Phillip Cloud

> > The one that immediately comes to mind is > > user does a read_csv with DuckDB, even passes in a table_name kwarg > > user calls list_tables and...

> I'm not opposed to explicit `list_` methods but I think we need to consider the user with less SQL experience and think about how we want to handle certain...

I'm not a big fan of APIs that encode flags with dynamic values like strings when there's an API to encode that information in a statically knowable and searchable way...

`con.list` seems fine. Just to be clear `con.list` would be the method for listing all tabular objects. Agree that `list_rels` complicates things with a new noun that maybe unnecessarily unfamiliar.

> I understand that this data-dependent case works if the filters are not fused, but ibis expressions shouldn't have assumptions about the underlying data. Removing filters for this case *removes*...

> By not fusing the operations these two will behave differently depending on the data which might be considered a downside as well. This can't be addressed statically (which means...

The workarounds aren't that clear IMO. It's not like we'd recognize the pattern and suggest an alternative. However, fusing makes the solution of "spell things out bit by bit" impossible...

I don't think that matters too much, given that this is incorrect behavior given what the user has written. I'm still not understanding what the rationale is for keeping the...

Reposting my comment from https://github.com/ibis-project/ibis/pull/9065#issuecomment-2082817013 I took the weekend to think about some of the issues here and did a bit of research on the SQL standard specification of predicate...

@sfc-gh-heshah Any chance this PR is still viable? It would be really great to properly support arrays (including `undefined` values).