record chromosome information in provenance?

[petrelharp]

Previously (before there was top-level metadata) we stored all the metadata in provenance. This still happens, but it hasn't been updated to include information about chromosomes.

It seems like this is no longer strictly necessary, but actually a nice thing; besides, I think it'd be straightforward as it's just the same code as for top-level metadata.

[bhaller]

Previously (before there was top-level metadata) we stored all the metadata in provenance. This still happens, but it hasn't been updated to include information about chromosomes.

True. I think we still put that stuff in provenance for backward-compatibility; back when we first added the top-level metadata, we wanted to still leave stuff in the old place too, so as not to break code that relied on it being there.

It seems like this is no longer strictly necessary, but actually a nice thing; besides, I think it'd be straightforward as it's just the same code as for top-level metadata.

Not hard, yeah; writing new stuff out to the JSON info is super easy because the JSON library that SLiM uses is so easy to use. But isn't it just duplicative? Why do you want it in provenance as well? If anything, it feels like maybe we ought to remove the other duplicative info from provenance, now that some time has passed and probably nobody is relying on it any more...? But if you see a benefit to adding it, then I'm happy to do so.

[petrelharp]

Well, provenance is just different than metadata: provenance is the iterative record of the various operations that produced the tree sequence (history); metadata describes the current tree sequence. So, purely in the interest of having helpful/informative provenance records, I'd vote to add this in?

[bhaller]

OK, fair enough. Should happen in the next pass. :->

[bhaller]

Done. I put these keys under "parameters" since that's where most of the other stuff that is duplicated from the metadata has been put. I didn't increment file_version; you can just check for the presence of these keys and use them if they're there. So I guess in a sense they're considered "optional", and SLiM <= 5.0 just didn't write them out, and now it does. As far as I can tell there's no formal schema for the provenance stuff, beyond tskit's schema for the top-level keys of the provenance; so I didn't do any schema change. Seems fine, trivial change.