The ICON_LES page is currently (and since a while) broken. This was useful for testing #101 but should be fixed.
My initial investigation showed that the EUREC4A_ICON-LES_control_DOM03_reff_native.zarr dataset seems to be corrupted. While .zmetadata shows there should be a time variable, there is none. For some reason, this doesn't get picked up correctly by the client library as an error, and instead some bytes of the error response seem to be interpreted as time values. These (garbage) values don't fit into a human time range, and thus are decoded as CFtime objects instead of np.datetime64, which in turn crashes matplotlib.
Possible fixes for the notebook:
[x] fix the dataset by (re-) uploading the time variable
[ ] throw the (likely unusable) dataset out of the eurec4a_intake catalog
[ ] some code in the notebook which catches the situation and keeps it running for the other datasets
The ICON_LES page is currently (and since a while) broken. This was useful for testing #101 but should be fixed.
My initial investigation showed that the EUREC4A_ICON-LES_control_DOM03_reff_native.zarr dataset seems to be corrupted. While
.zmetadata
shows there should be atime
variable, there is none. For some reason, this doesn't get picked up correctly by the client library as an error, and instead some bytes of the error response seem to be interpreted as time values. These (garbage) values don't fit into a human time range, and thus are decoded as CFtime objects instead ofnp.datetime64
, which in turn crashesmatplotlib
.Possible fixes for the notebook:
time
variableeurec4a_intake
catalog@observingClouds any thoughts?