Allow multiple sketch types (ksize, DNA/prot, etc.) in a database?

[ctb]

In #556 (comment), @luizirber suggested:

Can you build a LinearIndex with 3 different k-sizes? I think it should be possible to add signatures with more than one sketch (in this case it would be one signature with 3 minhashes, one for each k), and then search/gather will select the compatible sketch to compare. In this case the compatibility-checking is internal, and not exposed to other parts of the code.

(I think this also provides a clean way to add new sketches, without having to check conditions around the codebase to make them work.)

and I wanted to open up a separate issue to discuss this.

Briefly, I think:

• it's a fine idea to allow constructing and searching databases with multiple sketches per signature!
• it could be a user experience disaster to actually support this from the command line.

I envision people getting really confused by inconsistent results when they do a search and discover that they are getting wildly inconsistent results because they accidentally missed including 50% of the ksize=21 signatures.

Thoughts?

[ctb]

Well, practically speaking, this is now working fine for a variety of our Index classes - Zip files, directories, pathlists, signature collections - and I haven't noticed any problems myself. Closing for now. 🤷