eleanora eleanora - 4 months ago 32
Python Question

Data munging in pandas

I have a CSV file with lines look like:

ID,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#,


I can read it in with

#!/usr/bin/env python

import pandas as pd
import sys

filename = sys.argv[1]
df = pd.read_csv(filename)


Given a particular column, I would like to split the rows by ID and then output the mean and standard deviation for each ID.

My first problem is, how can I remove all the non-numeric parts from the numbers such as "100M" and "0N#" which should be 100 and 0 respectively.

I also tried looping over the relevant headers and using

df[header].replace(regex=True,inplace=True,to_replace=r'\D',value=r'')


as suggested in Pandas DataFrame: remove unwanted parts from strings in a column .

However this changes 98.4 into 984.

Answer

use str.extract:

In [356]:
import io
import pandas as pd
t="""ID,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#"""
df = pd.read_csv(io.StringIO(t), header=None)
df

Out[356]:
   0     1     2    3    4    5     6    7    8     9    10    11    12   13  \
0  ID  98.4  100M  55M  65M  75M  100M  75M  65M  100M  98M  100M  100M  92M   

   14   15  
0  0#  0N#  

In [357]:
for col in df.columns[2:]:
    df[col] = df[col].str.extract(r'(\d+)').astype(int)
df

Out[357]:
   0     1    2   3   4   5    6   7   8    9   10   11   12  13  14  15
0  ID  98.4  100  55  65  75  100  75  65  100  98  100  100  92   0   0

If you have float numbers then you can use the following regex:

In [379]:
t="""ID,98.4,100.50M,55.234M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#"""
df = pd.read_csv(io.StringIO(t), header=None)
df

Out[379]:
   0     1        2        3    4    5     6    7    8     9    10    11  \
0  ID  98.4  100.50M  55.234M  65M  75M  100M  75M  65M  100M  98M  100M   

     12   13  14   15  
0  100M  92M  0#  0N#  

In [380]:    
for col in df.columns[2:]:
    df[col] = df[col].str.extract(r'(\d+\.?\d+)').astype(np.float)
df

Out[380]:
   0     1      2       3   4   5    6   7   8    9   10   11   12  13  14  15
0  ID  98.4  100.5  55.234  65  75  100  75  65  100  98  100  100  92 NaN NaN

so (\d+\.?\d+) looks for groups containing \d+ 1 or more digits with \.? optional decimal point and \d+ 1 or more further digits after the decimal point

EDIT

OK edited my regex pattern:

In [408]:
t="""Name,97.7,0A,0A,65M,0A,100M,5M,75M,100M,90M,90M,99M,90M,0#,0N#"""
df = pd.read_csv(io.StringIO(t), header=None)
df

Out[408]:
     0     1   2   3    4   5     6   7    8     9    10   11   12   13  14  \
0  Name  97.7  0A  0A  65M  0A  100M  5M  75M  100M  90M  90M  99M  90M  0#   

    15  
0  0N#  

In [409]:    
for col in df.columns[2:]:
    df[col] = df[col].str.extract(r'(\d+\.*\d*)').astype(np.float)
df

Out[409]:
     0     1   2   3   4   5    6   7   8    9   10  11  12  13  14  15
0  Name  97.7   0   0  65   0  100   5  75  100  90  90  99  90   0   0