jltsiren / gbwtgraph Goto Github PK

GBZ should be a proper file format instead of just GBWT and GBWTGraph serialized in the simple-sds format. The new format will be:

1. Header
2. Tags
3. GBWT
4. GBWTGraph

GBWTGraph will lose tags, because they are now redundant.

Take this GFA as in.gfa:

H	VN:Z:1.0
S	11	G
S	21	G
W	sample	1	A	0	1	>11
W	sample	1	B	0	1	>21

And do:

bin/gfa2gbwt -b in
bin/gfa2gbwt -C in
cp in.gbz out.gbz
bin/gfa2gbwt -D out

It won't wound-trip the .gg and .gbwt through GBZ; it will instead fail with:

terminate called after throwing an instance of 'sdsl::simple_sds::InvalidData'
  what():  Invalid number of buckets
Aborted