在Pandas DataFrame中对布尔值的列进行排序

Question

I am attempting to learn time series. I want to find the dates that are linked with a boolean value that is True . I then assigned the boolean to the pd.DataFrame .

I assigned the boolean statements to the column named 50+ , like this:

How do I sort the True rows from the column 50+ ?

I have searched the internet and have not found a solution. Since I passed the 50+ from a boolean to the dateframe, doesn't this make it a normal string that can be sorted by the sort value function?

Answer 1

You need to specify the column name:

>>> import pandas as pd
>>> import numpy as np
>>> np.random.seed(123)

>>> idx = pd.date_range('2018-10-05', periods=7, freq='D')

>>> df = pd.DataFrame({'data': np.random.randn(idx.size),
...                    '50+': np.random.choice([0, 1], size=idx.size).astype(bool)},
...                   index=idx)

>>> df
                data    50+
2018-10-05 -1.085631   True
2018-10-06  0.997345   True
2018-10-07  0.282978  False
2018-10-08 -1.506295  False
2018-10-09 -0.578600  False
2018-10-10  1.651437   True
2018-10-11 -2.426679  False

>>> df.sort_values('50+')
                data    50+
2018-10-07  0.282978  False
2018-10-08 -1.506295  False
2018-10-09 -0.578600  False
2018-10-11 -2.426679  False
2018-10-05 -1.085631   True
2018-10-06  0.997345   True
2018-10-10  1.651437   True

>>> df.sort_values('50+', ascending=False)
                data    50+
2018-10-05 -1.085631   True
2018-10-06  0.997345   True
2018-10-10  1.651437   True
2018-10-07  0.282978  False
2018-10-08 -1.506295  False
2018-10-09 -0.578600  False
2018-10-11 -2.426679  False

If you're uncertain, you can always check the docstring .

The default is ascending=True , which will put False s first, because they are just 0s under the hood. (While True is 1.)

If you'd like to filter to rows where this column is True, you can use:

>>> df[df['50+']]
                data   50+
2018-10-05 -1.085631  True
2018-10-06  0.997345  True
2018-10-10  1.651437  True

Answer 2

I want to find the dates that are linked with a boolean value that is True.

You don't need to sort anything for this. You need only use Boolean indexing; in other words, construct a Boolean series or array with the same length as your dataframe index and apply it via __getitem__ , called by the syntax [] .

So, instead of "assigning the Boolean to pd.DataFrame ", just index the index !

index_filtered = df.index[df['50+']]