Re: [xsl] Merging two sets of files

[Emmanuel Bégué <medusis@xxxxxxxxx>]

Do I understand your requirements correctly -- you need to output
- a new version of book.xml with associated catalog information from
the drug database
- standalone documents from book.xml for topics that don't have
corresponding drug info
- a new version of each of the 10,000 for which information can be
found in book.xml (not sure about that last part -- requirements 3 and
4?)

I would build two temporary documents:
- a new book.xml with all associated catalog information
- a new big catalog file (from all the relevant little drug files),
with all associated book information

and then in a second pass, cut those big documents to output the
required result files as needed.

One way to do two passes with Saxon is to use saxon:next-in-chain
(variables are really painful to use).

Hope this helps.
Regards,
EB


2012/4/3 Emma Burrows <Emma.Burrows@xxxxxxxxxxx>:
> I'm currently using XSLT 2.0 (using Saxon 9.3 via Oxygen 12) to merge two
sets of XML files together based on a third file which is a kind of lookup
table. However, I'm coming across a problem when I need to effectively merge
two source files into the same output file, and I need some suggestions on a
change of approach.
>
> I have the following XML files:
>
> - Main document - let's call it book.xml
> This contains various types of topics, including about 4000 topics related
to drugs, each identified by a unique id.
>
> - Ancillary drug information files auto-generated from an online drug
database.
> These are about 10,000 little XML files, each named after the unique id of
the drug information in the online catalogue.
>
> - An XML file - let's call it lookup.xml - that is essentially a look-up
table, matching ids in book.xml to one or more drug catalogue ids, and vice
versa. However, not all records in book.xml have an entry in lookup.xml.
>
> Now my requirement is to convert book.xml from its current proprietary
format into a DITA-based specialisation, and while I'm doing that:
>
> 1- Output the records with no corresponding catalogue entry as standalone
documents.
>
> 2- Merge each drug record in book.xml that has catalogue entries with the
corresponding auto-generated catalogue file(s), based on lookup.xml.
>
> 3- If a record in book.xml has more than one catalogue id in lookup.xml, I
need to copy the book.xml record into every one of the corresponding
auto-generated files.
>
> 4- If more than one record in book.xml corresponds to one catalogue id in
lookup.xml, I need to merge all the book.xml records with that same catalogue
file.
>
> 5- Make sure the converted and merged files are referenced in the correct
location in the book's hierarchy.
>
> I expect we'll ultimately do something more sensible like use conref rather
than tamper with the auto-generated files, but merging them is my current
brief as it stands.
>
> Point 4 is the immediate stumbling block because my solution to fulfilling
points 2 and 3 was as follows:
>
> 1. Convert the book.xml drug record into the desired DITA format and place
that in a variable.
> I'm doing this based on a matched template, so this happens whenever the
processor "encounters" a drug record as it travels book.xml. This ensures that
I can export records with no catalogue id and keep track of where the record
was in the hierarchy.
>
> 2. Use the lookup.xml file to find corresponding catalogue ids for that
record.
>
> 3. For each catalogue id, open the corresponding catalogue file using
document(), and result-document it to a new file with the contents of the
variable inserted in the XML.
>
> The problem is that in step 3, I can't reopen a document that was previously
created by the transform, so I can't "add" a new book.xml record to the
contents of an already generated catalogue file, even by outputting a new file
with a different name.
>
> I can see that I'll probably need a process with an intermediate step,
perhaps using lookup.xml to guide the processing so I can group records with
the same catalogue id. But the only trouble with that is what to do with
records that don't appear in lookup.xml...
>
> Anyway, I hope all this is clear and I'm open to ideas. :)