Bottleneck is a collection of fast NumPy array functions written in Cython.
Let's give it a try. Create a NumPy array:
>>> import numpy as np
>>> arr = np.array([1, 2, np.nan, 4, 5])Find the nanmean:
>>> import bottleneck as bn
>>> bn.nanmean(arr)
3.0Moving window mean:
>>> bn.move_mean(arr, window=2, min_count=1)
array([ 1. , 1.5, 2. , 4. , 4.5])Bottleneck comes with a benchmark suite:
>>> bn.bench()
Bottleneck performance benchmark
Bottleneck 1.1.0dev; Numpy 1.11.0
Speed is NumPy time divided by Bottleneck time
NaN means approx one-third NaNs; float64 and axis=-1 are used
no NaN no NaN NaN NaN
(10,) (1000,1000) (10,) (1000,1000)
nansum 34.0 3.9 34.1 6.4
nanmean 117.1 5.6 118.5 7.3
nanstd 196.5 4.2 202.6 7.0
nanvar 190.0 4.5 195.6 6.8
nanmin 29.9 1.1 30.4 1.4
nanmax 29.8 1.0 30.6 1.8
median 52.6 0.7 60.0 2.9
nanmedian 48.0 2.4 55.9 6.0
ss 13.4 3.8 3.1 1.1
nanargmin 54.3 4.0 54.3 5.5
nanargmax 54.7 4.0 55.3 6.6
anynan 13.9 1.5 14.2 179.1
allnan 14.4 209.0 14.1 194.1
rankdata 34.4 1.5 36.1 2.2
nanrankdata 47.4 20.8 41.6 34.6
partsort 4.5 0.8 4.9 1.1
argpartsort 2.5 0.7 2.5 0.5
replace 8.2 1.9 8.2 2.0
push 149.9 10.4 152.7 14.5
move_sum 294.7 219.7 301.4 336.1
move_mean 709.9 248.0 729.5 489.8
move_std 1191.6 133.3 1223.3 363.8
move_var 1165.8 183.3 1202.6 379.1
move_min 237.6 22.2 235.3 48.6
move_max 242.4 22.6 239.9 73.1
move_argmin 364.6 79.9 385.0 235.4
move_argmax 368.7 80.7 384.2 257.4
move_median 435.8 48.7 435.2 138.5
move_rank 587.4 3.8 653.6 10.8Only arrays with data type (dtype) int32, int64, float32, and float64 are accelerated. All other dtypes result in calls to slower, unaccelerated functions. In the rare case of a byte-swapped input array (e.g. a big-endian array on a little-endian operating system) the function will not be accelerated regardless of dtype.
|
http://pypi.python.org/pypi/Bottleneck |
|
http://berkeleyanalytics.com/bottleneck |
|
http://github.com/kwgoodman/bottleneck |
|
http://groups.google.com/group/bottle-neck |
Bottleneck is distributed under a Simplified BSD license. See the LICENSE file for details.
Requirements:
| Bottleneck | Python 2.7, 3.4, 3.5; NumPy 1.11.0 |
| Compile | gcc or clang or MinGW |
| Unit tests | nose |
| Cython | Optional for released version of Bottleneck |
If Cython is installed on your computer then you can install either a released version of Bottleneck (PyPI) or a development version (GitHub).
If Cython is not installed on your computer then you can only install a released version of Bottleneck (PyPI). Cython is not required because the Cython files have already been converted to C source files in Bottleneck releases.
To install Bottleneck on GNU/Linux, Mac OS X, et al.:
$ sudo python setup.py installTo install bottleneck on Windows, first install MinGW and add it to your system path. Then install Bottleneck with the commands:
python setup.py install --compiler=mingw32Alternatively, you can use the Windows binaries created by Christoph Gohlke: http://www.lfd.uci.edu/~gohlke/pythonlibs/#bottleneck
After you have installed Bottleneck, run the suite of unit tests:
>>> import bottleneck as bn
>>> bn.test()
<snip>
Ran 114 tests in 18.978s
OK
<nose.result.TextTestResult run=114 errors=0 failures=0>