29887 – [XSLT30] Assertion on attributes of xsl:for-each-group is too strict, does not take shadow attribs into account

This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.

Bug 29887 - [XSLT30] Assertion on attributes of xsl:for-each-group is too strict, does not take shadow attribs into account

Summary: [XSLT30] Assertion on attributes of xsl:for-each-group is too strict, does no...

Status:	RESOLVED FIXED

Alias:	None

Product:	XPath / XQuery / XSLT
Classification:	Unclassified
Component:	XSLT 3.0 (show other bugs)
Version:	Candidate Recommendation
Hardware:	PC Windows NT

Importance:	P2 normal
Target Milestone:	---
Assignee:	Michael Kay
QA Contact:	Mailing list for public feedback on specs from XSL and XML Query WGs

URL:
Whiteboard:
Keywords:

Depends on:
Blocks:

Reported:	2016-09-29 22:12 UTC by Abel Braaksma
Modified:	2016-10-13 21:38 UTC (History)
CC List:	0 users

See Also:

Attachments

Description Abel Braaksma 2016-09-29 22:12:12 UTC

This bug is courtesy of Charles Foster, he noticed it in a Skype chat.

In de XSD we have:

count((@group-by, @group-adjacent, @group-starting-with, @group-ending-with)) = 1

But this doesn't work with shadow-attributes. I propose something like the following should work (not tested):

count((
(@group-by|@_group-by)[1], 
(@group-adjacent|@_group-adjacent)[1], 
(@group-starting-with|@_group-starting-with)[1],
(@group-ending-with|@_group-ending-with)[1]
)) = 1

Rationale: if both the shadow- and normal attribute exists, the shadow attribute takes precedence. This also accounts for situations where you would end up with @_group-by and @group-ending-with appearing, which is wrong, because the first will end up after de-shadowing as @group-by.

Comment 1 Michael Kay 2016-10-13 15:23:00 UTC

There are a few other cases of the same problem.

Line 180
<xs:assert test="not(exists(@select) and (exists(* except xsl:fallback) or exists(text()[normalize-space()])))"/>

exists(@select) ==> exists(@select|@_select)

Line 327, 690, 728, 1374
<xs:assert test="every $e in subsequence(xsl:sort, 2) satisfies empty($e/@stable)">

$e/@stable ==> $e/(@stable|@_stable)

Line 371, 503, 530, 552, 615
<xs:assert test="not(exists(@type) and exists(@validation))">

==> not(exists(@type|@_type) and exists(@validation|@_validation))

Line 737

count((@group-by, @group-adjacent, @group-starting-with, @group-ending-with)) = 1

==> count((@group-by|@_group-by, @group-adjacent|@_group-adjcent, @group-starting-with|@_group-starting-with, @group-ending-with|@_group-ending-with)) = 1

(could use the comma operator throughout but it feels more expressive this way)

Line 746

if (exists(@collation) or exists(@composite)) then (exists(@group-by) or exists(@group-adjacent)) else true()"

==> as elsewhere, replace @X by @X|@_X

Line 807, 814, 821, 890, 

Line 1085
test="not(exists(@name) and normalize-space(@visibility) = 'private' and exists(xsl:context-item))"

this one is more tricky because it's looking at attribute values. But the rule is obsolete because xsl:mode no longer has an xsl:context-item child. Slightly surprised Saxon's static type checking lets that through.

line 1182
<xs:assert test="every $prefix in (@stylesheet-prefix, @result-prefix)[. ne '#default']
            satisfies $prefix = in-scope-prefixes(.)"/>

we can keep this as is.

line 1186
test="if (exists(@value)) then empty((@select, @count, @from)) and @level='single' else true()">
            
Usual @X by @X|@_X except that @level='single' becomes (@level='single' or @_level)

line 1134
<xs:assert test="if (normalize-space(@static) = 'yes')
                           then empty((*,text()))
                           else true()">
Leave as is.

etc. etc. All the remaining assertions fit into one of the above patterns.

Comment 2 Michael Kay 2016-10-13 21:38:47 UTC

The schema has been updated and tested to fix these problems.