CODEWITHSUNDEEP
pythoncsvnumpy
maximus
maximus

Reputation: 608

Numpy : read data from CSV having numerals as string

I'm reading a .csv file in python using command as:

data = np.genfromtxt('home_data.csv', dtype=float, delimiter=',', names=True)

this csv has one column with zipcode which are numerals but in string format, for eg "85281". This column has values as nan:

data['zipcode']
Output : array([ nan,  nan,  nan, ...,  nan,  nan,  nan])

How can I convert these values in string to integers so as to get an array of values and not of 'nan's.

Upvotes: 3

Views: 1861

Answers (2)

B. M.
B. M.

Reputation: 18628

you must help genfromtxt a little :

 data = np.genfromtxt('home_data.csv',
 dtype=[int,float],delimiter=',',names=True,
 converters={0: lambda b:(b.decode().strip('"'))})

each field is collected as bytes. float(b'1\n') return 1.0 , but float(b'"8210"') give an error. the converters option allow to define for each field (here field 0) a function to do the proper conversion, here converting in string(decode) and removing (strip) the trailing ".

If home_data.csv is :

zipcode,val
"8210",1
"8320",2
"14",3

you will obtain :

data -> array([(8210, 1.0), (8320, 2.0), (14, 3.0)], dtype=[('zipcode', '<i4'), ('val', '<f8')])
data['zipcode'] -> array([8210, 8320,   14])

Upvotes: 1

userE
userE

Reputation: 249

Maybe not the most efficient solution, but read your data as string and convert it afterwards to float:

data = np.genfromtxt('home_data.csv', dtype=float, delimiter=',', names=True)


zipcode = data['zipcode'].astype(np.float)

Btw., is there a reason you want to save a zipcode as a float?

Upvotes: 1

Related Questions