Reading CSV File using Numpy

Question

from numpy import genfromtxt
dataPoints = genfromtxt(temp.csv, dtype='datetime64[D],i8',delimiter=' ')

CSV File: temp.csv

2014-05-19 10
2014-05-20 11
2014-05-21 12

Output from print(dataPoints)

[ nan  10.]
[ nan  11.]
[ nan  12.]

Edit

[1969-12-31 00:00:00 1970-01-11 00:00:00]
[1969-12-31 00:00:00 1970-01-12 00:00:00]
[1969-12-31 00:00:00 1970-01-13 00:00:00]
[1969-12-31 00:00:00 1970-01-14 00:00:00]
[1969-12-31 00:00:00 1970-01-13 00:00:00]

perhaps because a date is not a number

daouzli
– daouzli

2014-05-27 22:02:14 +00:00
Commented May 27, 2014 at 22:02 — daouzli
– daouzli, Commented May 27, 2014 at 22:02
@daouzli who said they all had to be numbers?

Apollo
– Apollo

2014-05-27 22:02:48 +00:00
Commented May 27, 2014 at 22:02 — Apollo
– Apollo, Commented May 27, 2014 at 22:02
the dtype default value in the docs

daouzli
– daouzli

2014-05-27 22:04:37 +00:00
Commented May 27, 2014 at 22:04 — daouzli
– daouzli, Commented May 27, 2014 at 22:04

Ricardo Cárdenes · Accepted Answer · 2014-05-27 22:06:04Z

4

Just tell NumPy what types to expect. E.g.:

>>> genfromtxt('temp.csv', dtype='datetime64[D],i8')
array([(datetime.date(2014, 5, 19), 10L),
       (datetime.date(2014, 5, 20), 11L), (datetime.date(2014, 5, 21), 12L)], 
      dtype=[('f0', '<M8[D]'), ('f1', '<i8')])

answered May 27, 2014 at 22:06

Ricardo Cárdenes

9,1941 gold badge23 silver badges35 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Ricardo Cárdenes Over a year ago

As for "why" are you getting nan instead of the proper type... NumPy should figure out the types for each column on its own, but dates don't seem to be on their coercion list and it's probably getting confused by the fact that your string starts on a number.

Apollo Over a year ago

after using your dtype I get the output in my edit above. This doesn't look right. Did I do something wrong?

Ricardo Cárdenes Over a year ago

Maybe a different version of NumPy? Anyway, do you want to operate with the dates, or just read them? If you're happy with a plain string you can just use "S10,i8" instead. No need for delimiter

Ricardo Cárdenes Over a year ago

Also, i8 may not be suitable for you. You may need to specify a larger integer

Apollo Over a year ago

I would like to work with dates, so some kind of datetime would probably be ideal. How do I check the version of NumPy I'm using?

Collectives™ on Stack Overflow

Reading CSV File using Numpy

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related