Ryan Michael

Doing it async as part of load changes the cost of loading data which is something we should think about. If we do it async we have to coordinate query/preparation...

Maybe the best solution is to separate the casting use case from the timestamp-assignment use case. In the case of casting, you're describing operations in Fenl, so it's possible to...

In the immediate-term, could we change from silently dropping the column to returning an error on data load? We could suggest converting the decimal to a float in the error...

From the old repo: > As part of the quick-fix for [#321 (closed)](https://github.com/kaskada/kaskada/-/issues/321) we dropped decimal columns. At some point, we should figure out how to handle them. Likely medium...

My concern with `last_n` is the loss of generality - it's not possible to express an operation like "a list of all the url's viewed since the last purchase", or...

Seems cool. Might be worth documenting a use-case somewhere

@bpot returning an iterator could get complicated with the retries - this seems like a more straightforward approach.

Adding bug, because we don't emit the current value when it changes on window reset (to null)

I think we chose this behavior because there was no other way to produce "daily emitting at end of window" style aggregations. I guess it's true that we can hack...

Possibly related - I've seen cases where these produce different results: ``` foo | when(p1) | when(p2) foo | when(p1 and p2) ``` This may be more predictable if we...