[Vim-latex-devel] (La)TeX compilation

[Luc H. <her...@fr...>]

* On Tue, Jul 15, 2003 at 12:13:51AM -0700, Srinath Avadhanula <sr...@fa...> wrote:
> On Tue, 15 Jul 2003, Luc Hermitte wrote:
> > - Why
> >     exec 'silent! 0r! '.g:Tex_CatCmd.' '.a:filename
> >   and not just:
> >     if fileexists(a:filename)
> >       exec 'silent! 0r '.a:filename
> >     endif
> >   ?
> 
> I tried doing :r {fname} before but found that it actually creates a
> new buffer in the process. I think that creating unused buffers is a
> greater evil than spawning a shell command. If you notice, I take a
> lot of care to only create one temporary buffer in a session. 

What the hell ?!
I wonder why ":r" and ":r! cat" differ so much. And what is the need of
keeping an unused buffer ...
Anyway. Thanks for the info!

> Also r!  {fname} seems to spawn the shell in the background in gvim,
> so its not too intrusive... 

But, problem: what are the correct external-program amiga, macos, ... ?

> I submitted a feature request about a function filecontents() in vim
> which would basically return a <NL> seperated string of the contents
> of a file, but I have a feeling its going to be neglected like always.
> Sometimes I really wish viml had a few more basic building blocks.

So do I.

> > - Why
> >     if search(a:regexp, 'w')
> >       let retVal = 1
> >     else
> >       let retVal = 0
> >     endif
> >   instead of:
> >       let reVal = search(a:regexp, 'w') > 0
> >   ?

> Heh :) I guess it does make things more compact. I would go so far as to
> say that I should have simply done:
> 
> let retVal = search(..., ...)

Yes and no. "search(...)" is not "boolean-like" as it is an int.
"search(...)>0" is ; as it values 0 or 1.

> [...]
> So whats correct? Depending on return values from boolean algebra might
> lead to inconsistencies, so be warned.

I am. :)

> BTW, Luc: how did you get the logic for what steps to follow to
> generate the DVI file correctly? 

Mainly from Tomer Kol's initial program. So I will not be surprised to
learn that it differs from the ones we can find in others
Makefile/scripts.

> I notice that TKMakeDVI first runs bibtex if file.aux contains
> \bibdata even before running latex. 
> But this seems to be incorrect in situations where the modification is
> to remove a bibliographic entry... i.e, if initially the file
> contained
> 
> \bibliography{fname}
> 
> and I removed this line or changed fname and then did TKMakeDVI, then it
> would wrongly make bibtex scan fname... I hope you see the problem.

Yes indeed.
However, bibtex is also run after latex. And if the .bbl file has
changed then we know that latex must be re-run.
May be to much compilations and bibtex executions are made. But the
result will be correct.

I don't remember why I/we considered the execution of bibtex after latex
wasn't enough. Probably that sometimes bibtex need to be run twice. But
I don't remember when -- I haven't use LaTeX for a while.

So far, it seems that the following may be enough:

0.0- need_to_rerun = 1
x.0- while need_to_rerun
x.1- latex
1.2- if !grep(\bibdata, mainfile.'.tex') => 1.3
     if fileexists(mainfile.'.bbl') => backup
     bibtex
     if .bbl and .bll.bak differs => need_to_rerun = 1
x.3- if grep(Rerun, mainfile.'.log')
x.0- while need_to_rerun
x.1- latex
y.2- nothing
...


BTW: there is a thing i think is a strench of tex-tools: bibtex,
makeindex, dvips, ... are used as 'makeprogram's => the errors can be
seen in a :cwindow.
This requires that compiler/tex.vim defines a function/command that
reset &efm to (pdf)(La)TeX settings -- without having to resource this
file.

-- 
Luc Hermitte
http://hermitte.free.fr/vim/