Python 3.13 incompatibility: del arg_values['frame']

[jklymak]

With Python 3.13.2 and xmitgcm 0.5.2

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[3], line 3
      1 td = '../results/StraightSlopeExpStrat002Dye/input/'
      2 td = '../results/StraightSlopeUpsidedownStrat001/input/'
----> 3 with xm.open_mdsdataset(td, iters='all', endian='<', prefix='dissipation') as ds:
      4     display(ds)
      5     ds=ds.isel(j=0, j_g=0).isel(time=-4)

File ~/Dropbox/CanyonMix/.pixi/envs/default/lib/python3.13/site-packages/xmitgcm/mds_store.py:166, in open_mdsdataset(data_dir, grid_dir, iters, prefix, read_grid, delta_t, ref_date, calendar, levels, geometry, grid_vars_to_coords, swap_dims, endian, chunks, ignore_unknown_vars, default_dtype, nx, ny, nz, llc_method, extra_metadata, extra_variables)
    164 frame = inspect.currentframe()
    165 _, _, _, arg_values = inspect.getargvalues(frame)
--> 166 del arg_values['frame']
    167 function_name = inspect.getframeinfo(frame)[2]
    169 # auto-detect whether to swap dims

ValueError: cannot remove local variables from FrameLocalsProxy

Doesn't occur for Python 3.12. Looks like python removed del for whatever arg_values is.

Thanks!

[hmkhatri]

Is there a workaround for this error? I had the same error (Python 3.13.2 and xmitgcm 0.5.2).

[mjlosch]

Not a real fix, but I found that commenting out line 166 let's you use open_mdsdataset also with python 3.13
I guess this https://docs.python.org/3/whatsnew/3.13.html#defined-mutation-semantics-for-locals is the relevant change in python, and one needs to port the code following these notes https://docs.python.org/3/whatsnew/3.13.html#porting-to-python-3-13 (I have not clue what they mean).

[jklymak]

@mjlosch - do you have commit bit here? I have a couple of small change PRs, one of which fixes an incompatibility with modern xarray. I could easily put in a PR for this as well, but not sure anyone is maintaining.

[mjlosch]

No, I can only submit PRs just like anyone else. I agree, it's a bit slow, even slower than https://github.com/MITgcm/MITgcm itself (o:

[hmkhatri]

Tagging @rabernat @raphaeldussin @cspencerjones

Any idea who is currently maintaining xmitgcm?
Some computability issues with latest python and xarray versions need to be resolved.

[IvanaEscobar]

@timothyas Maybe we need broader merging capabilities to this repo?

[mjlosch]

Fixed with #353, thanks @IvanaEscobar