I can understand if it's not needed, but what do you mean by wrong? Can you give an example?

Yes. Now the example on the official page use int(pct/100.*np.sum(allvals)) at the perceptual_absolute(foo):

scores = [116, 133, 18, 19, 4, 2]
def perceptual_absolute(pct, allvals):
        absolute = int(pct/100.*np.sum(allvals))
        return '{:.2f}%\n({:d})'.format(pct, absolute)
plt.pie(scores.values(), explode=explode,
                            autopct=lambda pct: perceptual_absolute(pct, list(scores.values())),
                            pctdistance=1.1, labeldistance=1.2, startangle=0)

int(pct/100.*np.sum(allvals)) get 132 but the correct is 133, given by round(pct/100.*np.sum(allvals)).

The actual code here in master (maybe not deployed yet) use int(round(pct/100.*np.sum(allvals))) but round(foo) already returns a int type. The call int(foo) is not used at the code.

For consistency sake with the rest of our examples, it should probably just be np.round and let the formatter take care of making it an integer.

In this case may be necessary to do int(np.round(foo)) because np.round(foo) returns numpy.float type.

because np.round(foo) returns numpy.float type.

This is actually a bug in numpy (numpy/numpy#15297) that is fixed in more recent versions. See numpy/numpy#15840

So if you do pip install --upgrade numpy that behavior should go away.

Edit: woops I didn't read this line from late in the PR:

No, we didn't change np.round, only ndarray.round.

Yeah, fair enough - I didn't realize that '{:d}'.format(np.round(3.4)) fails with ValueError: Unknown format code 'd' for object of type 'float'. print('%d'%np.round(3.4)) works just fine.

Interesting point:

print('%d'% 56)
print('%d'% 56.0)
print('%d'% 56.9)

All print "56".

print('%d'% '-')
print('%d'% 45)

All print "-". The conversion is forced before the string replacement.

I have edit to use int(np.round(foo)) instead round(foo).