Pandas: Set a value on a data-frame using loc then iloc

Question

Let's assume that I have the following data-frame:

df = pd.DataFrame(np.zeros((3, 5)), columns=["feature_a", "feature_b", "feature_c", "feature_d", "e"])

    feature_a   feature_b   feature_c   feature_d   e
0      0.0         0.0         0.0         0.0     0.0
1      0.0         0.0         0.0         0.0     0.0
2      0.0         0.0         0.0         0.0     0.0

However, please note that the data-frame that I am dealing with is much larger. Then, what I would like to do is update the value of the middle two feature columns, so that the result would be:

    feature_a   feature_b   feature_c   feature_d   e
0      0.0         0.0         0.0         0.0     0.0
1      0.0         8.0         8.0         0.0     0.0
2      0.0         0.0         0.0         0.0     0.0

What I have tried that I assumed it would work:

feature_columns = df.filter(like="feature").columns.values
df.loc[:,feature_columns].iloc[1,[1, 2]] = 88

For me it's utterly important that this is done following the pattern that I have tried. The reason for that is:

1. The columns that I wish to update strictly contain a pattern.
2. Once I have the columns selected I know the row and columns index that I wish to update.

To conclude, my question is how can I go from the starting data-frame to the resulting data-frame, while having a solution that will follow the way of doing things that I have tried.

Answer 1

This is how you deal with the specific example you show in your question:

import pandas as pd

df = pd.DataFrame(np.zeros((3, 5)), columns=["feature_a", "feature_b", "feature_c", "feature_d", "e"])
feature_columns = df.filter(like="feature").columns.values

sli = df[feature_columns].iloc[1,[1,2]]
df.loc[sli.name, sli.index] = 88
print(df)
# output
#        feature_a  feature_b  feature_c  feature_d    e
#     0        0.0        0.0        0.0        0.0  0.0
#     1        0.0       88.0       88.0        0.0  0.0
#     2        0.0        0.0        0.0        0.0  0.0

Depending on the precise application, you might have to change the exact syntax around a bit, but the idea is broadly applicable: harvest the columns and indices of your selection, then use those to slice df.loc , then assign to that slice. That should get around the "assign to copy" issue you're running into.