Feature request: pandas.DataFrame.setdefault()

[Sereger13]

We often find ourselves checking for column existence before assigning a value to it, e.g.:

df = pd.DataFrame()
col = 'ColumnA'
val = 'Value1'
if col not in df.columns:
    df[col] = val

And we were thinking it would be nice if there was a way to tell pandas: add this column with that value(s) but only if it does not exist already - somewhat similar to dict.setdefault() so we could say, for instance:

df.setdefault(col, val)

Doesn't necessarily have to be that syntax obviously - just a suggestion. What do you guys think?

[sinhrks]

can u describe the usecase? I'm not sure on what timing a new column is created based on the specified default.

[jreback]

This does it, though it operates on an element-by-element basis.

In [30]: df = pd.DataFrame()

In [31]: df.combine_first(pd.DataFrame({'A' : [1,2,3]}))
Out[31]: 
   A
0  1
1  2
2  3

In [32]: pd.DataFrame({'A' : [1,2,3]}).combine_first(pd.DataFrame({'A' : [4,5,6]}))
Out[32]: 
   A
0  1
1  2
2  3

[jorisvandenbossche]

API wise, this could be implemented as an additional keyword to DataFrame.assign.

Eg something like df.assign(col=val, if_exists='ignore'|'overwrite'), although not sure it is worth the extra keyword (the beauty of assign is that it takes currently no keyword arguments apart from the columns to add)

[shoyer]

I think this could be worth considering as part of the broader question of whether DataFrame and Series should satisfy the mapping ABC.

[Sereger13]

If we decide to go DataFrame.assign route, can I also suggest adding inplace=True|False parameter so we could avoid copying the entire DataFrame when adding a column.

[mroeschke]

Due to lack of activity and interest, closing