Sean MacAvaney

You can modify the CI in this branch, that's alright.

> Can you give "write" access temporarily Done!

I'm having trouble running some of the tests locally: ``` /opt/miniconda3/lib/python3.10/importlib/__init__.py:126: in import_module return _bootstrap._gcd_import(name[level:], package, level) test/util/docs/test_subset.py:2: in from .data import FakeDocs E ModuleNotFoundError: No module named 'docs.data' ```...

Great, thanks!

Sorry this is taking a bit. Supporting lookups by index is a bit tricky with the prefixing approach.

I'm a bit curious -- do you have a specific use care for breaking apart the 3 sub-datasets? I see the advantage in letting them be reused across years of...

I think I'm leaning toward materializing the unions, on the basis of reducing the complexity. If we're doing that, the dataset becomes a relatively routine one -- an iterator over...

Thanks! I'm building the tests and such now :)

I think it's mostly there! I'm only having some problems with v2. First, when I try to do a lookup, it fails in `DocsSubset.docs_count`, I think since the `docs_count()` isn't...

Thanks @bpiwowar and @andreaschari !