Performance of various xpaths
|
View:
New views
7 Messages
—
Rating Filter:
Alert me
|
 |
Performance of various xpaths
by robertsmme
::
Rate this Message:
Hi,
I am applying
a great number of xpaths to a very large document. The results are
slooooow.
Virtually all
the xpaths are of the form //ns:parent/ns1:child where child could be an
attribute.
I understand
the reason for the performance hit is that the above xpath apparently requires a
full document scan.
With saxon
xpath is this the case. Would I get better performance using Saxon 9 over
jaxen 1.1.1? Is there a big hit because the underlying document is JDOM
structure?
Thanks
Martin
------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
|
|
Re: Performance of various xpaths
by Michael Kay
::
Rate this Message:
It's not clear from your question whether the slow
performance you are reporting is coming from Saxon or from
Jaxen.
Saxon has an optimization for XPath expressions starting
//x - the first time such an expression is used for a given element name x, it
scans the document to find all the x elements, and on subsequent requests, this
index is reused. However, this only works for native Saxon tree models (Tiny
Tree and Linked Tree), it doesn't work for external models like DOM or JDOM
(though actually, it would be easy enough to implement).
For this and other reasons, Saxon will usually perform
quite a bit faster on its native tree implementation than on third-party tree
models.
If you have predicates in the path expressions, there are
potential further gains from the join optimizer in Saxon-SA. However, if you're
executing individual path expressions from a Java application there's a limit to
what Saxon can achieve because it can't see the whole picture at compile
time.
Michael Kay
------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
|
|
Re: Performance of various xpaths
by robertsmme
::
Rate this Message:
Michael,
Thank you for your response.
Do you have any API that work the other way round -
i.e. given a node in a document you can tell if an XPATH matches the
node?
Martin From: saxon-help-bounces@... [mailto:saxon-help-bounces@...] On Behalf Of Michael Kay Sent: 11 July 2008 11:24 To: 'Mailing list for the SAXON XSLT and XQuery processor' Subject: Re: [saxon] Performance of various xpaths It's not clear from your question whether the slow
performance you are reporting is coming from Saxon or from
Jaxen.
Saxon has an optimization for XPath expressions starting
//x - the first time such an expression is used for a given element name x, it
scans the document to find all the x elements, and on subsequent requests, this
index is reused. However, this only works for native Saxon tree models (Tiny
Tree and Linked Tree), it doesn't work for external models like DOM or JDOM
(though actually, it would be easy enough to implement).
For this and other reasons, Saxon will usually perform
quite a bit faster on its native tree implementation than on third-party tree
models.
If you have predicates in the path expressions, there are
potential further gains from the join optimizer in Saxon-SA. However, if you're
executing individual path expressions from a Java application there's a limit to
what Saxon can achieve because it can't see the whole picture at compile
time.
Michael Kay
------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
|
|
Re: Performance of various xpaths
by Michael Kay
::
Rate this Message:
If
the XPath expression takes the form of an XSLT pattern, then you can compile it
as a pattern and test it against a node. In 9.1 there's a convenient API for
this, for the first time: XPathCompiler has a method compilePattern(), which
constructs a pseudo-expression whose effective boolean value is true when
the context item matches the
pattern.
Michael Kay
------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
|
|
Re: Performance of various xpaths
by robertsmme
::
Rate this Message:
Michael,
That looks promising, what do you mean by "the XPath
expression takes the form of an XSLT pattern". What is an XSLT
Pattern?
Can you use the xpath //parent/child in your reply
please?
Martin From: saxon-help-bounces@... [mailto:saxon-help-bounces@...] On Behalf Of Michael Kay Sent: 11 July 2008 11:46 To: 'Mailing list for the SAXON XSLT and XQuery processor' Subject: Re: [saxon] Performance of various xpaths If
the XPath expression takes the form of an XSLT pattern, then you can compile it
as a pattern and test it against a node. In 9.1 there's a convenient API for
this, for the first time: XPathCompiler has a method compilePattern(), which
constructs a pseudo-expression whose effective boolean value is true when
the context item matches the
pattern.
Michael Kay
------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
|
|
Re: Performance of various xpaths
by Michael Kay
::
Rate this Message:
An XSLT pattern is anything you can write in the match
attribute of xsl:template. It's a subset of XPath defined at
Simple paths like //parent/child will work fine, but there
are many XPath expressions that are not valid patterns, for example "2+2", or
"count(//parent/child)", or "following-sibling::x".
Michael Kay
------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
|
|
Re: Performance of various xpaths
by andrew welch
::
Rate this Message:
2008/7/11 <martin.me.roberts@...>:
> Michael, > That looks promising, what do you mean by "the XPath expression takes the > form of an XSLT pattern". What is an XSLT Pattern? > > Can you use the xpath //parent/child in your reply please? The thing is... don't you want all sorts of different //parent/child combinations? As in: //foo/bar //foo/baz //other/else etc. So, if Saxon builds an index for each use of //, unless you use the very same combination of parent and child you won't see much of a benefit (or perhaps just the "parent" is indexed, so for xpaths with same parent you would see the benefit?) Anyway, in XSLT you could define a key that indexes the parent and child names: <xsl:key name="parent-and-child" match="*" use="concat(name(parent::*), '-', name(.))"/> and then do: key('parent-and-child', 'ns:parent-ns2:child') instead of //ns:parent::ns2:child to get the nodes your after - once the key is built the access time should be constant. -- Andrew Welch http://andrewjwelch.com Kernow: http://kernowforsaxon.sf.net/ ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08 _______________________________________________ saxon-help mailing list archived at http://saxon.markmail.org/ saxon-help@... https://lists.sourceforge.net/lists/listinfo/saxon-help |
| Free Forum Powered by Nabble | Forum Help |