@akeeste I tried running the Capytaine with the example cases as well as the demo cases, and all of them came out with the "new" format. Were you using Capytaine 1.3 when you tried it?

@jtgrasb Yes I used Capytaine v1.3 with the demo.py file as is. However I still don't see any changes to how the body name or real/imaginary indices format in readCapytaine. Before merging this PR, I would like to better pinpoint what option causes this change and fix it on the Capytaine side if possible.

Adam

@akeeste This is strange. I have checked a lot of potential sources for this error, but can't seem to figure it out.

• The .nc files already in our repo (sphere_full.nc, etc.) work fine with our current readCapytaine script.
• Any new runs using the cases already set up without any changes (sphere.py, demo.py, etc.) lead to the "new" format with changes to the body name and real/imaginary indices.
• We are both using Capytaine v1.3 so that doesn't seem to be the issue (I also tried updating to v1.4 and it didn't change anything)
• I am using MATLAB 2021b, but I also tried MATLAB 2020b and actually got a different error shown below:
  

Let me know if you have any ideas where these error might be coming from.

Thanks,
Jeff

@jtgrasb Thanks for the update. I'm not sure where our workflows are differing or why R2021b and R2020b give different errors. At the next team meeting let's assign a couple different developers to try and recreate this issue and pin down what is happening.

We are both using Capytaine v1.3 so that doesn't seem to be the issue (I also tried updating to v1.4 and it didn't change anything)

Nothing changed in these modules of Capytaine for years.

I expect the change to come from either xarray or the library used in the backend by xarray to export the netCDF file.
The choice of a backend engine by xarray depends on what is installed in your environment.

Maybe the "old" format could be reproduced using a conda environment with an old version of xarray and/or explicitely setting the engine argument of to_netcdf.

@akeeste and @jtgrasb I can confirm this issue with the body name. I just reran the OSWEC case in Capytaine v1.4, and ran BEMIO in 2022a. I got the following error:

Error using split
First argument must be text.

Error in readCAPYTAINE (line 91)
hydro(F).body = split(hydro(F).body,'+');

Error in bemio (line 4)
hydro = readCAPYTAINE(hydro,'oswec.nc');

@mancellin's suggestion that these issues could be due to a change in one of the Capytaine dependencies, like xarray, is a likely cause.

I can confirm that this PR resolves the issue for v1.4, and also works for the prior v1.3 release. I'm going to go and merge it so that dev is compatible with Capytaine v1.4 .

@jtgrasb and @akeeste please open a new PR if there are other issues that need to be resolved.