mailing-list for TeXmacs Users

[Henri Lesourd <address@hidden>]

> I use Tools->Update->Styles after altering the .ts file.

OK, so the cache problems should not apply to your current problem, then.

> Anyway, it's all academic because, thanks to your suggestion, the 
> problem is solved. Specifically, you said:
>
> > This is why in my previous mail, I suggested to
> > redefine <bibitem*|...> _itself_.
> >
> > This way you shunt completely the old code and start completely
> > from scratch, instead of having to reason in detail about how
> > a slight change to this old code operates.
>
>
> But, as you said in your last mail, the key is to start from SCRATCH,

Yes, this is a very useful key, not only in the context
of TeXmacs macros, indeed...

The main problem with software is that what you believe it does
when you read it and the way it behaves in the real world are
two **different** things. Therefore, starting from scratch is
the simplest way to wipe out doubts about this gap. Of course,
a more elaborated approach is needed if you want to maintain a
big piece of software, work together with other people, etc.

But nevertheless, reminding the following simple facts remains
extremely useful in any case :

-> software as you read it tends to be extremely deceiving. This
  is not an exception, but should be considered rather as the
  general rule ;

-> and especially, doing untested **hypothesis** about how the
  software behaves, and trying to reason in your mind using
  such untested hypothesis IS, and WILL REMAIN, extremely
  error-prone in ANY case ;

=> To summarize, what you can do by reading the code (and it
  is indeed extremely useful to do this) is only GUESS what
  the software does. But to be SURE is another matter (namely,
  it is a matter of testing).

I state it explicitely because it seems that lots
of people (especially the promoters of the various
software "technologies", whatever they could be at
one point in time) tend to underestimate, to forget
it, or even worse, to believe that in some (magical)
future, this aspect of software could be removed. But
of course, this is just plain irrealistic.

It's exactly the same as when you learn mathematics : everybody
will tell you that one cannot do mathematics just by reading
a paper or a book : you need a pencil a a sheet of paper, to
be able to redo what the author does, and fully become conscious
of what the real matter in the paper/book is. Things are not
different in software.


> I think that, by retaining render-bibitem in the chain, this 
> compact-item was what was ultimately causing some or all of the 
> confusion. I think you were suggesting something along these lines in 
> your last message by saying that there was another expansion of MFO 
> under bibitem*?

There is for sure another expansion of MFO under bibitem*, otherwise I don't
see how bibitem* would be able to change the alignment as it does.


> However, just dispensing with all the old code in bibitem is a little 
> drastic - for starters, it messes up the citation texmacs-labels 
> (obvious enough, as this is the only other thing that bibitem* does 
> apart from render the bibtex-label in text, cf. std-automatic.ts).

Yes, it is drastic. I wonder if this bug has been solved in 1.0.6. If not,
we will need a more fine-grained approach.