File path name for NumPy's loadtxt()

Question

I was wondering if somebody had some information on how to load a CSV file using NumPy's loadtxt(). For some reason it claims that there is no such file or directory, when clearly there is. I've even copy/pasted the full path (with and without the leading / for root), but to no avail.

from numpy import *
FH = loadtxt("/Users/groenera/Desktop/file.csv")

or

from numpy import *
FH = loadtxt("Users/groenera/Desktop/file.csv")

The documentation for loadtxt is very unhelpful about this.

Solution 1

This is probably not a loadtxt problem. Try simply

f = open("/Users/groenera/Desktop/file.csv")

to make sure it is loadtxt's fault. Also, try using a Unicode string:

f = open(u"/Users/groenera/Desktop/file.csv")

Solution 2

You might have forgot the double slash, "//". Some machines require this.

So instead of

FH = loadtxt("/Users/groenera/Desktop/file.csv")

do this:

FH = loadtxt("C:\\Users\\groenera\\Desktop\\file.csv")

Comments

[DSM]

Unrelated, but: from numpy import * -- don't do this. It'll lead to very hard-to-diagnose bugs in your code because many builtin functions will be shadowed by similarly-named numpy versions with dangerously different behaviour in certain cases. Either import functions explicitly from numpy import loadtxt or use the standard abbreviation, import numpy as np and then np.loadtxt.

[rajpy]

Can you copy 'file.csv' file current working directory and try out. From the documentation, I don't think so you can use file path.

[Warren Weckesser]

Show the result of the following when run in a terminal: cd /Users/groenera/Desktop; pwd; ls -l

[astromax]

@DSM Good advice about the potential namespace conflicts.

[astromax]

@rajpy I've done so and it works when the script and the csv file are in the same directory, but this is not what I want. The reason why I'm so insistent on it being in a different directory is because my script outputs data files of its own which need to be separate from the location I've placed my data files.

[astromax]

@Warren Weckesser That's the first thing I tried. The directory exists and the files are there.

[Warren Weckesser]

"Trust, but verify." I understood that's what you said; I just wanted to see it for myself. Have you tried the suggestion in @askewchan's answer?

[astromax]

I just tried it. Although my problem was that I'm importing many modules in a bad way (using from MODULE import *). I believe there is a loadtxt namespace clash, so what fixed it for me was to do: import numpy as np (then make sure I'm using np.loadtxt() explicitly).

[Warren Weckesser]

scipy's loadtxt (i.e. scipy.loadtxt) is just an alias for numpy.loadtxt (i.e. they are the same thing), so I don't see how using the name from scipy would have caused the problem.

[astromax]

The open command seems to work. On a completely different note, though, importing numpy loadtxt like so: import numpy as np f = np.loadtxt('/Users/groenera/Desktop/file.csv') seemed to do the trick. As far as I knew there wasn't a function in the standard library called loadtxt. I think the problem was that SciPy has a loadtxt function, and since I was importing everything from both I wasn't paying attention to which one was being called (and assuming that it was NumPy's version). For the future is there a way to test where a function is being called from in a situation like this?

[askewchan]

@astromax scipy.loadtxt is numpy.loadtxt, so I'd be very surprised if that was the problem (scipy imports numpy functions and provides them). You can see what file it comes from with <function>.__module__. So, for me np.loadtxt.__module__ returns 'numpy.lib.npyio' and so does scipy.loadtxt.__module__

[astromax]

Uh oh. Well, it's entirely possible that I just had an error in the path (it is a pretty long path). But from what I remember importing numpy differently is what fixed it for me. Thanks for all of the help, though!

[askewchan]

@astromax Hm.. It's also possible that some other system state was screwed up somehow, and the good old restart/reload is what fixed it.