Re: lybook-db etc etc.

[David Kastrup]

Graham Percival <address@hidden> writes:

> On Thu, Oct 27, 2011 at 07:25:19AM +0200, David Kastrup wrote:
>> Graham Percival <address@hidden> writes:
>> 
>> > On Wed, Oct 26, 2011 at 05:13:56PM +0200, David Kastrup wrote:
>> >> 
>> >> In my opinion, the whole lybook-db stuff needs to go.  Instead, Lilypond
>> >> is run _once_ for all snippets of a lybook source, generating _one_
>> >> PostScript file.
>> >
>> > ... so instead of only generating snippets it needs, you want to
>> > generate a full set of snippets for each language, thereby making
>> > "make doc" take roughly 5 times as long as it currently does?
>> 
>> Wrong.  Thereby making "make doc" do about 5 times the real work as it
>> currently does, taking a fraction of the time.
>
> Really?  if we have exactly the same @lilypond[] in
> es/learning.tely and fr/learning.tely, how do you avoid
> recompiling that @lilypond[] if you're running it on a
> .itely_file-by-.itely_file basis?

I don't.  I already _said_ that it will do 5 times the real work.  But
about 1% of the thrashing and reinitialization.  And that is what is
eating up 95% of the time (the vast majority of snippets is trivial and
totally dominated by reinitializing).  And five times 1% is still just
5%.

> Let me clarify that: "our development team is in a complete mess".
>
> but whatever.  go ahead and write some patches.

That would not be "some patches".  It would be about two months of work.

> as long as they don't break stuff, i'm fine with pushing them.

The work _will_ break stuff.  For example, Lilypond will need to tell
GhostScript bounding boxes for generating snippet images.  For included
PostScript code, it does not know them.

> if somebody wants to complain, they can do so right now, or during the
> review phase.

Cough, cough.  We are not at a point of time where complaining makes the
least bit of sense.

> unless john or jan speaks up in the next 24 hours, there's a good
> chance that they'll reappear in 3 weeks asking "why is xyz broken",
> and then we can have another fun round of hating each other for the
> broken situation.

There will be no code in a state fit to complain about in 3 weeks.

> to emphasize: i don't know how the current system works, i don't
> care, and i'm taking no blame if any work on this is rejected or
> reverted in x days or weeks.  we are collectively horrible at
> voicing objections to build system changes ahead of time, so any
> work is done at your own risk.

It has nothing to do with the build system at all.  This concerns how
Lilypond, in particular lilypond-book, goes about batching its work.

If lilypond-book finishes its work on typical large documents in 5% of
the previous time without touching the lybookdb database, I have little
doubt that somebody _will_ do the job of weeding out all the complex
build-system stuff trying to meddle with the internal operation of
lilypond-book.

-- 
David Kastrup