C. Titus Brown
C. Titus Brown
based in part on experiences in https://github.com/sourmash-bio/sourmash/issues/3534, I think we could improve our human-readable output for inspecting taxonomy results. this is particularly important as we head towards more comprehensive examination...
There are some troubling redefinitions, in particular...
in https://github.com/sourmash-bio/sourmash/pull/3606, I disabled `test_do_sourmash_index_zipfile_append` because it was not working properly under py3.13 - it seemed to be pulling in warnings from other tests running in other threads. we should...
In #3545 The use of `CollectionSet` for `LinearIndex` (used by `RevIndex` to store sketches) means that we don't store original sketches, but rather the rescaled ones. This breaks code that...
While working on https://github.com/sourmash-bio/sourmash/pull/3545, I noticed that the `CounterGather` tests were not thorough - in particular, the `union_found` stuff is not tested. We should upgrade these tests, and also explicitly...
PR https://github.com/sourmash-bio/sourmash/pull/3545 adds a full implementation of a memory-based `RevIndex`. Might be worth adapting it as a replacement for `MultiIndex`, which currently does a linear search w/o indexing of in-memory...
In https://github.com/sourmash-bio/sourmash/pull/3545, we upgrade the memory-based `RevIndex` to support the full `Index` protocol. I think we could, with relatively little effort, support multithreaded zip file search to `RevIndex`-based `CounterGather`.
https://github.com/sourmash-bio/sourmash/pull/3545/ adds basic support for picklists, but they should be examined for both speed and functionality with more care. In particular, it would be worth evaluating the speed impact of...
the `index()` function in commands.py is pretty ugly, and is getting uglier in https://github.com/sourmash-bio/sourmash/pull/3545. Consider factoring Python-based index creation out to another class, perhaps under `index/`.
in https://github.com/sourmash-bio/sourmash/issues/3558, there's not much of a point in doing a prefetch step, since the RocksDB-based `RevIndex` is quite efficient. Nonetheless, we need to implement the prefetch/`CounterGather` stuff, which adds...