Refactor DataFrame to fill NaNs with data from other column, combine rows

Question

Desire is input DataFrame reduced to name and color columns. If updated_name|color has data, use it. If '' (Nan in true data set) then either use the value entered in updated_name|color column if it exists, and if not use the value in original_name|color column.

In real data set got about as far as the DataFrame below using:

mask = df['updated_name'].notnull() | df['updated_color'].notnull()

I am stuck on next steps.

An example DataFrame is worth a thousand words, minimum viable as follows:

Input DataFrame:

sql_Key	original_name	original_color	updated_name	updated_color
1	Happy Bear	Red	Green Bear
1	Happy Bear	Red
1	Happy Bear	Red
2	Grumpy Bear	Blue
2	Grumpy Bear	Blue		Purple
2	Grumpy Bear	Blue
3	Sleepy Bear	Yellow	Energy Bear
3	Sleepy Bear	Yellow
3	Sleepy Bear	Yellow		Bright Yellow

Desired Results:

sql_Key	name	color
1	Green Bear	Red
2	Grumpy Bear	Purple
3	Energy Bear	Bright Yellow

Answer 1

This works, though I feel like it could be simpler: for each field in each group, get the first non-null "updated" value, otherwise get the first non-null "original" value.

g = df.groupby('sql_Key')

pd.concat(
    [
        g[f'updated_{v}'].first().fillna(g[f'original_{v}'].first()).rename(v)
        for v in ['name', 'color']],
    axis=1)

                name          color
sql_Key                            
1         Green Bear            Red
2        Grumpy Bear         Purple
3        Energy Bear  Bright Yellow