Karl Gutwin

I'd argue in favor of keeping this sort of behavior (well, fixing the bugs and inconsistencies, of course). It feels really straightforward to pass a column reference to a user-created...

This doesn't require `math.round`, as you can reproduce the missing decimals without it: ``` from [ { a=1.23 } ] derive { b=4.56 } ``` When you select the "Query...

Rewriting the original input to the new syntax: ``` from tb1 append tb2 select {new_col1=c1*2,new_col2=c2/c3} ``` Still produces the wrong output as of 0.13.3: ```sql WITH table_0 AS ( SELECT...

Yeah, I agree that we don't want to assume that the SQL dialect supports `UNION ALL BY NAME`. Actually, my question is whether or not we can establish PRQL's semantics...

I retested this and the fix in #5066 may have resolved the issue; well, at least the test case in the OP no longer fails as of e6650494. ``` (venv)...

Apparently this works to hide the bubble map scale: ``` options: { scales: { size: { axis: "size", display: false, } } } ``` TypeScript complains about `axis: "size"` so...

Looking again, is there a reason why the check for `if TYPE_CHECKING:` is present at the top of `model.py.jinja` when importing the model's lazy imports?

so close, looks like just one more cache (`cargo-nextest`) needs to be invalidated...

It's so close, now! https://github.com/PRQL/prql/actions/runs/16014368312/job/45178110951?pr=5346#step:15:880 Looks like the doctests somehow also need the linker patch.

I'm not sure why setting RUSTFLAGS didn't have any effect on the doctest compilation failure. Cross fingers that the upstream regression gets fixed, then we don't need this. https://github.com/rust-lang/rust/issues/141626#issuecomment-3027216890