On 9/4/10 11:13 PM, Robert Bradshaw wrote:
On Sat, Sep 4, 2010 at 7:33 PM, Jason Grout<jason-s...@creativetrax.com> wrote:
In improving fast_callable, I'm working on removing fast_float so we don't
have to maintain it. Of course, the pickle jar complains that, with the
fast_float code gone, a pickle in the pickle jar doesn't unpickle.
So...
1. Do people care if we can't unpickle fast_float objects? (that would save
a lot of effort if it's not needed)
2. There seem to be two ways to unpickle fast_float objects: (a) keep the
old code around, or (b) convert them to fast_callable objects. (a) leads to
a bigger maintenance problem for us, while (b) leads to a (lot?) more
effort, and possibly slight incompatibilities. Any preference?
I doubt there's a lot of fast_float pickled objects laying around, but
it may be worth deprecating them before removing them entirely. Also,
in case some pickle is holding on to a reference of a fast_float, it
may make sense to always keep the objects around (e.g. they could
unpickle into something that only raises exceptions when used) to
prevent larger pickles from breaking.
yes; good point. Carl Witty already did some of this when he first
wrote fast_callable. I'll add the deprecation warning, but keep the old
code. Note that people that have been using "fast_float" in the last
year or so have already been using fast_callable by default.
One thing I am doing, though, is removing support for fast_float from
the objects in Sage (and making the _fast_float_ methods just simple
wrappers around the _fast_callable_ methods). So the idea is that the
old fast_float evaluation code will still be there, so pickled objects
("compiled" functions) can still be executed. However, fast_float
objects using the old code won't be able to be created (easily).
Jason
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
