splitting at underscore in python and storing the first value

I have a pandas data frame like df with a column construct_name

construct_name
aaaa_t1_2
cccc_t4_10
bbbb_g3_3

and so on. I want to first split all the names at the underscore and store the first element (aaaa,cccc, etc.) as another column name.

Expected output

construct_name name
aaaa_t1_2 aaaa
cccc_t4_10 bbbb

and so on.

I tried the followingdf['construct_name'].map(lambda row:row.split("_")) and it gives me a list like

[aaaa,t1,2]
[cccc,t4,10]

and so on

But when I do

df['construct_name'].map(lambda row:row.split("_"))[0] to get the first element of the list I get an error. Can you suggest a fix. Thanks

5 Answers

Just use the vectorised str method split and use integer indexing on the list to get the first element:

In [228]:
df['first'] = df['construct_name'].str.split('_').str[0]
df
Out[228]: construct_name first
0 aaaa_t1_2 aaaa
1 cccc_t4_10 cccc
2 bbbb_g3_3 bbbb

df['name'] = df['construct_name'].str.split('_').str.get(0)

or

df['name'] = df['construct_name'].str.split('_').apply(lambda x: x[0])

After you do the split, you should get the first element (using [0]). And not after the map.:

In [608]: temp['name'] = temp['construct_name'].map(lambda v: v.split('_')[0])
In [609]: temp
Out[609]: construct_name name
0 aaaa_t1_2 aaaa
1 cccc_t4_10 cccc
2 bbbb_g3_3 bbbb

Another way of using the vectorised str.split method is passing the expand=True flag which then returns one column for each of the split parts.

>>> s = pd.Series( ['aaaa_t1_2', 'cccc_t4_10', 'bbbb_g3_3'], name='construct_name')
>>> s.str.split('_', expand=True) # to see what expand=True does 0 1 2
0 aaaa t1 2
1 cccc t4 10
2 bbbb g3 3
>>> s.str.split('_', expand=True)[0] # what you want, select first elements
0 aaaa
1 cccc
2 bbbb

This would be specially useful if you wanted to keep the first and second values for example.

In terms of the general behaviour of the expand=True flag, note that if the input strings do not have the same number of underscores you ca get Nones:

>>> s = pd.Series( ['aaaa_t1_2', 'cccc_t4', 'bbbb_g33'], name='construct_name')
>>> s.str.split('_', expand=True) 0 1 2
0 aaaa t1 2
1 cccc t4 None
2 bbbb g33 None

split take an optional argument maxsplit:

>>> construct_name = 'aaaa_t1_2'
>>> name, rest = construct_name.split('_', 1)
>>> name
'aaaa'