Bottleneck 0.6 release candidate 1

[Keith Goodman]

Test reports of the first release candidate of Bottleneck 0.6 would be
greatly appreciated. Include OS, python and numpy versions, and the
result of running bottleneck.test() and, optionally (well, it's all
optional), bottleneck.bench()

You can download Bottleneck 0.6rc1 here:
https://github.com/kwgoodman/bottleneck/downloads. And install with
"python setup.py install"

Or, if you'd rather pull from github, then use the 0.6rc1 tag. And
install with "make all".

Release notes: https://github.com/kwgoodman/bottleneck/blob/master/RELEASE.rst

[chrisspalm]

Works fine on win32.

In [1]: import bottleneck as bn

In [2]: bn.__version__
Out[2]: '0.6.0rc1'

In [3]: bn.test()
Running unit tests for bottleneck
NumPy version 1.6.1
NumPy is installed in C:\Python27\lib\site-packages\numpy-1.6.1-py2.7-
win32.egg\numpy
Python version 2.7.3 (default, Apr 10 2012, 23:31:26) [MSC v.1500 32
bit (Intel)]
nose version 1.1.2
..........................................................................................
----------------------------------------------------------------------
Ran 90 tests in 53.083s

OK
Out[3]: <nose.result.TextTestResult run=90 errors=0 failures=0>

In [4]: from platform import platform

In [5]: platform()
Out[5]: 'Windows-7-6.1.7600'

[Keith Goodman]

I'm glad all tests pass on win32/python2.7.3/numpy1.6.1. Thank you.

[chrisspalm]

Here's the bench too on the same machine.

In [4]: bn.bench()
Bottleneck performance benchmark
Bottleneck 0.6.0rc1
Numpy (np) 1.6.1
Scipy (sp) 0.9.0
Speed is NumPy or SciPy time divided by Bottleneck time
NaN means one-third NaNs; float64 and axis=1 are used
High-level functions used (mode='fast')

no NaN no NaN no NaN NaN
NaN NaN
(10,10) (100,100) (1000,1000) (10,10) (100,100)
(1000,1000)
median 3.57 1.71 2.34 3.37
2.97 5.74
nanmedian 89.44 22.43 5.21 55.33
10.95 1.87
nansum 7.41 4.07 5.63 4.06
0.40 0.68
nanmax 1.64 1.16 1.01 1.43
1.09 1.07
nanmean 17.50 8.69 9.38 9.92
1.70 1.83
nanstd 24.90 6.43 9.11 10.66
1.02 1.31
nanargmax 5.94 3.86 6.58 3.51
0.44 0.69
ss 3.84 2.82 5.51 1.91
1.40 1.41
rankdata 23.56 17.00 13.55 12.55
8.63 5.36
partsort 1.16 1.81 2.40 2.60
3.49 6.68
argpartsort 0.62 1.78 2.01 2.10
3.41 6.30
replace 2.88 2.21 2.22 2.88
2.21 2.00
anynan 2.47 3.07 2.84 1.96
3.84 46.74
move_sum 8.58 100.22 623.85 5.32 48.87
495.47
move_nansum 20.11 112.98 683.94 9.58 12.79
118.75
move_mean 8.46 71.84 531.29 5.27 48.54
487.48
move_nanmean 21.71 81.16 563.07 10.88 12.03
108.60
move_std 10.49 24.87 222.32 13.43 88.00
818.85
move_nanstd 26.43 38.19 328.89 15.69 16.64
146.23
move_max 3.64 17.67 147.75 2.21 16.34
161.52
move_nanmax 14.47 20.61 163.93 7.34
6.91 62.78

Reference functions:
median np.median
nanmedian local copy of sp.stats.nanmedian
nansum np.nansum
nanmax np.nanmax
nanmean local copy of sp.stats.nanmean
nanstd local copy of sp.stats.nanstd
nanargmax np.nanargmax
ss scipy.stats.ss
rankdata scipy.stats.rankdata based (axis support added)
partsort np.sort, n=max(a.shape[1]/2,1)
argpartsort np.argsort, n=max(a.shape[1]/2,1)
replace np.putmask based (see bn.slow.replace)
anynan np.isnan(arr).any(axis)
move_sum sp.ndimage.convolve1d based, window=a.shape[1] // 5
move_nansum sp.ndimage.convolve1d based, window=a.shape[1] // 5
move_mean sp.ndimage.convolve1d based, window=a.shape[1] // 5
move_nanmean sp.ndimage.convolve1d based, window=a.shape[1] // 5
move_std sp.ndimage.convolve1d based, window=a.shape[1] // 5
move_nanstd sp.ndimage.convolve1d based, window=a.shape[1] // 5
move_max sp.ndimage.maximum_filter1d based, window=a.shape[1] //
5
move_nanmax sp.ndimage.maximum_filter1d based, window=a.shape[1] //
5

[Keith Goodman]

I tested on 64-bit ubuntu 11.10 with python2.7.2/numpy1.6.2 and python
3.2.2/numpy1.6.1. All tests passed. So I guess we're ready for a
release.