Re: [gnulib-tool-py] gnulib-tool vs pygnulib: application vs module

[Bruno Haible]

Hi Dmitriy,

> Recently I've thought that it could be a great
> idea in addition to usual gnulib-tool give the user a possibility to write
> own scripts basing on functions and classes from pygnulib package. Here are
> some advantages:
> 1. It can make life easier, because sequence of commands can even more
> convenient than commands in the shell.

It is a very good idea. In fact, the MODULES.html.sh script, which generates
http://www.gnu.org/software/gnulib/MODULES.html, is currently making
thousands of invocations to gnulib-tool and is therefore extremely slow
(takes half an hour to execute). It would be 100 times faster if rewritten
in Python, after your project is finished.

And in the long run, you know that gnulib-tool is only part of the GNU
Build System. It is conceivable that people will want to do things with
gnulib-tool that we cannot imagine today, and this will be so much easier
to implement if the core of gnulib-tool consists of Python classes.

> What do you think about this idea? If it is a good one, it would be better
> to say it now, when the development process is at beginning. It will be
> hard to implement this feature later.

It is still possible to implement this feature later. (Reshuffling code
with the objective of giving it better structure is called "refactoring",
and is easy when there is a good test suite.) Of course I'll love it if
you can make a clean class design from the existing code, but it is not
a hard requirement.

Bruno