Re: [netmod] XPath node type tests

[Martin Bjorklund <[e-mail]>]

Robert Wilton <[e-mail]> wrote:
> 
> 
> On 23/10/2017 10:10, Martin Bjorklund wrote:
> > Andy Bierman <[e-mail]> wrote:
> >> On Fri, Oct 20, 2017 at 9:24 AM, Robert Wilton <[e-mail]>
> >> wrote:
> >>
> >>> Hi Lada,
> >>>
> >>> Thanks for the explanation, that makes sense.
> >>>
> >>>
> >>> On 20/10/2017 16:27, Ladislav Lhotka wrote:
> >>>
> >>>> Hi Rob,
> >>>>
> >>>> Robert Wilton <[e-mail]> writes:
> >>>>
> >>>> Hi,
> >>>>> XPATH 1.0 defines the following three node-type tests:
> >>>>>
> >>>>> 1) comment()
> >>>>> 2) processing-instruction(<opt arg>)
> >>>>> 3) text()
> >>>>>
> >>>> For completeness, node() is the fourth one.
> >>>>
> >>>> My assumption is that a YANG tree doesn't contain any nodes of type
> >>>>> 'comment' or 'processing-instruction' and hence these filters would
> >>>>> never match any nodes.
> >>>>>
> >>>> Yes. FWIW, Yangson library raises NotSupported exception upon
> >>>> encountering these.
> >>>>
> >> But a server or client should ignore PIs, not reject the XML.
> >>
> >> I think text() and node() are just filter tests.
> >>
> >>    /foo/*[text()] would return all the child nodes of /foo that are leaf
> >>    or
> >> leaf-list
> >>
> >> text() returns a boolean (0 or 1).  Do not use it for value testing:
> > No.  text() will select the text node children of the context node.
> This is presumably because text() is evaluated as "child::text()".

Yes.

> >>    /foo/*[text() = 'fred']  // wrong!
> > This actually works.  text() selects all text nodes (just one for a
> > leaf), and then that text node is compared to the string 'fred'.
> For clarity, am I right in my interpretation that a leaf is not itself
> a text node, but instead a leaf is an element node that contains a
> direct child text node?

Yes.

> Presumably, it is only leaf and leaf-list element nodes that can have
> these direct child text nodes.

Yes.

> I can see how this make sense for a XML document, but it does feel a
> bit non intuitive for a YANG data tree

Maybe, but since we use XPath, we need to conform to the data model
used by XPath (see section 5 of the xpath spec).

> and it may be helpful if this
> is documented somewhat ...

RFC 7950 refers to the data model of XPath (See section 6.4 of RFC
7950), but I agree that it could have had more text.  Specifically, it
could have stated how nodes are mapped to elements, that only
leaf/leaf-list have text nodes; that annotations are mapped to
attribute nodes (ok, not really in 7950...); that there are no
processing-instruction and comment nodes.

> 
>   /foo/*[. = 'fred']  // correct
> 
> Presumably this test isn't quite the same, since child container and
> list nodes would also be included in the comparison (i.e. by
> concatenating all their descendant leaf values together into a single
> string)
> whereas the expression with the text() check will only
> include the values of direct child leaf and leaf-list nodes (as YANG
> is currently defined today).

Yes.


/martin