[WIP] add signatures() method to both LCA and SBT indices

[ctb]

NOTE: this is a PR into #556.

Some quick notes for @luizirber -

• the SBT class has insert already defined but it takes different arguments than Index.insert.
• at least at the moment LCA indices don't support inserting.

[luizirber]

In the rust side it's taking a signature (sig47, in this case) and building the SigLeaf internally. (This is the code that does the conversion).

Issue: with this design you can't set metadata (I'm setting it to empty). We don't use metadata that much, but might be good to clarify what is the intended use for it (only internal properties for SBT operations? user-provided metadata?)

[ctb]

I think all signature metadata should be in SourmashSignature. Maybe we can allow insert to take hints or something? Dunno. But for now is it OK if I change tree.insert to take SourmashSignature objects?

[luizirber]

Yup, taking a SourmashSignature is the way to go

[ctb]

alas, it is not as easy as that! 'insert' is currently used to add bloom filter nodes too. What do you think about de-deprecating add_node, and having insert construct a SigLeaf and then use add_node underneath? Then add_node would be used for bloom filter nodes.

On Fri, Dec 13, 2019 at 05:14:22PM -0800, Luiz Irber wrote: Yup, taking a SourmashSignature is the way to go -- You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: #796 (comment)

-- C. Titus Brown, ctbrown@ucdavis.edu

[luizirber]

alas, it is not as easy as that! 'insert' is currently used to add bloom filter nodes too.
What do you think about de-deprecating add_node, and having insert construct a SigLeaf and then use add_node underneath? Then add_node would be used for bloom filter nodes.

I think that's fair. In this case we would have the Index abc implemented for the SBT with minhash leaves, but not for the 'regular' SBT (bloom filters for internal and leaf nodes).

The deprecation warning was for pointing to libraries using the SBT directly that they should think about using .insert() instead, but since add_node would continue working I don't think we need to deprecate it.

[luizirber]

maybe do

sig = load_one_signature(utils.get_test_data(f))
tree.insert(sig)

here?