W3C home > Mailing lists > Public > www-ql@w3.org > January to March 2001

RE: [www-ql] <none>

From: Ingo Macherius <macherius@darmstadt.gmd.de>
Date: Fri, 2 Mar 2001 17:35:37 +0100
To: <www-ql@w3.org>
Message-ID: <NDBBKNAAEKOBBDPKKDDCOEKEDIAA.macherius@darmstadt.gmd.de>
Jonathan,

> > Thus semantic
> > identity MUST be always given when transforming FLWR-XQuery to XML Query
> > Algebra expressions and from there to another syntax. There can
> > not be more power in XQuery or any other syntax then there is in the
> > Algebra. I hope I got this right ?

> I am not sure I understand what you mean by FLWR-XQuery. Do you mean
> XQuery? There is only one language named XQuery, and FLWR is just one of
> the ten or so expression types in XQuery.

I can not agree there is just one language named XQuery, formerly known as
"W3C XML Query". However, there is just one proposed syntax for it, yet,
which I refer to as FLWR-XQuery. The other syntax will be at least
XML-XQuery, while others like XSLT-XQuery or XPath-Xquery are worth while
thinking of.

Thus I do not agree we can limit the term XQuery to a specific syntax
suggestion.

> I am also not sure that you and I mean the same thing by the phrases
> "semantic sugar" and "semantic identity". If I convert an XQuery
> expression
> into the algebra, I may not be able to convert back from the
> algebra to the
> original XQuery expression. If I convert an XQuery expression

This raises orthogonality questions. If certain constructs derived by
building an algebra tree parsing FLWR-XQuery have an ambigous mapping back
into FLWR-XQuery, the question is whether FLWR-XQuery is orthogonal enough.
E.g. currently a fair amount of what can be done in WHERE can also be
epxressed as an XPath' filters(or whatever you call the path expression in
FLWR-XQuery). Another issue is AND, OR etc. which directly (even
syntactically) collide with existing XPath keywords.

I understand there currently can not be a literal roundtrip
FLWR->Algebra->FLWR. But unless this becomes feasible, FLWR-XQuery must not
be considered orthogonal enough. This may require some heuristics to
disambiguate the rewrite to FLWR syntax.

> into the XML
> representation of the parse tree, the information content in the two
> syntaxes is identical.

As explained above, a true XML syntax will hopefully dissolve
XPath-Expressions, FLWR-expression and and XSLT-ish RETURN sections into
something which closely resembles Algebra expressions. I have to second
Michael Rys here (even if I'm not so sure if he wasn't joking when writing
his mail regarding this issue). For me this is what I expect from the XML
syntax. Thus a roundtrip through the parse tree, or the XML syntax, should
loose any syntactic dependencies from FLWR-Xquery. Just as relational
algebra is independed of SQL syntax.

Second, I think it's a bit vague to talk of "information content of
syntaxes". We can only compare equality of query expressions on the
operational semantic level, not in terms of parse trees similarities. This
is what query optimization by equality and query rewrite for physical models
is all about.

Regards,
	Ingo Macherius
Received on Friday, 2 March 2001 11:33:54 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 22 July 2006 00:10:17 GMT