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 26766 - [XSLT30] xsl:merge implementation issues, error conditions and non-streaming behavior
Summary: [XSLT30] xsl:merge implementation issues, error conditions and non-streaming ...
Status: CLOSED FIXED
Alias: None
Product: XPath / XQuery / XSLT
Classification: Unclassified
Component: XSLT 3.0 (show other bugs)
Version: Last Call drafts
Hardware: PC Windows NT
: 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:
: 26765 (view as bug list)
Depends on:
Blocks:
 
Reported: 2014-09-09 23:10 UTC by Abel Braaksma
Modified: 2014-09-12 09:57 UTC (History)
0 users

See Also:


Attachments

Description Abel Braaksma 2014-09-09 23:10:45 UTC
The first two following questions were originally reported to me by my programmer, Eugene Fotin.

While implementing the new xsl:merge syntax, we stumbled upon the following questions, to which I believe the specification has no clear answer:

> what error should be raised if both @for-each-stream and @for-each-item are present? 

I think that, in line with similar constructs that have mutual exclusive attributes, we should introduce a specific static error condition for this.

> what to do when @for-each-stream is present and streamable="no"
> is absent?

From the current text, the semantics of for-each-stream are described in terms of streamability. I think that in the case of the effective value of streamable="no", the behavior should be similar to the fn:collection function (except that its arguments do not return collections, the behavior is more like fn:document), i.e., whether or not it is stable is implementation dependent. If we mean that, I think we should say so explicitly.

> the type of @for-each-stream is xs:string*, however the spec 
> demands it to return uris, in fact, this is a MUST. Making it 
> just xs:string* seems too liberal.

Why not change the argument to be xs:anyURI*? Then a type error would be the logical error if an argument is not a URI.

> the behavior of dereferencing the uris is described as the 
> same as fn:doc. Does this preclude raising the same errors 
> if dererencing fails? (this remark also applies to xsl:stream, 
> which defines no errors of its own).

We are more explicit about error conditions with fn:streaem-available, stipulating the errors of fn:doc-available that it inherits. I think it is worth mentioning this for xsl:stream and xsl:merge/@for-each-stream as well.
Comment 1 Abel Braaksma 2014-09-09 23:11:58 UTC
*** Bug 26765 has been marked as a duplicate of this bug. ***
Comment 2 Michael Kay 2014-09-10 11:00:05 UTC
> what error should be raised if both @for-each-stream and @for-each-item are present?

Agreed, for consistency we should allocate an error code.

> what to do when @for-each-stream is present and streamable="no" is absent?

Presumably you mean streamable="yes" is absent?
I would say, treat for-each-stream="XX" as if they had written for-each-item="(XX)!doc(.)".

> the type of @for-each-stream is xs:string*, however the spec 
> demands it to return uris, in fact, this is a MUST. Making it 
> just xs:string* seems too liberal.

It is our consistent practice to accept xs:string where a URI is required, e.g. in doc(), xsl:stream, resolve-uri() etc. An xs:anyURI is OK, because of type promotion.

> the behavior of dereferencing the uris is described as the 
> same as fn:doc. Does this preclude raising the same errors 
> if dererencing fails? (this remark also applies to xsl:stream, 
> which defines no errors of its own).

Yes, we should add this: The process of obtaining a document node given a URI is the same as for the docFO30 function (including error conditions).
Comment 3 Michael Kay 2014-09-12 09:12:17 UTC
The WG agreed with the proposed changes.
Comment 4 Michael Kay 2014-09-12 09:22:23 UTC
I will take the opportunity to add xsl:stream and xsl:merge-source to the list of elements to which error XTSE1505 applies (validation and type are mutually exclusive).
Comment 5 Michael Kay 2014-09-12 09:36:18 UTC
I have also taken the opportunity to specify, for avoidance of doubt, that [xsl:]default-validation has no effect on whether xsl:stream and xsl:merge-source perform validation; they perform no validation unless explicitly requested.
Comment 6 Michael Kay 2014-09-12 09:46:16 UTC
Comment #3 is an misrepresentation of the WG decision, especially for para 2. Here is the actual decision:

DECISION para 1: the WG accepted allocating an error code for the mutual exclusive arguments

DECISION para 2: instead of the proposed solution, the WG agreed with requiring streamable="yes" if for-each-stream attribute is present

DECISION para 3: no change, xs:string is in line with other functions and constructs

DECISION para 4: marked editorial, to add error conditions
Comment 7 Michael Kay 2014-09-12 09:57:53 UTC
Changes applied. The error code allocated is XTSE3195:

<error class="SE" code="3195"><p>The two attributes <code>for-each-item</code> and
               <code>for-each-stream</code> are mutually exclusive. The <elcode>xsl:merge-source</elcode> element may contain
               either or neither, but not both. It is a <termref def="dt-static-error"/> if both are present.
               Furthermore, if the <code>for-each-stream</code> attribute is present, the only permitted value (and the default value)
               of the <code>streamable</code> attribute is <code>yes</code>.
            </p></error>