kendalltau function now returns a nan tuple if empty arrays used as para... #3885
Conversation
…arameters instead of raising a runtime exception
|
A regression test would be good. |
|
Should I do something? Or is that what Travis CI is for? |
|
TravisCI will run the tests to try catching regressions, but I was thinking that adding a new test would be appropriate. For example, a test could be added that would have caused a RuntimeError but no longer raises the exception after your change. |
|
Ah, that makes sense. |
|
Changes Unknown when pulling 54704c2 on ctokheim:master into * on scipy:master*. |
|
Changes Unknown when pulling 061a5a8 on ctokheim:master into * on scipy:master*. |
| @@ -633,6 +633,9 @@ def test_kendalltau(): | |||
| assert_(np.all(np.isnan(stats.kendalltau([2,0,2], [2,2,2])))) | |||
| assert_(np.all(np.isnan(stats.kendalltau([2,2,2], [2,0,2])))) | |||
|
|
|||
| # empty arrays provided as input | |||
| assert_(np.all(np.isnan(stats.kendalltau([], [])))) | |||
There was a problem hiding this comment.
assert_equal has code to handle nan, so this can be simplified to assert_equal(stats.kendalltau([], []), (np.nan, np.nan)).
There was a problem hiding this comment.
Sure, I changed the checking for both the tie case and empty array case to use assert_equal.
kendalltau function now returns a nan tuple if empty arrays used as para...
|
Looks good, merging. Thanks @ctokheim. |
...meters instead of raising a runtime exception. Previously, empty arrays would have cause a maximum recursion depth exceeded error in mergesort if the provided arrays were empty (issue #3884).