CI: temporary skip parquet tz test for pyarrow>=2.0.0 #37303
Conversation
jorisvandenbossche
added
IO Parquet
| @@ -765,6 +765,10 @@ def test_timestamp_nanoseconds(self, pa): | |||
| check_round_trip(df, pa, write_kwargs={"version": "2.0"}) | |||
|
|
|||
| def test_timezone_aware_index(self, pa, timezone_aware_date_list): | |||
There was a problem hiding this comment.
should be outside but temporarily is ok
note we actually do want to pin pyarrow in some older builds but i'll comment on the issues
|
is this not a genuine error https://travis-ci.org/github/pandas-dev/pandas/jobs/737659028 the s3 test |
|
Yes, I split a fix for the genuine failure into its own PR: #37304 |
Yes, indeed, we will want to pin some builds to pyarrow 1.0, and keep some at latest (separate issue, though, see also #35259 (comment) for some related discussion) |
|
Opened a dedicated issue for this (pinning pyarrow): #37306 |
|
1.1.x is not yet failing with pyarrow related failures. https://dev.azure.com/pandas-dev/pandas/_build/results?buildId=45669&view=results changing milestone here and will backport if/when ci starts to fail. |
test_timezone_aware_index was added in #36004 which is only on master |
|
Ah, yes, that's a good reason it is not failing on 1.1.x ;-) |
|
Does the 1.1.x branch have Travis? Because then #37304 might still need to be backported |
|
Hmm, it seems the Travis build on 1.1.x is failing with a conda solve issue |
|
ah, was just mentioning that over on the other issue. It last build on https://travis-ci.org/github/pandas-dev/pandas/builds/736681105 we have some failures that we may need to resolve if we can get it to build consistently. I started to investigate #36779 but my local setup was obviously incorrect. |
See #37286, this PR is not the actual fix, but at least ensures we don't have failing CI in other PRs (I prefer to skip the test instead of pinning pyarrow to <= 1.0, since all other tests are passing on 2.0, which is good to still run)