[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Axiom-developer] build-improvements and latex

From: Ralf Hemmecke
Subject: Re: [Axiom-developer] build-improvements and latex
Date: Wed, 08 Nov 2006 23:18:23 +0100
User-agent: Thunderbird (X11/20060909)

3) Files with unnecessarily long names are more awkward to use.
Ok I admit that reasons 2) and 3) are not very strong reasons...
Not even 1) is a reason. Why do you believe one can say

make dvi

in ALLPROSE and then say

xdvi myalps.dvi

to get the whole documentation although I have the convention
notangle removes .nw
noweave adds .tex  (result is .nw.tex)

Good question. If the rule is "add .tex" then the result is

  latex myalps.nw.tex --> myalps.nw.dvi

not myalps.dvi. You must be doing something more to hide the .nw.

Your suggestion would require that we use awkwqard names like

  xdvi axiom.sty.pamphlet.dvi


Wow. I must be doing something wrong. I already question whether I should use German to get my idea through. English doesn't seem to work. ;-)

OK. Let me repeat. In ALLPROSE and any project that is built with ALLPROSE there are many files. All of them have the extension .nw (forget about README, GPL and that). Actually all have a second extension. So the general form is file.ext.nw. Now my Makefile basically have the rule that notangle removes .nw and noweave adds .tex.

There is only *one* dvi file that will be produced and *not* one for every .nw file. So "xdvi axiom.sty.pamphlet.dvi" is absolutely no argument. One should not even look at something like axiom.sty.pamphlet.tex, because that is generated and could look like a big mess --- if you have really looked at files that come out of noweave you will know that theses files are not for humans even if they are .tex files.

So now the trick why I don't have myalps.nw.dvi.

The top-level project file is called myalps.tex.nw. Applying my rules gives myalps.tex and myalps.tex.nw.tex. The first one is just a wrapper that in shortended form looks like

\author{Ralf Hemmecke}
  \title{\xProject{} \LIBRARYVERSION}}

and that will be latex'ed. As you can see, it's some kind of template.
And my Makefiles make sure that the setup will be such that all other .nw.tex files will be included into the documentation.

There is another point that speaks against the current practise of translating each .pamphlet file into one .dvi, namely "literate programming". Ideas don't just live in one file. That completely misses the point that there are relations from one part of the system to another. For example, there should not be 5 Makefile.dvi files lying around in different directories. Since the build procedure should be described as a whole.

It would better to describe the interpreter as a whole rather than producing a .dvi file for each of the 187 files in src/interp. Tim's approach (if I understood that correctly) is to make *one* huge .pamphlet file. My approach is to have as many pamphlet file as you like but produce *one* human readable document from all of them to describe a _unit_ (i.e. the interpreter).

Eventually, we perhaps want just one document (a web-site, for example) that describe all of axiom (including the source code). For a web site the size doesn't matter. But still, there should be some backlinks from the human-readable format (.dvi) to the actual .pamphlets.

There will never be generated anything else than one dvi file.
And I don't understand, why people care so much about the names
of intermediate files. You can remove .nw.tex files and no harm is done. If make would already remove them, one would not even
see what strange things are done under the hat.

I don't think anyone is that worried about internal file names,
but this discussion is about making the makefiles "simpler" and
more obvious, so we do care what goes on under the hat, right?

Of course, we care about Makefiles and conventions and such but as long as that becomes simple, generated file names can be complicated.

Using a file name for dvi files that retains the original
.pamphlet extension is not so obvious to me.

I hope, this time my English is understandable.

But that naming scheme simplifies the understanding of the system (only in my opinion, of course). One doesn't have to
rename axiom.sty.pamphlet to axiom-sty.pamphlet.

In my opinion it is only an accident that axiom.sty.pamphlet
has this name and contains only one root chunk. In general this
is not the case.

So the only reason I see not to just add ".tex" would be portability. But which are systems we want to build Axiom on that cannot handle two dots?

I would guess: none.

I agree. Other opinions concerning filename portability?


reply via email to

[Prev in Thread] Current Thread [Next in Thread]