[xsl] RE : [xsl] recursivity and param

["Xavier Boully" <xavier.boully@xxxxxxxxxxxxxxxx>]

My example was too simple to
represent my need.
But it doesn't matter as you
gave a good answer to the
general question that was the
real need.

You are right, going to the
first child and the next
sibling is the perfect way to
scan the whole tree with the
best control on nodes.

As I briefly think of this, we
get a better symmetry between
the width and depth of a tree
than with the common way of
taking all the children.
In an other context than XSLT,
this method oblige to think of
2 stopping conditions, one for
the depth direction and one
for the width direction. I am
particularly happy with this
symmetry as it is easier then
to separate the tree scan
process from the calculation
needed for a problem whatever
it is.

Thanks Dimitre

Xavier

-----Message d'origine-----
De : Dimtre Novatchev
[mailto:dnovatchev@xxxxxxxxx]
Envoyi : jeudi 12 ao{t 2004
13:53
@ :
xsl-list@xxxxxxxxxxxxxxxxxxxxx
m
Objet : Re: [xsl] recursivity
and param


As other people have pointed
out, the specific problem you
provide does not require
recursive processing at all --
a simple XPath expression
suffices.

The answer to your more
general question (in case I
understand it
correctly) is provided by the
Tree Visitor pattern, which is
a generic pattern of
transformation by visiting
every node in a sequential
(document) order. This
provides for the finest level
of granularity possible with
XSLT. An obvious shortcoming
may be the very deep recursive
processing, however there are
proven ways to solve this
problem -- e.g. the DVC
(Divide and Conquer) approach.

As is the case with other
patterns, because of its poer
and generality, it should be
used only when there are no
other, more direct and
efficient ways to solve a
given problem.

For detailed code see:

http://www.biglist.com/lists/x
sl-list/archives/200402/msg008
66.html


Cheers,

Dimitre Novatchev.


On Thu, 12 Aug 2004 12:32:35
+0200, Xavier Boully
<xavier.boully@xxxxxxxxxxxxxxx
m> wrote:
> I would like to use
recursivity through the source
document with XSLT
> 1.0 and no extension.
>
> As far as I understood, when
I am at a node level, I have a
context
> that I can pass to children
level using with-param. But
what I would
> like to obtain is to catch
things from children treatment
to influence
> the context of the node
level: when I pass context to
a certain child
> I would like to take in
account things that happened
before while
> processing children I met
before (in order of source
document for
> example).
>
> Reading again that question
make me think that it is maybe
not clear
> enough. Let's try an
example.
> - Suppose that in source
tree we have some
<to-copy-element />
> elements
> - Suppose that I just want
my XSLTransformation to copy
to the result tree
> the <to-copy-element /> from
the source tree.
> - Suppose that I add a
condition that I must not copy
more than 10
> <to-copy-element />. In
other words I want to copy
only the first (in order
> of source document) 10
<to-copy-element /> elements
or all the
> <to-copy-element /> elements
if
count(//to-copy-element)<11.
>
> I think (but this is maybe a
wrong hypothesis) that at a
given node
> level, I have to pass to a
child information about the
number of
> copies done with children
processed before.
>
> Any ideas ?
>
> Xavier