'run_compute' using a distribution fails when bundle contains a 'mesh' dataset.

[tilakpawar]

Trying to compute a model using distributions fails while drawing parameters from a distribution if the bundle contains a 'mesh' dataset. The error occurs even if the 'lc' dataset is specified in the run_compute command. I attach a simple jupyter notebook to reproduce the error and compare two bundles, one with and one without 'mesh' dataset.

run_compute_using_distribution_issue.ipynb.zip

[kecnry]

Thanks for reporting this, definitely a bug and something we shouldn't allow. In theory, this should work if sample_mode = 'all', but could also get very expensive on memory to store a large number of copies of the mesh. For anyone coming across this before its fixed, disabling the mesh dataset (using b.disable_dataset(...) or setting the enabled parameter to False) should avoid the error.