I can add a categorical column to a Pandas DataFrame like so:
import pandas as pd
label_type = pd.api.types.CategoricalDtype(categories=["positive", "negative"], ordered=False)
d = {'col1': [1, 2], 'col2': [3, 4]}
df = pd.DataFrame(data=d)
# Create a new column, setting the value universally to "positive"
df['label'] = pd.Series(["positive"] * len(df), dtype=label_type).values
This is less elegant than this shorthand with other types:
df['label2'] = "positive" # sets entire column to str("positive")
but it seems like the underlying type is just a str
print(type(df['label'].iloc[0]))
<class 'str'>
so it seems like the column-type has to be known ahead of time to pandas.
Is there any way to add a categorical column to a dataframe without manually constructing the Series? For example,
df['label3'] = label_type("positive")
How about this:
df['col4'] = df.assign(col4 = 'positive')['col4'].astype(label_type)
df.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2 entries, 0 to 1
Data columns (total 4 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 col1 2 non-null int64
1 col2 2 non-null int64
2 label 2 non-null category
3 col4 2 non-null category
dtypes: category(2), int64(2)
memory usage: 412.0 bytes
Though you still get an str type:
type(df['col4'].iloc[0])
str
Since I think that in this case iloc[] will return a string representation of the category.
Or just do it in two steps:
df['col4'] = 'positive'
df['col4'] = df['col4'].astype(label_type)