Fixed point support
Tim Peters
tim.one at comcast.net
Sat Feb 22 12:21:28 EST 2003
[Carlos Ribeiro]
> I was reading the latest python-dev summary as posted here, and
> it seems that the debate about FixedPoint has surfaced again.
The python-dev incarnation is specifically about whether anyone is going to
do the work to follow up on Guido's willingness to fold
http://fixedpoint.sf.net/
into Python 2.3. At this point I'd rather not, and instead wait for an
implementation of IBM's proposed standard, at
http://www2.hursley.ibm.com/decimal/
That doesn't appear to be possible for 2.3. I expect I could implement that
fully within a week, but don't even have an hour for it. Aahz got a good
start but also seems to lack sufficient time to finish it.
> I remember it being discussed some time ago during my first period of
> activity here (circa 2001).
>
> My question is, isn't proper to discuss fixed point support here,
> at c.l.p, with a bigger audience?
In general, yes. Whether to incorporate the specific module in question
into 2.3 is more a topic for Python-Dev, though, given that it was discussed
& approved on c.l.py last year.
> Or is this discussion going to be fruitless (for lack of interest) or
> endless (because it is so contractory)?
Discussion is mostly useless regardless if nobody can make time to do the
work.
> I have a few words on this subject to share, but first I would like to
> know where is the proper place...
Is IBM's proposal enough for you? If yes, discussion over; if no, you're
unreasonable <wink>.
> [I hope I'm not opening another can of worms. After PEP308 and case
> sensitivity, I feel a little bit tired...]
If there are character sets with mixed-case digits, I'm sure the debate will
never end. Apart from that, there are thorny issues regarding (at least)
literal notation, threads, I/O, and coercions among the other numeric types,
which would need to be resolved for a truly smooth integration into Python.
The biggest "advantage" of the SF library is that it's just another library
module, and Python doesn't know anything about its internals; that cuts off
tons of potential debate.
More information about the Python-list
mailing list