Hi Manuel,
Jim Hefferon has specified keywords and categorizations now for all CTAN
packages (!). So I was thinking that texdoc could acquire an --apropros
option, which returns results based on keywords, in addition to the
usual package and doc names. Probably searching both Jim's list and the
one-line summaries would be best. Maybe even search descriptions, too.
Jim's data is an enhanced version of the Catalogue, in XML, dumped
nightly. ftp://tug.ctan.org/ftpmaint/az/texcatalogue.xml
Of course we wouldn't just dump in the whole thing, we'd have to write
something to extract just the keywords in a form that is good for us,
and transform package names (catalogue -> tl) where needed.
That shouldn't be hard.
I don't think Jim's characterizations, nice as they are, are directly
relevant for texdoc, since texdoc is about displaying documentation, not
browsing directory trees. You can view it all online at
http://az.ctan.org/ (Jim's test site), though.
Clearly this is future work, nothing to be done quickly or before the
release. (It came up at the conference.) I'm sending it now just so
it gets off my list and on to yours :).
Thanks,
k
)
Le 12/07/2010 23:40, Karl Berry a écrit :
> Jim Hefferon has specified keywords and categorizations now for all CTAN
> packages (!).
Impressive indeed. Is the categorization related to the "bytopic" page of the
catalogue? http://texcatalogue.sarovar.org/bytopic.html
> So I was thinking that texdoc could acquire an --apropros
> option, which returns results based on keywords, in addition to the
> usual package and doc names. Probably searching both Jim's list and the
> one-line summaries would be best. Maybe even search descriptions, too.
>
Yep, I had such a project very ****uely on my longer-term list (searching the
description). No doubt keywords and catergories will make it more effective.
(****ue projects I have about texdoc include making a GUI, which would also allow
to browse (as opposed to search) by category. I realise it looks very similar to
Jim's project for the future Ctan search interface, looking at your link below.)
> Jim's data is an enhanced version of the Catalogue, in XML, dumped
> nightly. ftp://tug.ctan.org/ftpmaint/az/texcatalogue.xml
> Of course we wouldn't just dump in the whole thing, we'd have to write
> something to extract just the keywords in a form that is good for us,
> and transform package names (catalogue -> tl) where needed.
> That shouldn't be hard.
>
Sounds good. Does "we" mean "TL"? I mean, would the new information end up in
texlive.tlpdb or would the extraction tool be specific to texdoc?
> I don't think Jim's characterizations, nice as they are, are directly
> relevant for texdoc, since texdoc is about displaying documentation, not
> browsing directory trees. You can view it all online at
> http://az.ctan.org/ (Jim's test site), though.
>
Thanks for the link. I'll see in due time if the categories can be usefully used
as keywords too.
> Clearly this is future work, nothing to be done quickly or before the
> release. (It came up at the conference.) I'm sending it now just so
> it gets off my list and on to yours :).
>
Thanks.
Manuel.
)
Is the categorization related to the "bytopic" page of the
catalogue? http://texcatalogue.sarovar.org/bytopic.html
Not related. His keywords and characterizations are in addition to the
"bytopic" tree. He has another tree using that (Fenn's topics):
http://az.ctan.org/characterization/choose_dimen/
(****ue projects I have about texdoc include making a GUI,
Keyword searching sounds a heck of lot easier than making a GUI :).
Does "we" mean "TL"?
Well, I don't mind working on the extraction script if need be.
I mean, would the new information end up in
texlive.tlpdb or would the extraction tool be specific to texdoc?
I was imagining that the keywords would be in a separate file part of
the texdoc package, not tlpdb, because nothing but texdoc would be
reading them, so why take up the space/time for everyone ...
thanks,
k
)
Le 14/07/2010 19:56, Karl Berry a écrit :
> Is the categorization related to the "bytopic" page of the
> catalogue? http://texcatalogue.sarovar.org/bytopic.html
>
> Not related. His keywords and characterizations are in addition to the
> "bytopic" tree. He has another tree using that (Fenn's topics):
> http://az.ctan.org/characterization/choose_dimen/
>
Ok, I see.
> Keyword searching sounds a heck of lot easier than making a GUI :).
>
Sure :-)
> I was imagining that the keywords would be in a separate file part of
> the texdoc package, not tlpdb, because nothing but texdoc would be
> reading them, so why take up the space/time for everyone ...
>
Right. I was asking because one could imagine adding keyword search to tlmgr
too, in which case it would be better to put the info in the tlpdb.
But a separate file shipped with texdoc and generated by a texdoc-specific
script is perfectly fine with me.
Manuel.
)
