Beginning to be disapointed

[Johann Hibschman]

Michael A Miller writes:

> The innards of _tableio doesn't use NumPy at all.  All it does is
> tokenize a file, run atof on it and return a list.  The list is
> stuffed into a dictionary and there certainly could be some
> weirdness in my reference counting there, but I've never
> experienced the sort of problem that Demers reports.  NumPy is
> used in a python wrapper to convert the list to an array.

Ah, that's probably not it, then.  Sorry to malign your module, but my
first reaction in this sort of thing is always to look for bugs in
extensions.  I had some bad C++ code in a module of my own once do
something like this.  It didn't consume memory and thrash the disk,
but it operated for a while, then seg-faulted mysteriously.

Bring on the code samples!  ;-)

-- 
Johann Hibschman                           johann at physics.berkeley.edu