Update TimeSeries handling of absolute times
#791
Conversation
Codecov Report
@@ Coverage Diff @@
## master #791 +/- ##
=======================================
Coverage 96.87% 96.87%
=======================================
Files 88 88
Lines 6019 6025 +6
=======================================
+ Hits 5831 5837 +6
Misses 188 188
Help us with your feedback. Take ten seconds to tell us how you rate us. |
weldx/core.py
Outdated
| if self.time is None and time.is_absolute: | ||
| # type: ignore[union-attr] | ||
| data = data.weldx.reset_reference_time(time.reference_time) |
There was a problem hiding this comment.
This is something we should probably discuss. It fixes the following problem:
We replaced
return ut.xr_interp_like(
self._data,
{"time": time.as_data_array()},
method=self.interpolation,
assume_sorted=False,
broadcast_missing=False,
)with
return ut.xr_interp_like(
data,
time.as_data_array(),
method=self.interpolation,
assume_sorted=False,
broadcast_missing=False,
)to get absolute times working correctly when using xr_interp_like. However, this causes the test case (ts_const, time_mul + pd.Timestamp("2020"), [1, 1, 1, 1, 1, 1, 1, 1], "m") of test_interp_time to fail (currently line 450 in test_core.py). The test just checks if the value of a constant TimeSeries that gets interpolated with an array of times is broadcasted correctly. Since the value is constant, reference times shouldn't matter, but after the change, they suddenly did. So I caught this special case and just added a reference time to a copy of the data prior to interpolation. The question is if we want to keep it that way, should it be handled differently or is that something that xr_interp_like should handle internally
Changes
Addresses the issues mentioned in #471
Related Issues
Closes #471
Checks
updated doc/update example/tutorial notebooksupdate manifest file