TST: Update unreliable num precision component of test_binary_arith_ops (GH37328) #37380
Conversation
|
Seems like a numerical precision issue indeed. Comparing the values: 16.964677878827604 and 16.964677878827608 I get the following tolerance levels: Therefore I changed: |
avinashpancham
force-pushed
the
GH37328
branch
from
83d584d to
cc04f06
Compare
|
I see that this issue is not limited to those values, I also get this error on #36838, where there's a mismatch between 0.8129387113107932 and 0.8129387113107933 The proposed change would catch it, but maybe we will need to relax the tolerances more to make it generic. |
|
@avinashpancham Merging master should fix the linting error |
|
xref #13339 which changed the first assert in |
|
@rhshadrach sorry but I dont understand what you mean. Could you maybe give some more explanation? |
|
@dsaxton thanks that solved it indeed |
Now that I read it again it looks more like a general comment, instead of feedback. My bad. |
|
CI failed due to: |
restarted. |
|
cc @jbrockmendel this seems reasonble to me. |
| # direct numpy comparison | ||
| expected = self.ne.evaluate(f"nlhs {op} ghs") | ||
| tm.assert_numpy_array_equal(result.values, expected) | ||
| tm.assert_almost_equal(result.values, expected) |
There was a problem hiding this comment.
can you add a comment why we are doing this and reference to this issue
There was a problem hiding this comment.
yah i think this makes sense, but should have a TODO indicating that it should be equality and we had to punt
|
@jreback merged master and CI is greenish. Errors are memory/OS errors unrelated to this change, do we want to run this again or is this fine? |
|
thanks @avinashpancham |
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff