Re: [xsl] Does the new structure include the same text content?

["Imsieke, Gerrit, le-tex gerrit.imsieke@xxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>]

Yes, it is often hard to go full round-trip.

What sometimes works better is this: If HTML rendering stylesheets exist 
for both formats, to tweak both stylesheets so that, with a little or 
large bit of normalization, identical textual outputs emerge.

I don't know enough about S1000D to know whether this approach is 
feasible though.

Gerrit

On 22.01.2021 16:58, ian.proudfoot@xxxxxxxxxxx wrote:
> Hi Gerrit,
> Good to know that I may be on the right track with the normalized text diff.
> It would be almost impossible to go back to the original SGML structure from the XML. The main difficulty is that a lot of the structure in the SGML uses inclusions to allow tables and figures in almost any location. That SGML feature was always a recipe for untidy documents!
>
> Ian
>
> -----Original Message-----
> From: Imsieke, Gerrit, le-tex gerrit.imsieke@xxxxxxxxx <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>
> Sent: 22 January 2021 11:45
> To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> Subject: Re: [xsl] Does the new structure include the same text content?
>
> Hi Ian,
>
> diffing normalized text output is a good approach in my experience.
> However, if the 4.1 structures differ significantly from 1.7 as you say, it might be a good idea to transform the 4.1 output back to 1.7 prior to the diff. Or maybe not "transform it back to match the input exactly", but only to such a degree that the text files will be the same if no content was lost or duplicated.
>
> Gerrit
>
> On 22.01.2021 12:28, ian.proudfoot@xxxxxxxxxxx wrote:
> > Hi everyone,
> >
> > I am working on a project to convert several thousand SGML files
> > (S1000D
> > 1.7) into a more recent XML version (S1000D 4.1). My finished XSLT
> > style sheet does the job that is expected.  However during the
> > development I did run into a problem where an error in the stylesheet
> > allowed the output to pass schema validation but by omitting some
> > content! For me thatbs very bad news and I was lucky to notice it.
> > Ultimately the final output will be verified by the subject matter
> > experts, but I really donbt want to give them any reason to doubt the
> > reliability of the conversion.
> >
> > This got me thinking about ways to verify the output text content
> > against the input despite significantly different structure. Is there
> > an established way to do that? If so what is it called and how well
> > does it work?
> >
> > Perhaps itbs something that I should build into the XSLT as it is
> > written? Or perhaps it could be run as a post process batch comparison
> > operation?
> >
> > My initial thought is to output normalized text from input and output
> > and compare the resulting text filesb&
> >
> > Ibve searched the archives, but I probably donbt know the correct
> > terminology to get any useful resultsb&
> >
> > Thanks in advance for all responses.
> >
> > Ian
> >
> > Ian Proudfoot
> >
> > Bembridge
> >
> > Isle of Wight
> >
> > United Kingdom