[Tutor] unitest with random inputs

[Steven D'Aprano]

On Wed, Jul 19, 2017 at 06:08:57PM +0200, Sydney Shall wrote:

> For the second error, regarding 'nan' finding 'nan', I received the 
> following output.
> 
> FAIL: test_nan_in_capitalsadvanced (__main__.TestPOC)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File 
> "/Users/sydney/Capital/Capital_with_productivity/Current_Versions/testPOCWP_V2.py", 
> line 323, in test_nan_in_capitalsadvanced
>     self.assertIn('nan', self.capitalsadvanced)
> AssertionError: 'nan' not found in array([ 12026.72641072, 
> 12434.53700436,  12185.56314119, ...,
>         12488.04714281,  12479.4662866 ,  12310.66016998])

NANs are tricky to work with.

To check whether a list contains a NAN value, you cannot reliably use 
"nan in list". How do we fix that?

To begin with, we need to use an actual floating point NAN, not the 
string "nan". So we can try this:

float('nan') in list

Alas, that's not enough. As Peter discussed earlier, for technical 
reasons, "Not A Number" floating point values (NANs) are designed to 
always fail equality tests. And for other technical reasons, Python 
tests for object identity before equality tests when evaluating the "in" 
operator. Don't worry if you don't understand the gory details yet: the 
message to take home is that NANs are special, and you cannot *reliably* 
test for them with the "in" operator.

So 

    self.assertIn(float('nan'), self.capitalsadvanced)


*will not work* reliably, if at all.

Instead, we can create our own test for a NAN:

    self.assertTrue(any(math.isnan(x) for x in self.capitalsadvanced))


Make sure you run "import math" at the top of your module.

(I haven't tested that code myself, so please try it, and if it doesn't 
work for some reason, let us know on the mailing list and somebody can 
fix it.)

But... I'm rather surprised that you need this test. Are you sure that 
your array capitalsadvanced will *always* contain at least one Not A 
Number value? Unless you put one in yourself, NANs generally indicate 
that a mathematical error has occurred somewhere.




-- 
Steve