CODEWITHSUNDEEP
pythonnumpy
Steven T
Steven T

Reputation: 95

Numpy Indexing First Boolean

In a boolean array, I am trying to obtain the column index of the first True. argmax works with at least one True, but understandably max(False) is 0. I'm wondering what the best method would be, given a very large array.

name = np.array(['a', 'b', 'c', 'd'])
boolarr = np.array([[True, False, False, True],
                    [False, False, True, True],
                    [False, False, False, False]])
colidx = np.argmax(boolarr,axis=1)
print(name[colidx]) #result: ['a', 'c', 'a'] desired: ['a', 'c', None]

Upvotes: 5

Views: 162

Answers (1)

user3483203
user3483203

Reputation: 51155

You can't change the behavior of argmax, since the maximum of a row with all False is 0. However, you can use any to determine the rows that contain all False, and use np.where to mask your result:

out = name[colidx]
np.where(boolarr.any(1), out, None)


array(['a', 'c', None], dtype=object)

Upvotes: 3

Related Questions