I guess we have to come up with an overlay design that can let us go from step handle to path handle without a locate().
Which I think means we might need to copy all the step handles from the backing graph into a map somewhere.
This PR does just that (in a pretty naive way): add a step->path cache in order to speed up PackedReferencePathOverlay::get_path_handle_of_step()
on a chr10 deconstruct test from "chicken-pangenome-sep23` with 8 cores on my desktop
xg/gbwt: 47 s 439,220 kb
gbz (before): 224 s 841,624 kb
gbz (now): 61 s 885,008 kb
quoting @adamnovak
I guess we have to come up with an overlay design that can let us go from step handle to path handle without a locate(). Which I think means we might need to copy all the step handles from the backing graph into a map somewhere.
This PR does just that (in a pretty naive way): add a step->path cache in order to speed up
PackedReferencePathOverlay::get_path_handle_of_step()on a chr10 deconstruct test from "chicken-pangenome-sep23` with 8 cores on my desktop
Resolves #172