CODEWITHSUNDEEP
pythonpython-3.xpandas
Maria Nazari
Maria Nazari

Reputation: 690

Mapping Dictionary Values to Pandas Data-frame Column

I have this dictionary:

 j =  {1: {'help': 2},
 2: {'chocolate': 1, 'help': 1},
 3: {'chocolate': 1, 'help': 1}}

and this dataframe:

df = pd.DataFrame({'docId': [1, 2, 3, 1, 2, 3, ],
                       'sent': ['help', 'chocolate', 'chocolate', 'help', 'help', 'help']})

and I want to match the values according to docId and term so it should look like this:

docId  sent        freq
1      help         2
2      chocolate    1
3      chocolate    1
1      help         2
2      help         1
3      help         1

I'm not sure how to accomplish this, I was working on using map and apply but I didn't get anywhere.

Upvotes: 3

Views: 1014

Answers (3)

BENY
BENY

Reputation: 323226

IIUC try something different using reindex

s=pd.DataFrame(j).stack().reindex(pd.MultiIndex.from_arrays([df.sent,df.docId])).reset_index()
s
Out[81]: 
        sent  docId    0
0       help      1  2.0
1  chocolate      2  1.0
2  chocolate      3  1.0
3       help      1  2.0
4       help      2  1.0
5       help      3  1.0

End Up using this method lookup 

df['Freq']=pd.DataFrame(j).lookup(df.sent,df.docId)
df
Out[95]: 
   docId       sent  Freq
0      1       help   2.0
1      2  chocolate   1.0
2      3  chocolate   1.0
3      1       help   2.0
4      2       help   1.0
5      3       help   1.0

Upvotes: 4

piRSquared
piRSquared

Reputation: 294218

Remake your dictionary

With tuples as keys, you can map the get method over zipped columns

J = {(x, y): v for x, V in j.items() for y, v in V.items()}

df.assign(freq=[*map(J.get, zip(df.docId, df.sent))])


   docId       sent  freq
0      1       help     2
1      2  chocolate     1
2      3  chocolate     1
3      1       help     2
4      2       help     1
5      3       help     1

Or don't

You can use a lambda in map that takes two arguments and pass the iterables that supply the arguments.

df.assign(freq=[*map(lambda x, y: j[x][y], df.docId, df.sent)])

   docId       sent  freq
0      1       help     2
1      2  chocolate     1
2      3  chocolate     1
3      1       help     2
4      2       help     1
5      3       help     1

Upvotes: 6

cs95
cs95

Reputation: 402343

How about a list comprehension? You can chain two dict.get calls (one for each level of nesting).

df['freq'] = [
    j.get(x, {}).get(y, np.nan) for x, y in df[['docId', 'sent']].values]
df

   docId       sent  freq
0      1       help     2
1      2  chocolate     1
2      3  chocolate     1
3      1       help     2
4      2       help     1
5      3       help     1

If you can guarantee all entries exist in j, you can simplify the above to, 

df['freq'] = [j[x][y] for x, y in df[['docId', 'sent']].values]
df

   docId       sent  freq
0      1       help     2
1      2  chocolate     1
2      3  chocolate     1
3      1       help     2
4      2       help     1
5      3       help     1

Upvotes: 6

Related Questions