fn:has-attributes

[ChristianGruen]

fn:has-attributes

[michaelhkay]

The has-children() function was introduced specifically to support XSLT streaming, as a way of testing (by a single-token lookahead) whether children exist, without actually swallowing them as the equivalent exists(child::node()) would do. It was then moved from XSLT into XPath (along with other streaming-oriented things such as innermost() and outermost()) on the basis of "no needless differences".

The functionality of has-attributes() can be readily achieved using exists(@*) and I really see no need for a special function to do that.

[michaelhkay]

Wouldn’t that be a typical task for an optimizer (rewriting exists(node()) to an internal has-children() check)?

We made a decision (rightly or wrongly) that XSLT streamability shouldn't depend on optimization rewrites, in order to make the rules fully interoperable.

The statement that "the function has general utility as a convenience function" clearly indicates, to someone experienced in reading standards documents, that the WG knew the justification was weak; features don't attempt to justify themselves otherwise.

[ChristianGruen]

The has-children() function was introduced specifically to support XSLT streaming, as a way of testing (by a single-token lookahead) whether children exist, without actually swallowing them as the equivalent exists(child::node()) would do.

Wouldn’t that be a typical task for an optimizer (rewriting exists(node()) to an internal has-children() check)?

The specification states that the function has “general utility as a convenience function”. I think the argument would also be valid for an attribute function.

After all, it probably takes more time to talk about the function than just adding or ignoring it, so… I have no strong opinion.

[michaelhkay]

I don't think this function adds much value and I propose we turn down the suggestion.

[ndw]

The CG agreed to close this issue without action at meeting 064