Re: [xsl] Unravelling Imports

Subject: Re: [xsl] Unravelling Imports
From: Michael Kay <mike@xxxxxxxxxxxx>
Date: Thu, 12 Apr 2012 10:38:58 +0100
I think Scott has identified the two restrictions that will be very hard to get around: (a) apply-imports, and (b) dependencies on the base URI of stylesheet modules (for example, use of document('')). But it's probably true that 95% of stylesheets are unaffected by those restrictions.

Combining system-allocated precedence values with user-allocated priorities is possible, but it requires parsing of match patterns to determine the default priority if no user-defined priority has been set. Dealing with union patterns is particularly tricky.

For named templates, global variables, etc, it's a question of recognizing the duplicates and dropping all but the last.

There's a lot of messy detail for edge stuff such as xsl:output and xsl:decimal-format: it's a classic case where you can get a 90% solution for 10% of the effort of a 99% solution.

Michael Kay
Saxonica

On 11/04/2012 20:55, Scott Trenda wrote:
Ihe,

As much as it sounds like a bad idea on the surface, I can understand where you could be coming from. For example, if you have an intricate client-side XSLT architecture with highly modularized stylesheets, then the overhead cost of HTTP requests to download/cache-verify the individual stylesheets could start to add up to a non-negligible overhead, especially on a bad/slow connection.

I thought I was in a similar situation a few years ago, and at the time, I actually mapped out what the transformation would look like and how it would work. I never got around to implementing it due to the potential risks, but if you have a very compelling reason for it, I do know it's possible, at least with XSLT 1.0. It involves a lot of mapping out the exact priorities, applying and copying, sorted by the computed priorities (for<xsl:import>s), and copying only the topmost elements in the priority list to avoid duplicate definitions. If you're only using<xsl:include>s, it should be easier, because your XSLT processor would throw stylesheet errors if there were any duplicate definitions, due to the single-level nature of includes.

There are some specific functionalities that may not be available in a monolithic model (<xsl:apply-imports/>, for one), and others for which I'm honestly not certain of the behavior of XSLT 1.0 processors (e.g. the scope of document('') in an imported or included stylesheet). But I do know it's possible and feasible within a limited scope.

~ Scott


-----Original Message----- From: ihe onwuka [mailto:ihe.onwuka@xxxxxxxxxxxxxx] Sent: Wednesday, April 11, 2012 2:36 PM To: xsl-list Subject: [xsl] Unravelling Imports

There is a supposed requirement that we deliver monolithic XSLT with includes unravelled (there are no imports).

I don't need to be convinced it's a bad and misguided idea but I won't close off anyone who wants to comment as they may think of stuff that I haven't.

The purpose of my post though is this.

Suppose we did go ahead and code with imports, would it be possible to define a transform that mechanically transformed everything into monolithic XML.

I'm thinking it would involve alot of use of priorities at the very least.

Ihe

Current Thread