BUG: Fix infinite recursion loop when pivot of IntervalTree is ±inf #46664
Conversation
johannes-mueller
force-pushed
the
46658-bugfix-inf-pivot-infinite-recursion
branch
2 times, most recently
from
8a9514b to
c82d0fb
Compare
|
looks good. if you make non-draft can merge. |
johannes-mueller
force-pushed
the
46658-bugfix-inf-pivot-infinite-recursion
branch
from
c82d0fb to
c0ba6cf
Compare
Please also see my comment to the issue. I am in particular stuck with what to do about the CI failure. |
johannes-mueller
force-pushed
the
46658-bugfix-inf-pivot-infinite-recursion
branch
2 times, most recently
from
93d50c4 to
4a11347
Compare
|
This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this. |
johannes-mueller
force-pushed
the
46658-bugfix-inf-pivot-infinite-recursion
branch
2 times, most recently
from
de9e54a to
157d836
Compare
Skipping now the newly introduced tests on 32-bit like other tests of |
| @@ -207,3 +207,21 @@ def test_interval_tree_error_and_warning(self): | |||
| ): | |||
| left, right = np.arange(10), [np.iinfo(np.int64).max] * 10 | |||
| IntervalTree(left, right, closed="both") | |||
|
|
|||
| @pytest.mark.skipif(not IS64, reason="GH 23440") | |||
There was a problem hiding this comment.
Prefer xfail over skipif just in case this gets fixed in the future
There was a problem hiding this comment.
You mean like this?
@pytest.mark.xfail(not IS64, strict=False, reason="GH 23440")
That would be inconsistent with the other skipifs in the file. So maybe we should change all of them?
There was a problem hiding this comment.
Yeah, but without the strict=False (unless you know this test is specifically flaky and passes sometimes)
Generally we use xfail if a test has a chance to be fixed in the future and skipif if the test is never expected to pass/covered somewhere else.
Can change the skipif in another PR if it meets the above criteria for xfailing.
johannes-mueller
force-pushed
the
46658-bugfix-inf-pivot-infinite-recursion
branch
from
157d836 to
d792429
Compare
Attempts to fix pandas-dev#46658. When the pivot of an IntervalTree becomes ±inf the construction of the IntervalTree comes to an infinite loop recursion. This patch tries to fix that by catching those cases and set the pivot to a reasonable value. Note that the tests are skipped on 32-bit systems (see pandas-dev#23440)
johannes-mueller
force-pushed
the
46658-bugfix-inf-pivot-infinite-recursion
branch
from
d792429 to
f2c75c3
Compare
|
cc @mroeschke when you are good |
|
Thanks @johannes-mueller |
…andas-dev#46664) Attempts to fix pandas-dev#46658. When the pivot of an IntervalTree becomes ±inf the construction of the IntervalTree comes to an infinite loop recursion. This patch tries to fix that by catching those cases and set the pivot to a reasonable value. Note that the tests are skipped on 32-bit systems (see pandas-dev#23440)
When the pivot of an IntervalTree becomes ±inf the construction of the
IntervalTree comes to an infinite loop recursion. This patch tries to fix that
by catching those cases and set the pivot to a reasonable value.
doc/source/whatsnew/v1.5.0.rstfile if fixing a bug or adding a new feature.