Am Di., 21. Jan. 2020 um 23:13 Uhr schrieb David Kastrup <d...@gnu.org>:
>
> Thomas Morley <thomasmorle...@gmail.com> writes:
> > David, you remember my suggestion to generate that "General Code
> > Reference" with an Index ... ?
>
> Yes, that may have helped. But it may also have delivered a haystack.
>
> I think we probably need a "good programming" corpus, but the snippets
> are supposed to be that. Navigating them still is too hard.
If you think of the "Snippets Manual" or the LSR, then I doubt they
are best suited to what I think we need.
Their snippets are too much targeted at delivering ready to use tools
for actual typesetting work.
As an example:
music-map has the problem that you can't really stop it recursing
deeper. One reason why you wrote map-some-music and for-some-music.
Alas, I always need to get to their doc-strings again, to understand
the differences and which one to use for which case.
And ofcourse their docstrings are not in any manual, afaict.
Sometimes I look in our regression-tests, which contains nice coding
examples, but you can't expect average user to look there.
And sometimes even the regression-test don't say anything, as an
example for no info at all:
$ git grep "ly:broadcast"
lily/dispatcher-scheme.cc:LY_DEFINE (ly_broadcast, "ly:broadcast",
scm/define-music-callbacks.scm: (ly:broadcast
(ly:context-event-source context)
Leading to IR entry:
Function: ly:broadcast disp ev
Send the stream event ev to the dispatcher disp.
Which is ununderstandable without any context.
Cheers,
Harm
