CODEWITHSUNDEEP
pythonpython-3.xpandasdataframe
wanaryytel
wanaryytel

Reputation: 3482

Populating Pandas dataframe row with a tuple of data using apply

I have a really simple dataframe for testing purposes. It looks like this: 

movieId | title     | genres              | Drama | Action | Comedy
1       | Toy Story | {'Drama', 'Comedy'} | 0     | 0      | 0

I want to reflect the set genres in booleans in the respective columns, so the desired result would be: 

movieId | title     | genres              | Drama | Action | Comedy
1       | Toy Story | {'Drama', 'Comedy'} | 1     | 0      | 1

So I tried this code with apply: 

def ttb(genreset):                                                
    return tuple(1 if g in genreset else 0 for g in all_genres)

all_genres = ('Drama', 'Action', 'Comedy')

df.T.loc[all_genres, :] = df.apply(lambda x: ttb(x.loc['genres']), axis=1)

But this resulted in an error that I can't really wrap my head around:

ValueError: shape mismatch: value array of shape (19,) could not be broadcast to indexing result of shape (19,1)

Do I need to somehow cast the return value of apply to have a fixed size or why doesn't it work as I would expect? I tried with more data as well, but always got the same error. Googling for the error gave lots of results, but offered no viable solutions for me.

Upvotes: 1

Views: 68

Answers (1)

cs95
cs95

Reputation: 402553

Call str.join followed by str.get_dummies:

v = df.genres.str.join(',').str.get_dummies(sep=',')

Or, if "Action" needs to be added explicitly, let's use reindex:

v = (df['genre']
     .str.join(',')
     .str.get_dummies(sep=',')
     .reindex(
         ['Comedy', 'Action', 'Drama'], 
         axis=1, 
         fill_value=0
     )
)

print(v)
   Comedy  Action  Drama
0       1       0      1

If you have many unique values and you're not sure what they are, you can always find their union:

u = set().union(*df.genres.tolist())

And now, use u to reindex the result.

If you need to add this back to your original DataFrame, use concat:

df = pd.concat([df, v], axis=1)

Upvotes: 3

Related Questions