python - ValueError when converting string to integer in Dataframe -

i trying replace strings in years column of dataframe below numbers in string. example, change zc025yr 025. code follows:

import urllib, urllib2 import csv stringio import stringio import pandas pd import os zipfile import zipfile pprint import pprint, pformat  my_url = 'http://www.bankofcanada.ca/stats/results/csv' data = urllib.urlencode({"lookuppage": "lookup_yield_curve.php",                          "startrange": "1986-01-01",                          "searchrange": "all"}) request = urllib2.request(my_url, data) result = urllib2.urlopen(request) zipdata = result.read() zipfile = zipfile(stringio(zipdata))  df = pd.read_csv(zipfile.open(zipfile.namelist()[0]))  df = pd.melt(df, id_vars=['date'])  df.rename(columns={'variable': 'years'}, inplace=true) 

the dataframe have looks this:

              date     years          value 0       1986-01-01   zc025yr             na 1       1986-01-02   zc025yr   0.0948511020 2       1986-01-03   zc025yr   0.0972953210 3       1986-01-06   zc025yr   0.0965403640 ..... 

however, if add code below in order restructure dataframe error valueerror: cannot convert float nan integer in line df['years'] = df['years'].str.extract('(\d+)').astype(int) strange because when @ year's data in csv file don't see there being 'nan' associated it.

#converting strings in column number of years df['years'] = df['years'].str.extract('(\d+)').astype(int) df['years'] = df.years/100 

thank you

try creating new function convert strings integer , call in series.apply method follows -

edit: adding logic default empty strings 0 , use different value if want handle empty strings in years colomn differently

import re def getyear(s):     x = re.search('(\d+)',s)     return int(x.groups()[0]) if x not none else 0 # or want handle 

then use function -

df['years'] = df['years'].apply(getyear)