xsl-list
[Top] [All Lists]

Re: [xsl] W3C Specification of fn:filter() -- is this a bug in the document or in Saxon?

2019-09-09 09:12:57
The alternative formulation wouldn't change anything. It would still have the 
same theoretical weakness that the rewritten expression might use different 
resources and therefore fail under different circumstances. It might make it 
less likely that the two formulations would differ in practice, but this is a 
specification, not a suggested implementation.

The convention in specifications is to ignore efficiency considerations when 
specifying functionality. Saying that A is equivalent to B carries implicit 
caveats, like, "provided you have enough memory and no-one turns the power 
switch off while waiting for it to finish".

Michael Kay
Saxonica

On 9 Sep 2019, at 14:20, Dimitre Novatchev dnovatchev(_at_)gmail(_dot_)com 
<xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com> wrote:

I'm aware that some languages have attempted to formulate rules in the 
language semantics making tail call optimization mandatory. The XSL and 
 > XQuery WGs considered several times whether to try and make the whole 
"errors and optimization" rules more formal and rigorous, and we decided we 
didn't have the skills and resources to do it, for the same reason that 
work on the XQuery formal semantics was abandoned.
  
Michael Kay
Saxonica

The original problem can be eliminated (and the same solution may be 
applicable in similar cases), if the "equivalent implementations" were 
replaced with non-recursive code, As in this case -- just use:

function($f as function(item()) as xs:boolean, $list as item()*) as item()*
{
  $list ! .[$f(.)]
} 

Thanks,
Dimitre 


On Sun, Sep 8, 2019 at 10:22 PM Michael Kay mike(_at_)saxonica(_dot_)com 
<mailto:mike(_at_)saxonica(_dot_)com> 
<xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com 
<mailto:xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com>> wrote:
The "errors and optimization" rule in XPath says that processors can quite 
legitimately rewrite one expression with another that has different resource 
requirements and that therefore has different failure characteristics. This 
is by design. It means that either of these formulations could fail with a 
stack overflow, and in that sense they are indeed equivalent.

I'm aware that some languages have attempted to formulate rules in the 
language semantics making tail call optimization mandatory. The XSL and 
XQuery WGs considered several times whether to try and make the whole "errors 
and optimization" rules more formal and rigorous, and we decided we didn't 
have the skills and resources to do it, for the same reason that work on the 
XQuery formal semantics was abandoned.

Michael Kay
Saxonica

On 9 Sep 2019, at 02:44, Dimitre Novatchev dnovatchev(_at_)gmail(_dot_)com 
<mailto:dnovatchev(_at_)gmail(_dot_)com> 
<xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com 
<mailto:xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com>> wrote:

 You can never guarantee that two expressions are equivalent in your
sense, because of "errors and optimization". Any construct might raise
an error - in the case of this example, stack overflow if the recursion
gets too deep.

What about tail-recursion?  

For years we have known recursive expressions whose tail-recursiveness is 
correctly recognized in BaseX and it provides correct evaluation regardless 
of the input size (recursion depth) but other processors fail miserably...

How much value for the developers would have been provided by the 
specification if it mandated proper handling of tail-recursion!!!

The value provided in a document is rather debatable when specifying 
"equivalent implementations" that blow up for reasonably long inputs 
(several thousand items isn't too high!) when other implementations could 
have been provided that demonstrate equivalence with much longer inputs 
(millions of items)

Also, why in an XPath specification give "equivalent implementations" in two 
different languages neither of which is XPath?

Cheers,
Dimitre

On Sun, Sep 8, 2019 at 5:54 PM Liam R. E. Quin 
liam(_at_)fromoldbooks(_dot_)org <mailto:liam(_at_)fromoldbooks(_dot_)org> 
<xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com 
<mailto:xsl-list-service(_at_)lists(_dot_)mulberrytech(_dot_)com>> wrote:
On Mon, 2019-09-09 at 00:18 +0000, Dimitre Novatchev
dnovatchev(_at_)gmail(_dot_)com <mailto:dnovatchev(_at_)gmail(_dot_)com> 
wrote:
The W3C F&O 3.1 spec (at
https://www.w3.org/TR/xpath-functions-31/#func-filter 
<https://www.w3.org/TR/xpath-functions-31/#func-filter> ) says:

Rules

The effect of the function is equivalent to the following 
[...]

Because "equivalent" means the two functions must produce the same
result
for for all possible values in the same set of arguments, 

That is one possible definition of "equivalent" but it is not the one
used in the Functions and Operators document...

You can never guarantee that two expressions are equivalent in your
sense, because of "errors and optimization". Any construct might raise
an error - in the case of this example, stack overflow if the recursion
gets too deep.

Liam

-- 
Liam Quin, https://www.delightfulcomputing.com/ 
<https://www.delightfulcomputing.com/>
Available for XML/Document/Information Architecture/XSLT/
XSL/XQuery/Web/Text Processing/A11Y training, work & consulting.
Carefoot Web-slave for historical images http://www.fromoldbooks.org/ 
<http://www.fromoldbooks.org/>


XSL-List info and archive <http://www.mulberrytech.com/xsl/xsl-list>
EasyUnsubscribe <http://lists.mulberrytech.com/unsub/xsl-list/293509> (by 
email <>)
--~----------------------------------------------------------------
XSL-List info and archive: http://www.mulberrytech.com/xsl/xsl-list
EasyUnsubscribe: http://lists.mulberrytech.com/unsub/xsl-list/1167547
or by email: xsl-list-unsub(_at_)lists(_dot_)mulberrytech(_dot_)com
--~--
<Prev in Thread] Current Thread [Next in Thread>