OMAS COCOs transformations do not appear to be adequate for em_coupling

[torrinba]

I'm not too familiar with what options that are available for COCOs transformations in OMAS, but from what I can tell they do not appear to be capable of supporting matrices which need to have different transformations applied for each dimension.

For example, the mutual_plasma_active matrix has units of inverse toroidal current (TOR) in one dimension (corresponding to pf_active coils) and units of flux (PSI) in the other dimension (corresponding to the poloidal flux generated by the coils).

This is a pretty unusual application for COCOs since it seems like the matrix would need to have the transformation PSI/TOR (which would be the same as PSI*TOR if TOR only involves a sign change), but I don't see any way to handle this in OMAS. Is this possible?

Currently em_coupling has no COCOs defined so any use of these variables from OMAS with non-native COCOs will produce incorrect results.

The other em_coupling matrices seem like they would have the same transformation or something trivial (TOR or None)

[orso82]

@torrinba I think we might have encountered a similar situation with the MSE. @jmcclena might remember. The approach there was to do set the COCOS transformation to be None for those variables, and handle the transformation of the individual elements by hand.

[jmcclena]

@orso82 That's what I was thinking at first too. But it was a bit easier for MSE, as each row had a different cocos. So we were able to set it with an array of cocoss

_cocos_signals['mse.channel.:.active_spatial_resolution.:.geometric_coefficients']=[None, None, 'TOR', None, None, None, 'TOR', None, None]#[DEL?]# -1.000000 # geometric_coefficients

[orso82]

@jmcclena I had forgotten about the array of COCOS!

@torrinba please take a look if you can use the same functionality (or perhaps you can extend it to work with matrixes)

[torrinba]

@orso82 the challenge with the em_coupling matrices is not that the rows or columns need to have different COCOs, it's that each dimension does. For example using the matrix A to compute b = A x, the vectors b and x have different COCOs so A needs to have a combination of those. Does OMAS support this?

[jmcclena]

@torrinba The logic for handling the array can be found here: https://github.com/gafusion/omas/blob/4ab29a3ffd8621a74699c3e8197c2b01f506f318/omas/omas_core.py#L1342

Perhaps we can add logic to split by *, and perform all transforms in the split for em_coupling?

[jmcclena]

Something like 1354 turns to:

norm = 1 
for transform in transform.split('*'):
    norm *= omas_physics.cocos_transform(self.cocosio, self.cocos)[transform]

[github-actions[bot]]

Stale issue message