Ben Jeffery

Might it be nice to have a general method: `tskit.genotypes_as_alleles(alleles, genotypes)`, `tskit.genotypes_as_ragged_alleles(alleles, genotypes)` then `Variant.genotypes_as_alleles` would use this?

This is probably possible, but likely not a drop-in replacement as we do some customisation to cope with things like allowing `None` at the top level. See https://github.com/tskit-dev/tskit/blob/main/python/tskit/metadata.py#L61 One quick...

I've changed the title here as we'll get validation caching in for 0.4.1

Interesting, thanks @hyanwong will do a proper think about this after 0.4.0

Bumping this to 0.3.7

Hmm, I guess it wouldn't be a breaking change to always have the node table be one row larger than it actually is and that row have the correct null/unknown...

The two options would be: - Get `tsk_node_table_expand_main_columns` to write the virtual root values to all the new space. I think this would be a perf and memory hit given...

Yeah, probably the right choice. Could be created lazily to avoid the memory penalty to non-users?

Just chatted to @szhan about this to confirm this will belong on the `Variant` class next to `genotypes` et al. Currently all of this class's methods are C based -...

I agree, we have a lot of code that does the copy-clear-mutate-add/append dance. I prefer making a new table then assigning it as in your first example.