59937 - XPath and XSLT function lib incomplete

Status: VERIFIED FIXED

(Core :: XSLT, defect, P3)

Description

[Axel Hecht]

The following functions are not implemented yet:

id()
normalize-space()
sum()
key()
unparsed-entity-uri() // what the heck does this? Example?
element-available()
function-available()

Axel

Comment 1

[Axel Hecht]

I assign this bug to me, at least for now.
I implemented the first three functions, patch will come in a minute.
I have no idea about the others, though. So somebody will have to take over.

Axel

Comment 2

[Axel Hecht]

Attachment: patch for sum() id() normalize-space(), includes 59649

Comment 3

[Axel Hecht]

More spam, accepting bug.
Please review this one, I'm not sure I did all the memory issues allright, and 
I didn't test the mozilla part. As I don't have a working example for the id()
function :-(

Axel

Comment 4

[Axel Hecht]

OK,
I have looked thru the problems with the id() function. We definitly need to 
work dependent on DTD, it's not the "id" attribute, but a as ID specified
attribute. Checking for "id" would work for HTML4, but we don't set the 
document type. Therefor I can't check that. I will leave the implementation
for getElementById() in, and return NULL for the standalone.
How is that?

Axel

Comment 5

[Axel Hecht]

I will attach a new patch, that does id() better, my code in 
standalone/Document.ccp was under a false impression about what
NodeDefinition::nextSibling is. I completely removed that code in favor of a
return NULL; and some comments.
I didn't return the NodeSet in the id() function, too, and didn't resolve the
function name.
The current patch runs most (I didn't test all) examples from docbook and quite
a few from the XSLT benchmark.

Adding format-number() to the list of unimplemented.
I will file a bug on the benchmark tests, a good amout of work there.

I'd like a review, I think the memory stuff is good, looking at it awake.

Thanx to dr and jag for pointing my nose on to my shameless assumptions on the
id() impl for standalone. We essentially have none, as we don't resolve DTDs 
and don't set the doctype, AFAI understood the code.

Adding dr to CCs

Axel

Comment 6

[Axel Hecht]

Attachment: new patch, has bug 59649, bug 60059, sorry, only one source tree

Comment 7

[Axel Hecht]

Attachment: oh, darn, forgot to recreate diff, this one is it :-(

Comment 8

[Axel Hecht]

The number functions should be revisited, to return NaN in error cases instead
of 0.0.

Axel

Comment 9

[Axel Hecht]

fix is in, r=peterv. Peter's code r=me.

xpath function lib complete

xslt functions key(), format-number(), unparsed-entity-uri(), 
element-available() and function-available() not yet implemented, but parsed
correctly and error handled.

Axel

Comment 10

[Dan Rosen]

Sounds good... You should probably be able to implement element-available() and
function-available() trivially... I'd suggest closing this bug (as fixed) in
favor of specific ones for the small number of leftover bits.

Comment 11

[Peter Van der Beken [:peterv]]

Marking fixed. Bugs have been filed per remaining function.

Comment 12

[Axel Hecht]

We haven't been verifying bugs for quite a while :-(