Sultan Orazbayev

This is a somewhat related problem (https://stackoverflow.com/questions/72171471/order-of-metadata-in-dask-groupby-apply) providing meta as a dict, appears to enforce specific column ordering and this leads to potential downstream errors.

I am both interested in the feature and interested in contributing to this. This would be especially handy with data that exceeds memory (so would be great to make this...

@vankesteren: while the PR is reviewed, it would be great if you could do an independent test-drive of the new `pattern` function.

Thanks for the suggestions! re: adding number of missing values: do you have a suggestion for the name of this column? `values_missing`?

1. No, it doesn't work with Numba 0.44 (and sparse 0.7). ```output import sparse import numba import numpy as np numba.__version__, sparse.__version__ >('0.44.0', '0.7.0') x=sparse.COO(np.random.rand(5*10**2, 5*10**2)).astype(np.float16) y = sparse.dot(x,x) z...

Thanks!

FWIW, there's [a related SO question here](https://stackoverflow.com/q/73314292/10693596).

Another possibility is to capture the error exits explicitly: https://bitbucket.org/snakemake/snakemake/issues/420/onerror-onsuccess-for-rules

It turns out that attempt must be passed via `resources:`. I tried experimenting with shell variable expansions, but in the end that was useless because on every attempt, the original...

This might be relevant (order of metadata affecting downstream operations): https://github.com/dask/dask/issues/9080