使用混合数据保存numpy数组

Question

I have a numpy array where every value is a float followed by an integer, eg:

my_array = numpy.array([0.4324321, 0, 0.9437212, 1, 0.4738721, 0, 0.49327321, 0])

I would like to save it like this:

0.4324321 0 0.9437212 1 0.4738721 0 0.49327321 0

But if I call:

numpy.savetxt('output.dat',my_array,fmt='%f %i')

I get an error:

AttributeError: fmt has wrong number of % formats.  %f %i

How can I fix this?

Answer 1

Your real problem is that printing out a 1D 8-element array gives you 8 rows of 1 column (or, if you force things, 1 row of 8 columns), not 4 rows of 2 columns. So, you can only specify a single format (or, if you force things, either 1 or 8 formats).

If you want to output this in a 4x2 shape instead of 1x8, you need to reshape the array first:

numpy.savetxt('output.dat', my_array.reshape((4,2)), fmt='%f %i')

This will give you:

0.432432 0
0.943721 1
0.473872 0
0.493273 0

The docs are a little confusing, as they devote most of the wording to dealing with complex numbers instead of simple floats and ints, but the basic rules are the same. You specify either a single specifier, or a specifier for each column (the in-between case of specifying real and imaginary parts for each column isn't relevant).

---

If you want to write it in 1 row of 8 columns, first you need to reshape it into something with 1 row of 8 columns instead of 8 rows.

And then you need to specify 8 formats. There's no way to tell numpy "repeat these two formats four times", but that's pretty easy to do without numpy's help:

numpy.savetxt('output.dat', my_array.reshape((1,8)), fmt='%f %i ' * 4)

And that gives you:

0.432432 0 0.943721 1 0.473872 0 0.493273 0 

Answer 2

The problem is that savetxt() will print one row for each array entry. You can force a 2D-array creating a new axis and then print the (1x8) format:

numpy.savetxt('output.dat', my_array[numpy.newaxis,:], fmt='%f %i'*4)