Beginning to be disapointed

[Johann Hibschman]

[Yes, I know.  I'm fouling up the reference chain.  But there's weirdness
afoot, and I can't do any better.]

Hugues Demers writes:

>> 2. how are you reading them in?

> I use a module written by M. Miller called TableIO to read my data because
> it's fast. It read text file into Numeric array

Aha.

This might be it.  IIRC, some of the innards of the reference counting
changed at some point in the NumPy distribution.  If you're using a
different version of NumPy than TableIO was designed for, there might
be a memory allocation problem.

Double-check that, since (to me) the errors you get sound like what
happens to me when I try to free something that I shouldn't.  It runs
for a bit, after the mistake, then dies mysteriously.

I bet that's it.  I bet TableIO fails to set a bit like OWN_DATA or
some such on the inputted arrays, and then something bad happens.
I don't know for sure, though.

As an experiment, try using Konrad Hinsen's ArrayIO module to read the
arrays instead.  That's pure python, so it will be a bit slow, but it
might help you track down the problem. (It's what I use, and I don't
find it to be too terribly slow.)

Good luck,

--Johann

-- 
Johann Hibschman                           johann at physics.berkeley.edu