Custom sort without specifying all values-CodePudding

Is there a way to sort values without having to specify all values in the list? I just want to, for example, sort by economics and then library. The remaining rows should be the same order as the original df.

order = ["economics","library"]

df   cat
0  library
1  economics
2  science
3  np.NaN

Expected Output:

1  economics
13 economics
0  library
...

df.sort_values("cat", key=lambda column = column.map(lambda: x:order.index(x)))

CodePudding user response：

Example

data = ['library', 'economics', 'science', 'aaa', 'bbb']
df = pd.DataFrame(data, columns=['cat'])

df

    cat
0   library
1   economics
2   science
3   aaa
4   bbb

Code

order = ["economics","library"]
out = df.sort_values('cat', key=lambda x: pd.Categorical(x, categories=order, ordered=True))

out

    cat
1   economics
0   library
2   science
3   aaa
4   bbb

CodePudding user response：

You can generate dictionary with enumerate for mapping values in cat column:

print (df)
         cat
0    library
1  economics
2    science
3        new
4  economics

order = ["economics","library"]

print ({v:k for k, v in enumerate(order)})
{'economics': 0, 'library': 1}

df = df.sort_values("cat", key=lambda x: x.map({v:k for k, v in enumerate(order)}))
print (df)
         cat
1  economics
4  economics
0    library
2    science
3        new