="12"/>
<para type="warning" position="13"/>
</rec>
<rec id="2">
<para type="warning" position="21"/>
<para type="warning" position="22"/>
<para type="warning" position="23"/>
</rec>
<rec id="3">
<para type="info" position="31"/>
<para type="warning" position="32"/>
<para type="warning" position="33"/>
</rec>
</table>
</xml>
For XPath _expression_ "//rec/para[1]", xmllint.exe outputs:
<para type="error" position="11"/><para type="warning" position="21"/><para type="info" position="31"/>
While my colleague said that the output should be: <para type="error" position="11"/>
According to his explaination, only after all evaluations of E2 against all nodes resulting from the evaluation of E1, the evaluation of E3 can begin. As for the example, the process looks like:
rec(id = "1") -> all para under rec (id = "1") -> rec(id = "2") -> all para under rec (id = "2") -> rec(id = "3") -> all para under rec (id = "3") , the result sequence then acts as the input
sequence of the evaluation of E3, so only the very first para node should be ouput. To me, this is a little bit like BFS (Breadth-First-Search).
On the contrary, I think the process should be:
rec(id = "1") -> all para under rec (id = "1") -> get the first para under rec (id = "1") -> rec(id = "2") -> all para under rec (id = "2") -> get the first para under rec (id = "2") -> rec(id = "3") -> all para under rec (id = "3") -> get the first para under rec (id = "3"). The first para node under each rec should be output, as xmllint.exe did above. Could be regarded as DFS (Depth-First-Search)?
I havn't found a clear definition about the parsing rule from the XPath spec. Or you can say that I have not understood the spec well :). Anyway, I cannot persuade him even I have shown him many XML tools that really performs the same
with xmllint.exe. Could someone give me the theoretical support?
My colleague used another XPath _expression_ to support his opinion: "//rec[1]/para[ type="warning"][2]" (Please focus on the latter part: "para[ type="warning"][2]", I purposely used rec[1] to avoid mix-up).
In this case, all evaluations of E2 against all nodes resulting from the evaluation of E1 have been done before the evaluation of E3. The output of para[ type="warning"] act as the input sequence of E3. Sounds reasonable? BFS?
As descibed in section 3.3.2 Predicates of XPath spec 2.0 (1.0 should follow it too): In the case of multiple adjacent predicates, the predicates are applied from left to right, and the result of applying each predicate serves as the input sequence for the following predicate.
So, what's the
nice distinction between E1/E2[E3] and E1[E2][E3]. They both evaluate using inner focus, a paragraph of the spec lists both together as if they completely consistent (but do they really follow the same parsing rule?):
Cited from 2.1.2 Dynamic Context (Spec 2.0): Certain language constructs, notably the path _expression_ E1/E2 and the predicate E1[E2], create a new focus for the evaluation of a sub-_expression_. In these constructs, E2 is evaluated once for each item in the sequence that results from evaluating E1. Each time E2 is evaluated, it is evaluated with a different focus. The focus for evaluating E2 is referred to below as the inner focus, while the focus for evaluatingE1 is referred to as the outer focus. The inner focus exists only while E2 is being evaluated. When this evaluation is complete, evaluation of the containing _expression_ continues with its original focus unchanged.
Any commens would be appreciated.
Hi Liam,
Could you have a look at this :)
Thanks,
Ming
