Support factors

[multimeric]

I know this is a known issue, but it I feel like it could do with an issue for tracking purposes.

Replication:

> duckplyr:::duckdb_rel_from_df(iris)
Error in duckplyr:::duckdb_rel_from_df(iris) : 
  Can't convert factor columns to relational. Affected column: `Species`.

[grantmcdermott]

Factor support could be implemented via enum types, no? Or, am I missing something obvious?

[BlackEdder]

Doing some digging it looks like there was a missing feature in duckdb, which made it complicated, but that bug is now solved! Hopefully factors can be enabled soon.

[krlmlr]

While we can add support for factors, enabling this will almost certainly unlock test failures that were hidden before because we had a fallback as soon as factors were present.

Action items:

• Enable
• Fix internal tests
• Run revdepchecks, fix

[TimTaylor]

@krlmlr - Any chance this could make it in to the 1.0.0 release? If it's a case of undoing https://github.com/tidyverse/duckplyr/commit/8da062f7915bf5134785b6684f0c396c56dd09d6, testing and then revdep checks (as opposed to needing a good understanding of internals) then I'd be happy to take a look / make a PR to that effect.

[krlmlr]

Thanks. It's unlikely to make it into 1.0.0: to be useful, we'll also need support for functions that accept factors as an argument. We'd also need support from the duckdb R package.

Happy to revisit after 1.0.0 is out. Also need to check the original trigger for https://github.com/duckdb/duckdb/issues/8561 .

[JerePlum99]

I wanted to bump this now that 1.0.0 is out - any timeline on when factors may be supported for duckdb + duckplyr?