Best to use an 'assert' command?

On 31 Jul 2015 8:41 pm, Jon Herman [email protected] wrote:

currently required to be a numpy array.

Could add a simple check at the beginning of every method, if type(Y) is list: Y = np.array(Y)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/70.

Hey Will. Does assert always need to throw an error when the assertion fails? In this case I think it'd be better to silently convert to np.array.

I don't think there is any issue with the methods failing if they're not called with the correct type arguments. An assert command would then raise the appropriate error (TypeError), and warn the user. Generally, it is best to fail early and gracefully than try to automatically cope for every possible permutation of types the procedure called be called with.

Hey Will. I originally brought this issue up with Jon, as I ran into errors trying to pass in a list. I'd be fine with the software failing, but it should produce a better error message. Below is the current error message:

Traceback (most recent call last):
  File "C:\Program Files (x86)\Python27\lib\site-packages\salib-e315cb0508ded16e92aba9a6c42b1fea41e78e90-py2.7.egg\SALib\analyze\sobol.py", line 21, in analyze
    if calc_second_order and Y.size % (2 * D   2) == 0:
AttributeError: 'list' object has no attribute 'size'

Thanks Dave. I don't have a preference either way, as long as all of the methods are modified consistently. (Maybe this falls under the general need for improved error messages).

Something I noticed, numpy will always cast a list to a np-array if you pass it into a function instead of using the object attributes. So for example:

>>> import numpy as np
>>> A = [2, 3, 4, 5]
>>> np.size(A)
4
>>> A.size
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'list' object has no attribute 'size'
>>>

So maybe the easiest thing to do is switch to np.* functions instead of object attributes throughout.

Hmm. I understand now. Or we could just create a 'data cleaning' module for when the command-line interface is used. Then the data will be coerced to the correct type before the functions are called.