Better PayPal penny testing

[rummik]

Should solve #2198

Edit: Please note, I don't know how to test this. I'll leave that up to @whit537

[rummik]

Ping @whit537. I don't know who else deals with PayPal right now.

[chadwhitacre]

IRC

Here's why I took out the test for 0.75.

[chadwhitacre]

The bad amount is 26.26. We compute a net of 25.74 and a fee of 0.52

I am unable to reproduce this result. Right now I am seeing masspay.py compute 25.75 given 26.26. I think we indeed need to back out ba2b611 (or the functional equivalent).

[chadwhitacre]

No wait ...

[chadwhitacre]

No, actually, yes ... ?

[chadwhitacre]

Yeesh. Now I am seeing masspay.py compute take 26.26 and compute 25.74, with PayPal assessing a fee of 0.51 as expected:

[chadwhitacre]

Here's the Python script I was using:

#!/usr/bin/env python

from decimal import Decimal as D


def round_(d):
    return d.quantize(D('0.01'))


for dollars in range(25,27):
    for cents in range(100):
        target = D('{}.{:0>2}'.format(dollars, cents))
        exact = target / D('1.02')
        inexact = round_(exact)
        check_exact = inexact * D('1.02')
        check_rounded = round_(check_exact)
        if check_rounded != target:
            print target, check_rounded, check_exact, inexact, exact

Here's a similar JavaScript script that @rummik came up with:

http://jsbin.com/rujoc/2

[chadwhitacre]

I rebased this on masspay-report (#2099) and modified it to go back to the previous test we had, prior to #2029.