Jerome Kelleher

Need this for #675 Check for tests in referring to #415 and enable.

Added 0.4.0 milestone, but no particular rush with this I think if the feature hasn't been released upstream.

It would be nice if ``tsinfer combine-sample-slices`` warned if the total number of sample provided is not the same as what's in the original sample data file, but it shouldn't...

Example [here](https://github.com/jeromekelleher/sc2ts/blob/314ed7b8a0274127bdb9043b17b04ed882631718/sc2ts/inference.py#L1112) of abstracting out the matching from final result building.

This maps pretty naturally on to the per-site array paradigm, so we can probably build the appropriate array given the input data (genotypes and base qualities), and use that as...

Seems reasonable - I figure there'll be a bunch of QC steps first on the sgkit dataset before passing in, so users will be aware of sgkit much more than...

Huh - I *thought* the problem was that we weren't updating mutation parents, but it looks like we are: https://github.com/tskit-dev/tsinfer/blob/6ed592f62d3c92c00eb918ef2bcb50c8c8171bad/tsinfer/eval_util.py#L481] @szhan can you make a reproducible example here please by...

The short version of avoiding this is to make sure you have only two alleles @szhan. It's probably simplest to stick with old style infinite sites simulations for your work,...

I'm going to leave this open as there are issues to be resolved (but good you have a temporary workaround @szhan)

Re VCFs, there's a position transform option which gives you control over what happens to the coordinates on outout to vcf.