v0.9.21

Full Changelog: v0.9.20...v0.9.21

• The need for json file for comparison is entirely dropped. Instead parquet metadata is used to store database information
• Profiling is more memory friendly and reliable
• Bug resulting from naming conventions in profile and compare is fixed

v0.9.20

Full Changelog: v0.9.19...v0.9.20

• Significant speedup and memory management with --engine polars
• Added profile sort to the utilities along with parquet metadata (mostly not user facing)
• Added --sparse to build-matrix (Much lighter and no performance drop according to small scale local tests)

v0.9.18

• Improved visualization module

v0.9.16

Full Changelog: v0.9.15...v0.9.16

• Matrix implementation is now in place with streamlined CLI
• Improved docs and more testing

v0.9.8

Full Changelog: v0.9.7...v0.9.8

• Now you can add to the existing matrix db
• Now matrix compare outputs a duckdb file and can be resumed
• The parquet can be extracted from duckdb file using a utility cli function

v0.9.4

• Zipstrain-light is removed in favor of the new matrix approach
• Genome-long matrices replace scaffold long ones

v0.9.3

Changes:

• Better memory handling when building matrix
• Legacy codes removed

v0.9.0

• Stable and efficient GPU engine for ANI calculations (CUDA and MPS)

v0.8.5

• Docker with cuda 11 support added
• Old ZipStrain functionalities are trimmed
• Genes are now optional but recommended

v0.8.4

• Improved way for building large profile matrix (batch my available memory)