Braden Shepherdson

I'm assigning this to the QPD team to look at the underlying metadata and any accidental assumptions around numeric keys vs. string-like keys. This may instead be an FE bug...

Problem 3 - using 2 decimal places for everything - was fixed in passing by #27616 .

I tried this and it worked for me. Then I looked back at this bug and saw a difference: - `[Created At] < "2024-06-30T23:59:59Z"` is allowed and queries correctly. -...

Yes, it works with the `T` and no time zone info. But if it used to work, then this is still a regression, since it can break existing questions needlessly....

I think this should remain with the QC team and be triaged by them. There's only a couple of basic conversions of the `visibility_state` fields on tables or fields in...

We're poised to overhaul the QP to use MLv2, and there's some significant changes that need to happen to support eg. Metrics v2. I think this is a worthy thing...

The first of those quirks is coming from MBQL lib, which doesn't return a "Zoom in" drill for category columns - only for bins, time series and geography. This is...

## Repro I was able to reproduce this locally with the following setup: - Create MBQL models "Orders Model" and "Products Model", but trivial wrappers for those two tables. -...

### Analysis The trouble is in `metabase.query-processor.middleware.add-implicit-joins`: - The main source is a model, so the query has a `:source-query` for the raw Products table. - When implicit join clauses...

I don't believe this is a regression. There's a workaround: either use the original table as the source instead of a transparent model; OR add an explicit join and use...