v2.0.0

Welcome Blini2!

This version brings further speed and RAM optimization, as well as improved clustering and a new distance-dump feature.

⚠️ Warning: sketch files (.blini) created with Blini v1 are incompatible with this version. Datasets will need to be re-sketched with Blini2 in order to work with it.

Changes from v1.1.0:

• New hashing makes sketch files smaller and reduces RAM requirement.
• Improved clustering algorithm finds more "central" representatives, and achieves clustering quality (silhouette score) that approaches that of quadratic hierarchical clustering.
• New distance-dump feature that prints out pairwise distances between query sequences, similar to mash dist and skani triangle but faster and slightly more accurate. Can be activated with -d.
• Sketches that are too small to be meaningful (<25 hashes) now create an error. This can be solved by either reducing the scale, or by using the -ignore-too-short flag.
• The default scale is now 40, which is suitable for sequences of length 1000 bases and above.

Existing flags were not changed, so existing workflows should keep working the same, as long as the sketch (.blini) files are updated.

Please report any bugs or unexpected behavior!