Skip to content

Gzipping all training files results in a nice reduction: add feature that allows scripts/modules to handle this #28

Description

@dkoslicki

For example, using the Metalign default training database (199807 genomes) and running

python MakeStreamingDNADatabase.py ${trainingFiles} ${outputDir}/${cmashDatabase} -n ${numHashes} -k 60 -v
python MakeStreamingPrefilter.py ${outputDir}/${cmashDatabase} ${outputDir}/${prefilterName} 30-60-10

results in uncompressed:

16G Mar 22 03:39 cmash_db_n1000_k60.h5
9.3G Mar 22 08:07 cmash_db_n1000_k60_30-60-10.bf
6.9G Mar 22 04:34 cmash_db_n1000_k60.tst

yet

4.6G Mar 22 03:39 cmash_db_n1000_k60.h5.gz
3.6G Mar 22 08:07 cmash_db_n1000_k60_30-60-10.bf.gz
3.6G Mar 22 04:34 cmash_db_n1000_k60.tst.gz

so ~2-4x compression.

Would need to either:

  • Enable MakeStreamingDNADatabase.py and MakeStreamingPrefilter.py to detect compressed training data and decompress it in the script or (better yet)
  • Enable decompression in the modules MinHash.py and Query.py themselves.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions