Match fragmentary samples

[hyanwong]

There are a few use-cases of wanting to match samples that are e.g. single reads against a tree sequence. These would be represented by samples that have long missing flanking regions. The current sample matcher does the following:

          result = self.find_path(
                matcher=local_data.matcher,
                child_id=self.sample_id_map[j],
                haplotype=haplotype,
                start=0,
                end=self.num_sites,
            )

I suspect it would be easy to override the start and end values here, and treat it a bit like a partial-length ancestor match. There might be fiddling later on to allow samples not to be matched anywhere though.

[hyanwong]

I think all it needs is this:

            start = 0
            end = len(haplotype)
            if haplotype[0] == tskit.MISSING_DATA:
                # could have missing flank at start
                start = np.where(haplotype != tskit.MISSING_DATA)[0][0]
            if haplotype[-1] == tskit.MISSING_DATA:
                # could have missing flank at end
                end = np.where(haplotype != tskit.MISSING_DATA)[0][-1] + 1
            result = self.find_path(
                matcher=local_data.matcher,
                child_id=self.sample_id_map[j],
                haplotype=haplotype,
                start=start,
                end=end,
            )