CODEWITHSUNDEEP
pythonpandasindexing
Christopher
Christopher

Reputation: 2232

Python/Pandas: Get index of item in a column

I have a Pandas dataframe(df) with following columns:

df["ids"]

0         18281483,1658391547
1           1268212,128064430
2                  1346542425
3  13591493,13123669,35938208

df["id"]

0      18281483
1       1268212
2    1346542425
3      13123669

I like to find out, in which order of "ids" the respective "id" can be found, and output the respective value in a new column "order". Following code was tried without success:

df["order"] = df["ids"].str.split(",").index(df["id"])

----------------------------------------------------------------------
TypeError: 'Int64Index' object is not callable

Is there a syntax error? I tried the split and index function with every row manually (by inserting the lists and string), and it worked.

Desired output:

df["order"]

0 0
1 0
2 0 
3 1

Upvotes: 0

Views: 265

Answers (3)

user3483203
user3483203

Reputation: 51165

Really shouldn't need to use apply here. On larger Dataframes it will be incredibly slow. Broadcasted comparison will work just fine.

(df["ids"].str.split(",", expand=True) == df["id"][:, None]).idxmax(1)

0    0
1    0
2    0
3    1
dtype: int64

Performance

d = {'ids': {0: '18281483,1658391547',
             1: '1268212,128064430',
             2: '1346542425',
             3: '13591493,13123669,35938208'},
      'id': {0: '18281483', 
             1: '1268212', 
             2: '1346542425',
             3: '13123669'}}

df = pd.DataFrame(d)
df = pd.concat([df] * 1000)

%timeit (df["ids"].str.split(",", expand=True) == df["id"][:, None]).idxmax(1)                 
7.51 ms ± 61.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

%timeit df.apply(lambda x: x['ids'].split(',').index(x['id']), axis=1)                         
54.1 ms ± 249 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

Upvotes: 0

Quang Hoang
Quang Hoang

Reputation: 150735

Try:

df['output'] = df.astype(str).apply(lambda x: x['ids'].split(',').index(x['id']), axis=1)

Output:

                          ids          id  output
0         18281483,1658391547    18281483       0
1           1268212,128064430     1268212       0
2                  1346542425  1346542425       0
3  13591493,13123669,35938208    13123669       1

Upvotes: 1

sushanth
sushanth

Reputation: 8302

Here is a approach,

def index_(ids, id):
    split_ = ids.split(",")
    if id in split_:
        return split_.index(id)
    else:
        return -1


print(
    df.assign(id = df1.id.astype(str))
        .apply(lambda x: index_(x.ids, x.id), axis=1)
)

0    0
1    0
2    0
3    1
dtype: int64

Upvotes: 1

Related Questions