Path expressions on match templates don't work with path segments encountered before the match directive

[cmlenz]

Before [18], the following worked:

<div xmlns:py="markup.edgewall.org/">
  <span py:match="div/greeting">
    Hello ${select('@name')}
  </span>
  <greeting name="Dude" />
</div>

Since [18], the patch expression “div/greeting” no longer matches because the <div> element is encountered before the py:match directive in the input. This should probably be fixed, though. One option is to keep a stack of already encountered elements in the Template._match filter and push them into the path evaluator so that it has the correct context.

[cmlenz]

Not going to get this in for 0.2.

[cmlenz]

And again…

[cmlenz]

This is not high priority right now, so clear the milestone.

[Carsten Klein <carsten.klein@…>]

Since the span represents the context node in which the matcher is being applied in, simply matching "div/greeting" would look for a child element "div" on the "span" element.

This is actually the correct way to implement it, see ​http://www.w3.org/TR/xpath/#path-abbrev for a few examples.

For example "divgreeting" would match all grandchildren of the context element, which is "span" that are descendants of a given "div" child element.

Does genshi know about ".." as a selector? If so, then "../greeting" would be the correct way to do it, or just "div[...]/greeting" for the not so optimal selection.

What do you think?