[Numpy-discussion] creating a Numeric array from a numpy array LEAKS memory

Pauli Virtanen pav@iki...
Fri Oct 24 13:52:06 CDT 2008

Fri, 24 Oct 2008 14:39:59 -0400, Jose Borreguero wrote:

> My bad. Using the intermediate list does *not* leak. Still, the original
> problems stays. Can anyone run the following code in their machine and
> see if they have leaks? Maybe it only happens to me :(*
> import numpy,Numeric
> big=10000000
> na=numpy.array([0.0,])
> for i in range(big):
>     Na=Numeric.array(na,Numeric.Float)*

Yep, leaks also here: (Numeric 24.2, numpy 1.2.0)

	import sys, numpy, Numeric
	na = numpy.array([0.0])
	for i in xrange(1000000):
	    foo = Numeric.array(na, Numeric.Float)
	print sys.getrefcount(na)

The getrefcount prints 1000002, so it seems like there's a refcount error 
somewhere. But

	na = numpy.array([0.0])
	for i in xrange(1000000):
	    foo = numpy.array(na, numpy.float_)
	print sys.getrefcount(na)

refcounts correctly.

Pauli Virtanen

More information about the Numpy-discussion mailing list