Re: A theory problem

[david.rosenborg@xxxxxxxxxxx]

Hi!

In our implementation of XSLT we use a simple set
of heuristics to avoid reordering:

Each subexpression is assigned zero or more of the
following properties:

SINGLETON : the expression can at most select a single node.

SIBLINGS: the expression will select nodes in document order that are siblings
of eachother.

DESCENDANTS-OR-SELF: the expression will only select the context node or
descendants
to it in document order.

Then we combine expressions in for example a path:

e1 / e2

The new expression that results  from this will have properties that are
calculated from the combination of the subexpressions. I don't have
time right know to list all the combinations but it isn't that hard to figure
them
out and I'll give an example:

If e1 have the property SIBLINGS and e2 have the property DESCENDANTS-OR-SELF
then the whole expression is guaranteed to be in document order.

So by assigning the properties to primitive expression and writing rules for
combinations we can avoid reordering in many situations.

</David>

_______________________________________________________________________
David.Rosenborg@xxxxxxxxxxxxxxxx



 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list