numpy.genfromtxt
accepts iterators as well as files. That means it will accept the output of itertools.islice
. Here, test.txt
is a five-line file:
>>> import itertools, numpy
>>> with open('test.txt') as t_in:
... numpy.genfromtxt(itertools.islice(t_in, 3))
...
array([[ 1., 2., 3., 4., 5.],
[ 6., 7., 8., 9., 10.],
[ 11., 12., 13., 14., 15.]])
One might think this would be slower than letting numpy
handle the file IO, but a quick test suggests otherwise. genfromtxt
provides a skip_footer
keyword argument that you can use if you know how long the file is...
>>> numpy.genfromtxt('test.txt', skip_footer=2)
array([[ 1., 2., 3., 4., 5.],
[ 6., 7., 8., 9., 10.],
[ 11., 12., 13., 14., 15.]])
...but a few informal tests on a 1000-line file suggest that using islice
is faster even if you skip only a few lines:
>>> def get(nlines, islice=itertools.islice):
... with open('test.txt') as t_in:
... numpy.genfromtxt(islice(t_in, nlines))
...
>>> %timeit get(3)
1000 loops, best of 3: 338 us per loop
>>> %timeit numpy.genfromtxt('test.txt', skip_footer=997)
100 loops, best of 3: 4.92 ms per loop
>>> %timeit get(300)
100 loops, best of 3: 5.04 ms per loop
>>> %timeit numpy.genfromtxt('test.txt', skip_footer=700)
100 loops, best of 3: 8.48 ms per loop
>>> %timeit get(999)
100 loops, best of 3: 16.2 ms per loop
>>> %timeit numpy.genfromtxt('test.txt', skip_footer=1)
100 loops, best of 3: 16.7 ms per loop