Too many dimensions ISTP logging warning for support data

[sdhoyt]

I'm getting the following ISTP compliance warnings from cdflib

WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable de_time, but de_time does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable esa_step, but esa_step does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable mode, but mode does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable tof0, but tof0 does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable tof1, but tof1 does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable tof2, but tof2 does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable tof3, but tof3 does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable cksm, but cksm does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable pos, but pos does not have that many dimensions
WARNING:cdflib.logging:ISTP Compliance Warning: direct_events is listed as the DEPEND_1 for variable direct_events, but direct_events does not have that many dimensions

Each of the variables is a 1 dimensional support_data type variable that is not tied to an Epoch, so DEPEND_0 is left undefined and the index coordinate is set to DEPEND_1. This way of defining this passed the SKTEditor check, so I think it should be ok from an ISTP perspective.

[jameswilburlewis]

Is that type of CDF in a publicly available data set? I'd like to see what IDL SPEDAS and PySPEDAS do with it....I don't think we've ever come across a CDF with DEPEND_N support variables having their own DEPEND_N attributes. SKTEditor has never complained about missing DEPEND_N attributes on variables marked as "support_data".

"ISTP compliance" is a bit of a murky concept anyway, since the de facto standard seems to be "what SKTEditor accepts", which can and has changed over time. But it might be "more correct", in this situation, to use a DEPEND_0 attribute rather than DEPEND_1. To my way of thinking, DEPEND_N attributes are purely local to that parent variable, with no concept of propagating them along chains of dependencies. Just my $.02 worth....

[bryan-harter]

So in this scenario, "esa_step", "de_time", etc are all shaped something like like [2], and they have the attribute DEPEND_1=direct_event? Is this error for cdf_to_xarray, or xarray_to_cdf?

[bryan-harter]

And do you have an example of the cdf that is either generated or you're reading in?

[bryan-harter]

I think I figured it out, I made changes to xarray_to_cdf so that we allow support_data to have its own DEPEND_1 variables without needing a DEPEND_0