Slicing WaveformModes does not work #63
Description
A simple slicing of a WaveformModes object would result in a KeyError, saying time_axis not found from self._metadata.
A MWE is here:
from nrcatalogtools import RITCatalog
rit_code = 'RIT:BBH:0001-n100-id3'
rit_catalog = RITCatalog.load(download=False)
h_rit = rit_catalog.get(rit_code)
print(h_rit[2])I managed to trace this back to one line in the SXS WaveformModes object (line 119):
obj, time_key = self._slice(key)and because of this, doing the following will result in the same error:
from nrcatalogtools.waveform import WaveformModes
h_rit_viewed = h_rit.view(WaveformModes)
print(h_rit_viewed)If I print out the __dict__ attribute before and after view(), the lists of attributes are indeed wildly different.
I suspect that it is how the WaveformModes are constructed that is not compatible with this view() method, but I have yet to find out why.
Here is an excerpt from the traceback:
File ~/.conda/envs/.../site-packages/sxs/time_series.py:463, in TimeSeries.__repr__(self)
[461] def __repr__(self):
[462] r = repr(self.ndarray)
--> [463] return f"{type(self).__name__}{r[max(0, r.find('(')):-1]}, time={self.time!r}, time_axis={self.time_axis})"
File ~/.conda/envs/.../site-packages/sxs/time_series.py:297, in TimeSeries.time_axis(self)
--> [297] return self._metadata["time_axis"]
KeyError: 'time_axis'Before getting to this step, there was a little issue with the metadata, so I did a little fix in catalog.py
if type(metadata) is not dict and hasattr(metadata, "to_dict"):
metadata = metadata.to_dict()
# ↓ These are new ↓
elif isinstance(metadata, dict):
metadata = dict(metadata.items())Finally, maybe this would be helpful; here are the versions of the packages I have:
numpy → 1.26.2
sxs → 2024.0.25
nrcatalogtools → latest master branch