Split "kmerseek index" into "index-create-sketch", "index-get-kmers", "index-rocksdb"

[olgabot]

Currently kmerseek index wraps three commands in Sourmash/Branchwater:

• index-create-sketch: sourmash scripts manysketch (uses as many CPUs as available, low-medium memory)
• index-get-kmers: sourmash sig kmers (takes a long time, but only 1 CPU and low memory)
• index-create-rocksdb: sourmash scripts index (uses as many CPUs as available, low-medium memory)

These were combined for convenience on the command line, but for creating Nextflow pipelines, we probably want to add the option to separate them so that they can be put into separate processes. For example, sourmash sig kmers only uses one CPU and takes a really long time, like 24hrs+ on larger datasets, but shouldn't be blocking on proceeding with doing kmerseek search.

To clarify, kmerseek-index would still wrap the call to all three of index-create-sketch, index-get-kmers, index-rocksdb.

Notice that each command is named search-VERB-something -- I like this style for clarity. Also taking suggestions on naming, the hardest problem!

[heuermh]

Notice that each command is named search-VERB-something -- I like this style for clarity. Also taking suggestions on naming, the hardest problem!

I think this sounds reasonable. It is also compatible with tab-completion, which should be something provided with kmerseek.