More descriptive error reporting when creating sample data file?

[hyanwong]

Off the back of #545 (reply in thread) it might be helpful to actually output (or save) the data that we attempt to stick into zarr, if the attempt fails. Perhaps we could wrap

tsinfer/tsinfer/formats.py

Line 181 in c1f3295

array[start:end] = buffered

in a try/except block, and then dump the "buffered" array somewhere? Presumably it might be too big to print out though?

However, this may not be worth doing, if we are moving to inputting via sgkit in the future - I don't know what the timescale is on that though.

[jeromekelleher]

I don't think it's worth the effort to be honest, it'd be a lot of work. In the linked discussion it's a metadata issue, which I agree we could give a more descriptive error on, but that's probably an upstream tskit thing (we could say "error encoding metadata key 'x', data 'y' is not JSON serialisable" or whatever)

[hyanwong]

that's probably an upstream tskit thing (we could say "error encoding metadata key 'x', data 'y' is not JSON serialisable" or whatever)

Yes, that seems sensible.

[benjeffery]

tskit-dev/tskit#1505