other python ideas

[Carlos Alberto Reis Ribeiro]

At 09:00 09/04/01 +0000, Remco Gerlich wrote:
>Recently people want to get rid of import, change the clear semantics of the
>assignment statement, add stricter interfaces, make more private members of
>classes, do bytecode optimization...

Sorry. I'm involved in more than one of the above, and I think that 
sometimes the *intention* get lost in the middle of the discussion. As for 
some of the points that you raised:

* I don't want to change the semantics of the assignment statement. I just 
wanted to have a way to differentiate between temporary objects created as 
intermediate results in expressions, and objects that are already bound to 
some name. In the end, Robin Thomas found a nice way to solve the problem 
*without* having to resort to a magic __assign__ method. I believe that 
this may lead to a big improvement, specially for NumPy operations.

* The discussion about interfaces is going nicely. Lots of proposals and 
counter-proposals are being raised, ranging from the full-Java-like 
interfaces to the lets-it-stay-as-it-is crowd. I think that, in the end, we 
will have a good mechanism for interface implementation and assertion.

* As for the bytecode optimization, after some discussion, I'm convinced 
that it is not a easy thing to do. But this is something that can be done 
*without* changing anyone's code, and it have the potential to benefit 
everyone. It's just a matter of looking at what to optimize. Simply 
dismissing optimization without looking at the potential gain is as bad as 
the opposite - proposing optimizations for the sake of optimization, 
without checking first what kind of gain could be attained with it.

Last but not least, I'd like to point out that I'm not comfortable with the 
rapid pace of development of the past few months. Some weird things are 
being proposed. I think that this is related to the recent changes related 
to  the Python development team. As things get more stable, the pace will 
be moderated naturally.

Python is a very good language as it is. We (the user community) will 
always be looking for improvements, specially if we can make them slowly, 
in a incremental way, without breaking anyone's code. That search for 
improvements, allied with a common sensical approach (most times anyway 
:-), is one of the best things in the Python core team.


Carlos Ribeiro