Re: [xsl] Best practices when reorganizing content?

["Michael Kay michaelkay90@xxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>]

I've been studying this problem for a while now, and I think I have some of
the answers but not all. XSLT 3.0 is certainly not especially good at
processing JSON in the "recursive descent" style using template rules. One of
the problems, as you say, is the absence of element names; another is the
absence of parent pointers. We're incrementally adding features to the 4.0
spec to address the problems but I'm conscious that we haven't yet joined all
the dots. You may find it interesting to read
https://github.com/qt4cg/qtspecs/issues/1786

I would be interested in the detail of what JSON you are processing and what
transformation you want to apply to it. We need this kind of input on use
cases to come up with the right feature set for the 4.0 spec.

Meanwhile, my Balisage 2022 paper gives a snapshot of my thinking at the
time:

https://www.balisage.net/Proceedings/vol27/html/Kay01/BalisageVol27-Kay01.htm
l


and draws on a couple of use cases of JSON transformations with 3.0 reported
at XML Prague 2016 -- see
https://www.saxonica.com/papers/xmlprague-2016mhk.pdf -- where sadly the
conclusion was that it was often easier to convert the JSON to XML, do the
transformation, and then convert it back.

Michael Kay
Saxonica

> On 21 Mar 2025, at 20:37, dvint@xxxxxxxxx
<xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx> wrote:
>
> My general stylesheet design is to try and leverage templates that match
things and produce output as they are found in the document order. In some
cases the templates will process the content and other times it will
ignore/hide the content. If I'm worried the format will evolve I'll try and
add a wildcard template <xsl:template match="*"> and report elements that
weren't in my original source to trigger a review oof the output.
>
> That works well for me with something that is generally in the same order of
output and content is nested nicely that I might ignore whole sections.
>
> I'm now trying to process JSON that was developed for a completly different
purpose and the content has to be transformed in a very specific order. Also
because it is coming from JSON i don't have nice element names to reference.
>
> Currently I start a template and then try and target the specific sections
with if statements to make the decision to output or not - more procedural.
>
> I run the risk of missing additions to the content that I might want to do
something with and I don't have a good way to try and report these additions
and potential deletions.
>
> I've got a fair amount of content being managed and it has given me an idea
of what is going on in the content that I didn't have when I started, such
that I'm going to see if I can try a different approach.
>
> The other issues is that the content doesn't have any sort of
model/scheam/dtd that is used to validate the results and it looks like anyone
can add any structure (or not follow existing patterns) when ever they want.
This structure has morphed over the last 2yrs of use so I have early content
designs with more recent designs.
>
> What I'm trying to achieve:
> - stylesheet that is less procedural
> - reporting of new structures that I might want to look into
> - error reporting when the content I receive doesn't match "my model view"
of what was intended
>
> Any suggestions greaty appreciated.
>
> ..dan
>
>

o?<