sirtom67 / upfirdn Goto Github PK

What steps will reproduce the problem?
1. Get a long audio waveform
2. Run upfirdn
3. Returns negative dimension errors in numpy

What is the expected output? What do you see instead?
We expect the files to be processed, but we get a numpy exception

What version of the product are you using? On what operating system?
Latest. Ubuntu 12.04

Please provide any additional information below.

The problem comes from int precision with long files and resampling ratios.
A cast to unsigned long as to be performed to compute the needed sample count 
out.

Here's the diff to solve it

diff -r 8182188fd9ef upfirdn/Resampler.h
--- a/upfirdn/Resampler.h   Sun Jan 02 22:14:30 2011 -0600
+++ b/upfirdn/Resampler.h   Thu Jun 20 14:53:27 2013 -0700
@@ -134,7 +134,7 @@
 int Resampler<S1, S2, C>::neededOutCount(int inCount)
 /* compute how many outputs will be generated for inCount inputs  */
 {
-    int np = inCount * _upRate;
+    unsigned long np = inCount * (unsigned long) _upRate;
     int need = np / _downRate;
     if ((_t + _upRate * _xOffset) < (np % _downRate))
         need++;

[deleted issue]

Would it be alright if I port this code to scipy? I think it would be useful to 
have this functionality there. The site says that the code is BSD licensed, is 
that correct?

Hi, I spent a few minutes to make upfirdn work on Python 3.  I haven't tested 
extensively yet, but the test() function passes for python 2.7.6 and 3.4.0 on 
Ubuntu 14.04.