BUG: handle outofbounds datetimes in DatetimeConverter #13801
Conversation
Current coverage is 85.23% (diff: 100%)@@ master #13801 diff @@
==========================================
Files 140 140
Lines 50415 50415
Methods 0 0
Messages 0 0
Branches 0 0
==========================================
+ Hits 42970 42972 +2
+ Misses 7445 7443 -2
Partials 0 0
|
| @@ -1222,6 +1222,14 @@ def test_secondary_y_irregular_ts_xlim(self): | |||
| self.assertEqual(left, ts_irregular.index.min().toordinal()) | |||
| self.assertEqual(right, ts_irregular.index.max().toordinal()) | |||
|
|
|||
| def test_plot_outofbounds_datetime(self): | |||
There was a problem hiding this comment.
Does pandas have the equivalent of @cleanup from the mpl tests? This is touching global state and can leak to other tests.
There was a problem hiding this comment.
In TestPlotBase (from which this testcase inherits), all settings are set to the defaul in the setup, and the figures closed in the teardown.
Or are there still other things that may leak?
|
working fine ;). No conflict with matplotlib so far in my usage. thanks for that! |
|
@sinhrks @TomAugspurger OK to merge? |
| @@ -216,7 +216,7 @@ def try_parse(values): | |||
| else: | |||
| values = [_dt_to_float_ordinal(x) for x in values] | |||
| except Exception: | |||
| pass | |||
| values = _dt_to_float_ordinal(values) | |||
There was a problem hiding this comment.
Allow me I do not understand converter dispatch logic.
Might be unrelated to the PR, but is it possible that input can have a type not registered for DatetimeConverter like below condition? If so, it may better to box the above with try-except to return input as it is.
There was a problem hiding this comment.
I am not completely sure, but by letting the DatetimeConverter accept for example strings (although string type is not registered), this makes it possible to do ax.set_xlim('2012-01-01', '2012-01-10')
|
@sinhrks further comments? If not, be my guest to merge :-) |
|
thx! |
xref #2579
This at least solves the direct negative consequence (erroring code by importing pandas) of registering our converters by default.