Re: [xsl] Optimizing preceding-sibling & following-sibling axes (with key()?)

["G. Ken Holman" <gkholman@xxxxxxxxxxxxxxxxxxxx>]

At 2010-04-07 20:56 +0200, Martynas Jusevicius wrote:
> I have such a variable definition (quite cumbersome):
>
> <xsl:variable name="descendant-headings"
> select="following-sibling::h:*[self::h:h1 or self::h:h2 or self::h:h3
> or self::h:h4 or self::h:h5 or self::h:h6][o2e:heading-level(.) &gt;
> o2e:heading-level(current())][preceding-sibling::h:*[self::h:h1 or
> self::h:h2 or self::h:h3 or self::h:h4 or self::h:h5 or
> self::h:h6][o2e:heading-level(.) = o2e:heading-level(current())][1] is
> current()]"/>
>
> Basically, I need to select all descendant (level-wise, not as in
> descendant axis) headings for the current() one, meaning all following
> headings with lower level which go before the next heading one the
> same level as current(). This expression gives me what I need.
> o2e:heading-level() function simply extracts the level number from
> elements name, like 'h3'.
>
> Now, I've read multiple times that preceding-sibling/following-sibling
> are slow. I wonder if this expression could be optimized?

I'm not sure why you are using that expression at all.  It is obvious 
because of the use of a user-defined function that you are running 
XSLT 2.0 ... your use-case is the exemplar I use in the classroom for:

  <xsl:for-each-group group-starting-with="...

... to infer depth from flatness.

> I also have a key definition which I use for similar purposes:
>
> <xsl:key name="heading-by-level" match="h:h1 | h:h2 | h:h3 | h:h4 |
> h:h5 | h:h6" use="number(substring-after(local-name(), 'h'))"/>
>
> I just can't figure out how combine this key with the sibling axes to
> get the same result and improve performance, i.e. that instead of
> traversing the whole document it would be enough to traverse key()
> values, like in key('heading-by-level', 1 to 6).
> Does it make sense?

Maybe for XSLT 1.0, but not that way.  I would simply use the 
preceding axis finding the closest heading and put the generated id 
of that heading as the lookup value.  Then when walking through the 
headings I would find the content for that heading from the key table 
instead of from the axes.  It was much faster than trying to use the axes.

But I think you just want to use the XSLT 2.0 grouping approach if 
you are trying to infer something like DocBook (with nested sections) 
from XHTML (with sibling sections).

I hope this helps.

. . . . . . . . . . Ken


--
XSLT/XQuery training:         San Carlos, California 2010-04-26/30
Principles of XSLT for XQuery Writers: San Francisco,CA 2010-05-03
XSLT/XQuery training:                 Ottawa, Canada 2010-05-10/14
XSLT/XQuery/UBL/Code List training: Trondheim,Norway 2010-06-02/11
Vote for your XML training:   http://www.CraneSoftwrights.com/s/i/
Crane Softwrights Ltd.          http://www.CraneSoftwrights.com/s/
G. Ken Holman                 mailto:gkholman@xxxxxxxxxxxxxxxxxxxx
Male Cancer Awareness Nov'07  http://www.CraneSoftwrights.com/s/bc
Legal business disclaimers:  http://www.CraneSoftwrights.com/legal