python – How do I read CSV data into a record array in NumPy?

The Question :

448 people think this question is useful

I wonder if there is a direct way to import the contents of a CSV file into a record array, much in the way that R’s read.table(), read.delim(), and read.csv() family imports data to R’s data frame?

Or is the best way to use csv.reader() and then apply something like numpy.core.records.fromrecords()?

The Question Comments :

The Answer 1

705 people think this answer is useful

You can use Numpy’s genfromtxt() method to do so, by setting the delimiter kwarg to a comma.

from numpy import genfromtxt
my_data = genfromtxt('my_file.csv', delimiter=',')

More information on the function can be found at its respective documentation.

The Answer 2

198 people think this answer is useful

I would recommend the read_csv function from the pandas library:

import pandas as pd
df=pd.read_csv('myfile.csv', sep=',',header=None)
array([[ 1. ,  2. ,  3. ],
       [ 4. ,  5.5,  6. ]])

This gives a pandas DataFrame – allowing many useful data manipulation functions which are not directly available with numpy record arrays.

DataFrame is a 2-dimensional labeled data structure with columns of potentially different types. You can think of it like a spreadsheet or SQL table…

I would also recommend genfromtxt. However, since the question asks for a record array, as opposed to a normal array, the dtype=None parameter needs to be added to the genfromtxt call:

Given an input file, myfile.csv:

1.0, 2, 3
4, 5.5, 6

import numpy as np

gives an array:

array([[ 1. ,  2. ,  3. ],
       [ 4. ,  5.5,  6. ]])



gives a record array:

array([(1.0, 2.0, 3), (4.0, 5.5, 6)], 
      dtype=[('f0', '<f8'), ('f1', '<f8'), ('f2', '<i4')])

This has the advantage that file with multiple data types (including strings) can be easily imported.

The Answer 3

80 people think this answer is useful

I timed the

from numpy import genfromtxt
genfromtxt(fname = dest_file, dtype = (<whatever options>))


import csv
import numpy as np
with open(dest_file,'r') as dest_f:
    data_iter = csv.reader(dest_f,
                           delimiter = delimiter,
                           quotechar = '"')
    data = [data for data in data_iter]
data_array = np.asarray(data, dtype = <whatever options>)

on 4.6 million rows with about 70 columns and found that the NumPy path took 2 min 16 secs and the csv-list comprehension method took 13 seconds.

I would recommend the csv-list comprehension method as it is most likely relies on pre-compiled libraries and not the interpreter as much as NumPy. I suspect the pandas method would have similar interpreter overhead.

The Answer 4

67 people think this answer is useful

You can also try recfromcsv() which can guess data types and return a properly formatted record array.

The Answer 5

17 people think this answer is useful

As I tried both ways using NumPy and Pandas, using pandas has a lot of advantages:

  • Faster
  • Less CPU usage
  • 1/3 RAM usage compared to NumPy genfromtxt

This is my test code:

$ for f in ; do  /usr/bin/time python $f; done
2.94user 0.41system 0:03.05elapsed 109%CPU (0avgtext+0avgdata 502068maxresident)k
0inputs+24outputs (0major+107147minor)pagefaults 0swaps

23.29user 0.72system 0:23.72elapsed 101%CPU (0avgtext+0avgdata 1680888maxresident)k
0inputs+0outputs (0major+416145minor)pagefaults 0swaps

from numpy import genfromtxt
train = genfromtxt('/home/hvn/me/notebook/train.csv', delimiter=',')

from pandas import read_csv
df = read_csv('/home/hvn/me/notebook/train.csv')

Data file:

du -h ~/me/notebook/train.csv
 59M    /home/hvn/me/notebook/train.csv

With NumPy and pandas at versions:

$ pip freeze | egrep -i 'pandas|numpy'

The Answer 6

7 people think this answer is useful

You can use this code to send CSV file data into an array:

import numpy as np
csv = np.genfromtxt('test.csv', delimiter=",")

The Answer 7

5 people think this answer is useful

I would suggest using tables (pip3 install tables). You can save your .csv file to .h5 using pandas (pip3 install pandas),

import pandas as pd
data = pd.read_csv("dataset.csv")
store = pd.HDFStore('dataset.h5')
store['mydata'] = data

You can then easily, and with less time even for huge amount of data, load your data in a NumPy array.

import pandas as pd
store = pd.HDFStore('dataset.h5')
data = store['mydata']

# Data in NumPy format
data = data.values

The Answer 8

5 people think this answer is useful

This is the easiest way:

import csv
with open('testfile.csv', newline='') as csvfile:
    data = list(csv.reader(csvfile))

Now each entry in data is a record, represented as an array. So you have a 2D array. It saved me so much time.

The Answer 9

4 people think this answer is useful

I tried this:

import pandas as p
import numpy as n

closingValue = p.read_csv("<FILENAME>", usecols=[4], dtype=float)

The Answer 10

4 people think this answer is useful

Using numpy.loadtxt

A quite simple method. But it requires all the elements being float (int and so on)

import numpy as np 
data = np.loadtxt('c:\\1.csv',delimiter=',',skiprows=0)  

The Answer 11

3 people think this answer is useful

This work as a charm…

import csv
with open("data.csv", 'r') as f:
    data = list(csv.reader(f, delimiter=";"))

import numpy as np
data = np.array(data, dtype=np.float)

The Answer 12

0 people think this answer is useful
In [329]: %time my_data = genfromtxt('one.csv', delimiter=',')
CPU times: user 19.8 s, sys: 4.58 s, total: 24.4 s
Wall time: 24.4 s

In [330]: %time df = pd.read_csv("one.csv", skiprows=20)
CPU times: user 1.06 s, sys: 312 ms, total: 1.38 s
Wall time: 1.38 s

Add a Comment