--print-clusters bug with --sphere option

[nlubock]

Hello,

There appears to be a bug when using the --print-clusters flag with the --sphere clustering option. The resulting clusters are just repeats of the centroid. Here's a reproducible example from the latest master e125759

> cat test.tsv
GGGGGGGGGGGGGGGGGGGG	50
AGGGGGGGGGGGGGGGGGGG	10
AAGGGGGGGGGGGGGGGGGG	5
AAAGGGGGGGGGGGGGGGGG	20
TGGGGGGGGGGGGGGGGGGG	20
TTTTTTTTTTTTTTTTTTTT	100

> starcode -d1 --sphere --print-clusters -i test.tsv
running starcode with 1 thread
reading input files
raw format detected
sorting
progress: 100.00%
spheres clustering
TTTTTTTTTTTTTTTTTTTT	100	TTTTTTTTTTTTTTTTTTTT
GGGGGGGGGGGGGGGGGGGG	80	GGGGGGGGGGGGGGGGGGGG,GGGGGGGGGGGGGGGGGGGG,GGGGGGGGGGGGGGGGGGGG
AAAGGGGGGGGGGGGGGGGG	25	AAAGGGGGGGGGGGGGGGGG,AAAGGGGGGGGGGGGGGGGG

Hope this helps!
Nate

[ezorita]

Indeed, there seems to be a bug in sequence transfer/canonical annotation. Thanks for reporting.

[ezorita]

This seems to solve the issue. Please pull the last commit to master.

Thanks!