octave-maintainers
[Top][All Lists]
Advanced

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

Re: Collapsed directories in libinterp


From: Rik
Subject: Re: Collapsed directories in libinterp
Date: Thu, 04 Jul 2013 12:48:19 -0700

On 07/04/2013 11:54 AM, Daniel J Sebald wrote:
> On 07/03/2013 09:42 PM, Rik wrote:
>> 7/3/13
>>
>> Jordi,
>>
>> I went ahead and collapsed interp-core and interpfcn into the existing
>> corefcn directory.
>>
>> --Rik
>
> This is going back to everything in one directory.  My feeling is that
> there really was some sense and usefulness to the categorization, but
> there was a lot of confusion as a result of having, perhaps, too deep of
> a hierarchy but mostly because of choosing names for directories that are
> so similar.
>
> With this change, now things like fft.cc are back in the same directory
> as mex.cc and jit-util.cc.  Those seem so far afield.  Also, DEFUN really
> is a good delineation.  (I wish that DEFUN could be modified to make
> Octave compile faster when touching one of those built-in
> routines--that's the one major complaint I've had from a development
> standpoint.)  I think that having a separation of "interp" and "corefcn"
> makes the most sense and perhaps bringing them back a level so that core
> functions fall under ./libcore or ./corefcn.  Core functions really
> shouldn't be under ./libinterp because things like fft.cc haven't
> anything to do with the interpreter.  The interpreter is some of the
> trickier code to deal with and best left to a few people to fix who
> really know what they are doing.  I rarely ventured there.  Core
> functions, on the other hand, are a little more broad and more
> straightforward--they are easy to work with once one understands the
> Octave data objects.
>
I think we are probably narrowing in on the correct level of separation. 
It used to be that everything, including octave.cc, was in a single src/
directory; that was too much in one directory.  Splitting into libinterp
and having several sub-directories was a good thing.  However I agree with
Jordi that the separation between interpfcn and interp-core didn't turn out
to be as useful as I first thought.  We could try splitting the existing
corefcn into a corefcn or mathfcn directory and an interpfcn directory.  We
could also try this arrangement for a few weeks and see how it goes.  In
the end it isn't particularly hard to change this division with Mercurial.

--Rik




reply via email to

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