Matthew Johnston

Okay. I'll plug that in later today and dig into the issue more. The built in date functions in sqlite _should_ get us what we need here.

@iwarshak I can't recreate this. The local schema in the test suite has this ```sql CREATE TABLE IF NOT EXISTS "products" ("id" INTEGER PRIMARY KEY AUTOINCREMENT, "account_id" INTEGER CONSTRAINT "products_account_id_fkey"...

Yea that solution looks super close to being the solution to use. I believe we'd want to use `compile_env` in this case ```elixir @datetime_type Application.compile_env(:ecto_sqlite3, :datetime_type, :iso8601) defp expr({:datetime_add, _,...

Fixed under `v0.17.2` thanks a ton @krwenholz, I had to alter your solution a bit, it was failing integration tests expecting that microsecond to be returned.

Options don't need to be compile time, but I don't think users want to alter the adapter at runtime. The trade off is that every time the the `datetime_add` is...

@newmanjeff I think we can, there is overhead associated with it https://www.sqlite.org/autoinc.html

I think the area to do that would be here https://github.com/elixir-sqlite/ecto_sqlite3/blob/d4689abb69b7010ffac096dabae825578cb8d4ef/lib/ecto/adapters/sqlite3/connection.ex#L1731-L1750 Just need to handle the case where two primary keys are provided and not to use autoincrement in that...

Perhaps it could also be an opt in using the `opts`?

A lot of the issues I am grappling with is trying to maintain some backwards compatibility with old the old ecto2 sqlite adapter and decisions made there and this one....

No updates on this. I haven't explored this more. If you want to, you can take a stab at it 😄