I'm developing a program that makes some floating points calculations. Is there any way to test my functions (which deliver floats) with doctests?
5 Answers
Sure, just format the floats with a reasonable format, based on your knowledge of what precision you expect them to exhibit -- e.g, if you expect accuracy to 2 digits after the decimal point, you could use:
''' Rest of your docstring and then...
>>> '%.2f' % funcreturningfloat()
'123.45'
'''

- 854,459
- 170
- 1,222
- 1,395
The following works for nosetests:
>>> 1/3. # doctest: +ELLIPSIS
0.333...

- 355,277
- 75
- 744
- 836
-
1This works for repeating decimals (because their most significant digits aren't affected by floating point error), but not in general. For example, ">>> 1 - 10e-17" returns 0.9999999999..., which wouldn't be matched if you test for "1.0..." – SigmaX Apr 13 '19 at 00:21
The documentation has a suggestion
Floating-point numbers are also subject to small output variations across platforms, because Python defers to the platform C library for float formatting, and C libraries vary widely in quality here.
>>> 1./7 # risky
0.14285714285714285
>>> print 1./7 # safer
0.142857142857
>>> print round(1./7, 6) # much safer
0.142857

- 295,403
- 53
- 369
- 502
You can use numtest - a doctest extension that simplifies the test of numerical results. https://pypi.python.org/pypi/numtest
>>> 1.0/3
0.333
Failed example: 1.0/3 Expected: 0.333 Got: 0.3333333333333333
>>> 1.0/3 # doctest: +NUMBER
0.333
import doctest
import numtest
doctest.testmod()
Process finished with exit code 0
All tests passed. No need for string formatting in your tests.

- 7,256
- 3
- 42
- 45
String format allows use test tuples.
>>> funcreturningfloattuple(1.0)
(1.0, 1.0)
>>> '%.2f, %.2f' % funcreturningfloattuple(1.0)
'1.00, 1.00'
>>> funcreturningfloattuple(1.4)
(1.3999999999999999, 1.3999999999999999)
>>> '%.2f, %.2f' % funcreturningfloattuple(1.4)
'1.40, 1.40'

- 131
- 6