2

The question is probaly extremely dumb, but i hurt my brain figuring out what to do

There is a pd.dataframe with N columns. I need to select some columns, referring by index of a column, then convert all values to numeric and rewrite that column in my dataframe

I've done it by column name reference (like df['a'] = pd.to_numeric(df['a']) but stuck with indices (like df[1] = pd.to_numeric(df[1])

What is the right way in this situation to dataframe column referencing? (python 2.7)

jezrael
  • 822,522
  • 95
  • 1,334
  • 1,252
Sasha Korekov
  • 323
  • 3
  • 16

2 Answers2

3

You can use ix for selecting columns and then apply to_numeric:

import pandas as pd

df = pd.DataFrame({1:['1','2','3'],
                   2:[4,5,6],
                   3:[7,8,9],
                   4:['1','3','5'],
                   5:[5,3,6],
                   6:['7','4','3']})

print (df)
   1  2  3  4  5  6
0  1  4  7  1  5  7
1  2  5  8  3  3  4
2  3  6  9  5  6  3

print (df.dtypes)
1    object
2     int64
3     int64
4    object
5     int64
6    object
dtype: object

print (df.columns)
Int64Index([1, 2, 3, 4, 5, 6], dtype='int64')

cols = [1,4,6]    
df.ix[:, cols] = df.ix[:, cols].apply(pd.to_numeric)

print (df)
   1  2  3  4  5  6
0  1  4  7  1  5  7
1  2  5  8  3  3  4
2  3  6  9  5  6  3

print (df.dtypes)
1    int64
2    int64
3    int64
4    int64
5    int64
6    int64
dtype: object

If columns are strings, not int (but it looks like int) add '' to numbers in list cols:

import pandas as pd

df = pd.DataFrame({'1':['1','2','3'],
                   '2':[4,5,6],
                   '3':[7,8,9],
                   '4':['1','3','5'],
                   '5':[5,3,6],
                   '6':['7','4','3']})

#print (df)

#print (df.dtypes)

print (df.columns)
Index(['1', '2', '3', '4', '5', '6'], dtype='object')

#add `''`
cols = ['1','4','6']
#1. ix: supports mixed integer and label based access     
df.ix[:, cols] = df.ix[:, cols].apply(pd.to_numeric)

#2. loc: only label based access
# df.loc[:, cols] = df.loc[:, cols].apply(pd.to_numeric)

#3. iloc: for index based access
# cols = [i for i in range(len(df.columns))]
# df.iloc[:, cols].apply(pd.to_numeric)

print (df)
   1  2  3  4  5  6
0  1  4  7  1  5  7
1  2  5  8  3  3  4
2  3  6  9  5  6  3

print (df.dtypes)
1    int64
2    int64
3    int64
4    int64
5    int64
6    int64
dtype: object
Pramit
  • 1,373
  • 1
  • 18
  • 27
jezrael
  • 822,522
  • 95
  • 1,334
  • 1,252
  • Thank you, it looks good. But is there a way to reference column with non-int header with int? I mean, referencing 6th column in `Index(['1', '2', '3', '4', '5', 'k'], dtype='object')` with `cols = [6]` ? – Sasha Korekov Jun 16 '16 at 02:45
  • Yes, you can use [`iloc`](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.iloc.html) - so 6. columns is `df.iloc[:,5]` (python count from 0) – jezrael Jun 16 '16 at 05:27
0

You may want to check the following post out Is .ix() always better than .loc() and .iloc() since it is faster and supports integer and label access?

A must [Different choice of Indexing] (http://pandas.pydata.org/pandas-docs/stable/indexing.html#different-choices-for-indexing)

Community
  • 1
  • 1
Pramit
  • 1,373
  • 1
  • 18
  • 27