CODEWITHSUNDEEP
pythonpandasnumpy
jottbe
jottbe

Reputation: 4521

How to get the integer positions of the True entries in a pandas indexer

I would like to get the integer positions of rows in a pandas series, where the series contains specific values, or for the positions in an boolean indexer, where the value is True.

If I have the following dataframe, I would like to get the integer positions of the True values in column label:

import pandas as pd
data=dict()
data["col_0"]=[-0.2518508748588564, -0.6339192005025384, -0.6938892242609978, 2.4470042529183402, 0.8977665869071174]
data["label"]=[False, False, True, False, True]
df2= pd.DataFrame(data)

I could of course do that like below, but it seems a big awkward to me and I wonder, if there is a cleaner way to do this (especially without a reset_index, but the solution should work no matter what index labels are used in the original dataframe):

ser= df2['label']
new_ser= ser.reset_index(drop=True)
new_ser.index[new_ser].to_list()

The result is of course [2, 4].

Upvotes: 1

Views: 140

Answers (1)

David Erickson
David Erickson

Reputation: 16683

You can use:

np.flatnonzero(df2['label'])

data=dict()
data["col_0"]=[-0.2518508748588564, -0.6339192005025384, -0.6938892242609978, 2.4470042529183402, 0.8977665869071174]
data["label"]=[False, False, True, False, True]
df2= pd.DataFrame(data)
np.flatnonzero(df2['label'])

Out[1]: array([2, 4], dtype=int64)

That makes it an np.array. To make a list, use:

[*np.flatnonzero(df2['label'])]
Out[2]: [2, 4]

Upvotes: 1

Related Questions