- From: Peter Fankhauser <fankhaus@ipsi.fhg.de>
- Date: Mon, 25 Feb 2002 11:48:30 +0100
- To: "'Chen Ya Bing'" <iscp0054@nus.edu.sg>
- Cc: <www-ql@w3.org>
- Message-ID: <000301c1bde9$f771e230$ad240c8d@pcpotato>
One should be careful with statements like "SQL is obviously more
powerful than XQuery".
In your example the statement is downright wrong.
The SQL query you give operates on a different schema;
select jno, sno, pno, price
from spj
where price > 100
The equivalent in XQuery will be sth. like
for $j in document("spj1.xml")/spj
where $j/price > 100
return
<spj>{$j/jno, $j/sno, $j/pno, $j/price}</spj>
This isn't any more complicated than SQL, is it? But of course, it
operates
on and returns a flat table.
You appear to want to have a powerful form of groupBy which
(re-)introduces
nesting. Sth. like:
for $j in document("spj1.xml")/spj
where $j/price > 100
return
<spj>{$j/jno, $j/sno, $j/pno, $j/price}</spj>
groupby jno, sno, pno
Such an extension does not enhance the expressive power of XQuery, but
it is
certainly useful (providing an important hint for query optimization,
especially when
implementing XQuery on top of a relational engine). Nothing in XQuery's
design
prevents the introduction of such a feature, but for XQuery 1.0 it may
be prudent
to not fully commit to creeping featurism.
If you're worried about the multiple predicates on price in the solution
I originally suggested,
try the following:
for $j in document("spj.xml")//project
let $s1 :=
(for $s2 in $j/supplier
let $p := $s/part[price > 100]
where not(empty($p))
return <supplier sno="{$s2/@sno <mailto:sno={$s2/@sno> }">
{$p}
</supplier>)
where not(empty($s1))
return <project jno ="{$j/@jno}">
{$s1}
</project>
Best regards,
Peter
-----Original Message-----
From: www-ql-request@w3.org [mailto:www-ql-request@w3.org] On Behalf Of
Chen Ya Bing
Sent: Montag, 25. Februar 2002 10:25
To: TAN Kuan Hui
Cc: www-ql@w3.org
Subject: Re: define views using xquery
I think SQL is obviously more powerful than XQuery, although it cannot
replace XQuery because SQL is for relational database and XQuery is for
hierarchical database.
the SQL I give does construct the output I want. the schema I gave (
spj(jno, sno, pno, price) ) is based on the FD: jno, sno, pno --->
price, which is in 3NF.
in this case, the SQL is never rigged and the comparison is fair.
anyway, thank you for your comments.
----- Original Message -----
From: TAN Kuan Hui <mailto:kuanhui@mobileworkspace.com>
To: Chen Ya Bing <mailto:iscp0054@nus.edu.sg>
Cc: www-ql@w3.org
Sent: Monday, February 25, 2002 4:53 PM
Subject: Re: define views using xquery
Firstly, you have to be realistic that XQuery is still in draft stage.
Selectively evaluate XQuery is not objective evaluation.
Your SQL did not construct the output, strictly speaking it only does a
search which is at best equivalent to a path selection. What you were
looking for is significantly more in-depth. You were not making a fair
comparison. IMO, XQuery is significantly more powerful than
SQL if your SQL is fully rigged end-to-end.
You can't use algebra to express your view unless the engine
exposes that layer. The surface syntax is XQuery.
rgds.
----- Original Message -----
From: Chen Ya <mailto:iscp0054@nus.edu.sg> Bing
To: TAN Kuan Hui <mailto:kuanhui@mobileworkspace.com>
Sent: Monday, February 25, 2002 16:25 PM
Subject: Re: define views using xquery
since I don't find a system which support filter function, I did not
test the xquery.
the SQL equivalent depends on how we store the document into database.
the simplest one may be: spj(jno, sno, pno, price)
select jno, sno, pno, price
from spj
where price > 100
it is much easier than XQuery.
for the algebra, I mean if I use algebra to express the view, it may be
simplier.
----- Original Message -----
From: TAN Kuan Hui <mailto:kuanhui@mobileworkspace.com>
To: Chen Ya Bing <mailto:iscp0054@nus.edu.sg> ; Peter
<mailto:fankhaus@ipsi.fhg.de> Fankhauser
Cc: www-ql@w3.org
Sent: Monday, February 25, 2002 2:29 PM
Subject: Re: define views using xquery
LET $p := document("spj.xml")//part[price > 100]
RETURN filter($p/../.. | $p/.. | $p//*)
BTW, what is your SQL equivalent for the example
you have given ?
NB: you can't code at the algebra layer and the surface syntax
at the same time. I don't think the 2 grammars can co-exist without
ambiguities.
combination of XML-Schema and XQuery is not a problem.
If you require your view function to return a typed result,
encapsulate it in a function.
rgds.
----- Original Message -----
From: Chen Ya <mailto:iscp0054@nus.edu.sg> Bing
To: Peter Fankhauser <mailto:fankhaus@ipsi.fhg.de>
Cc: www-ql@w3.org
Sent: Monday, February 25, 2002 9:34 AM
Subject: Re: define views using xquery
one of the actual difficulties is like that:
there is a XML document (spj.xml):
<db>
<project jno="j001">
<supplier sno="s001">
<part pno="p001">
<price> 100</price>
</part>
</supplier>
<supplier sno="s002">
<part pno="p001">
<price> 110</price>
</part>
</supplier>
</project>
</db>
we want to define a view which selects such project, supplier and part
that their sub element price is greater than 100. the view definition
may be:
define view test as
For $j In document("spj.xml")//project[//price >100]
Return
<project jno = {$j/@jno}>
{ For $s In $j/supplier[//price >100]
Return
<supplier sno = {$s/@sno}>
{ For $p In $s/part[price > 100]
Return
<part pno = {$p/@pno}>
{$p/price}
</part>
}
</supplier>
}
</project>
we can see that for such a simple view, we need to repeat the predicate
"price >100" three times, which will lead to a very low performance of
query processing, because we need to search the tree three times. the
view definition is not only complex and wordy, but also low-performance.
Is there any other way to write the query?
as a view definition, one of favorite feature is that it is declarative,
so that we don't need to write a very complex definition.
can we just use a more simple way to write view, can XML Algebra provide
any help?
the view is just to put a predicate to filter some elements. if we
define a view that change some structures, the view definition will be
more complex.
until now, we do not consider the problem of view updatability and the
combination of XML-Schema and XQuery.
Best regards,
----- Original Message -----
From: Peter Fankhauser <mailto:fankhaus@ipsi.fhg.de>
To: 'Chen Ya <mailto:iscp0054@nus.edu.sg> Bing' ; www-ql@w3.org
Sent: Thursday, February 21, 2002 9:40 PM
Subject: RE: define views using xquery
As a starter from the scientific side you may check out
some recent papers by Serge Abiteboul and others (one of
Serge's favorite themes is views in all disguises).
I'd be also be interested in the actual difficulties that you
encountered when trying views with XQuery. In principle
one can regard a view as a "stored" query which can be
composed with other queries. As one of XQuery's virtues
is compositionality, another is closure, so I wonder where exactly
the difficulty lies.
(1) Is it the combination of XML-Schema and XQuery that
you (presumably) need.
(2) Is it updatability of views that concern you?
(3) Is it that your notion of "simple view" doesn't translate
to an XQuery notion of a "simple query"?
(4) Which sort of views do you want to define?
Have you got any concrete examples, usecases, requirements?
Where exactly do you encounter inefficiency?
Best regards,
Peter
-----Original Message-----
From: www-ql-request@w3.org [mailto:www-ql-request@w3.org] On Behalf Of
Chen Ya Bing
Sent: Donnerstag, 21. Februar 2002 10:02
To: www-ql@w3.org
Subject: define views using xquery
I'd like to use XQuery as a view definition language to define XML
views. but I find that it is difficult to use it as the view definition
language, because a simple view definition have to lead to a complex
XQuery. even more, it may be inefficient. comparing with view definition
in SQL, it is inconvenient.
can anybody give me advices of how to utilize XQuery to define XML view
more efficiently?
and how will the XQuery WD support the view facility.
Best regards,
Chen
Received on Monday, 25 February 2002 05:50:53 UTC