Mutation's "inherited" state

[jeromekelleher]

Getting the state that a mutation inherited is fiddly right now, and has been annoying me quite bit recently. We should add a method to the Python object at least to get it, but it would also be nice to have access to it in numpy for reasoning efficiently about reversions and so on.

Does inherited_state feel like the right term here?

The numpy stuff would depend on us making some decisions about vectorised access to the string columns which I guess I should open a seperate issue on.

tables = ts.tables
assert np.all(tables.mutations.derived_state_offset == np.arange(ts.num_mutations + 1))
derived_state = tables.mutations.derived_state.view("S1")
assert np.all(tables.sites.ancestral_state_offset == np.arange(ts.num_sites + 1))
ancestral_state = tables.sites.ancestral_state.view("S1")
# Compute the "inherited state" for each mutation
inherited_state = ancestral_state[ts.mutations_site]
mutations_with_parent = ts.mutations_parent != -1
parent = ts.mutations_parent[mutations_with_parent]
assert np.all(parent >= 0)
inherited_state[mutations_with_parent] = derived_state[parent]

# Check
for mut in ts.mutations():
    state0 = ts.site(mut.site).ancestral_state
    if mut.parent != -1:
        state0 = ts.mutation(mut.parent).derived_state
    assert state0 == inherited_state[mut.id].decode()

[benjeffery]

inherited_state sounds good to me, only other thing I could come up with parent_state but I think that could get confused with the parent node state.

[hyanwong]

I have wanted this a number of times too, so +1 from me. The name seems reasonable.

[hyanwong]

Presumably we would need to make sure that compute_mutation_parents() has been run before calculating this.

[jeromekelleher]

Yes, it's a bug really that we don't fully validate the mutation parents on tree sequence creation.