[Python-Dev] Hi, and a question about filename completion via readline in Python 2.3

[Fernando Perez]

Hi all,

since this is my first post here, I figured I'd briefly introduce myself.  I'm 
the lead developer of ipython (http://ipython.scipy.org), a shell for 
interactive python work.  I've been a fan (and big advocate) of python over 
the last 2 years, mainly for scientific computing.  For those interested, you 
can find a brief overview of ipython at:

http://ipython.scipy.org/misc/ipython_scipy03.pdf

Besides thanking the python team for this fantastic language, I have a 
question concerning some changes apparently made to the C part of readline 
support for  python 2.3.  If I understood correctly this thread:

http://mail.python.org/pipermail/python-list/2002-February/085063.html

as of python 2.3, readline will NOT default anymore to providing filename 
completions when nothing matches in the available namespaces.  There is 
mention in that thread of making it a configurable parameter, but I failed to 
find any specifics on how to do that.

While perhaps I'm missing something, so far (after getting this reported as an 
ipython bug from 2.3 users) I've failed to be able to recover filename 
completion under 2.3.

While I'm sure you found good reasons for this change, under some environments 
it turns out to be really quite nasty.  One of the strengths of ipython is its 
tight integration with the underlying system environment, and its users rely 
constantly on filename/directory completion for navigating their filesystem as 
  they work.  I personally (and many others) use ipython as my main python 
shell, and I fell it provides a superior working environment to anything else 
I've seen out there (especially if you need to combine python work with 
manipulating underlying files).

So ultimately my concerns are:

- is there currently a way to re-enable readline filename completion, as 
suggested in that thread?

- if not, can such a change be made?

I hope if some of you test ipython, you'll find out that actually that kind of 
integration with the filesystem can make a very strong argument for 
python-based shells as fantastic tools for many data-driven environments.  I 
have already knowledge of several projects which are either using or 
considering ipython as the underlying shell for scientific data analysis 
environments, and this change will be a major setback for their usability.

Again, many thanks for all your hard work on python.

Best regards,

Fernando.