-1 (non-binding). I can see where you are coming from, but in my opinion checks in the codebase should be used to direct runtime control flow. Otherwise I think they belong in proper tests.
On Thu, Dec 5, 2019 at 9:56 AM Jarek Potiuk <jarek.pot...@polidea.com> wrote: > Here is a quick vote on using asserts in Airflow code. > > It is distilled from the discussion > https://lists.apache.org/list.html?dev@airflow.apache.org. > > Here are the two options: > > *[+1]* Allow using asserts in some specific cases.* > *[-1]**: Forbid using asserts.* > > The voting will last till Monday 4 pm CET. The committers have binding > votes, but everyone is encouraged to call advisory - non-binding - votes as > well. > > Consider that my +1 (binding) vote. > > > * [+1] The case are clearly "strictly meant for developers" assertions > (None fields mainly) - which are more like type annotations and can be > stripped away when optimising. If those asserts are stripped out, another > exception will be thrown out shortly. If we agree to that we will add some > clear rules for those asserts in CONTRIBUTING.md and make it part of code > review process to check if assertions are "proper". > > ** [-1] Forbidding using asserts is mainly due to ambiguities when to > use/when to not use asserts. If we agree to that, we will forbid using > asserts via pre-commits and remove all assertions in our code. > > J. > -- > > Jarek Potiuk > Polidea <https://www.polidea.com/> | Principal Software Engineer > > M: +48 660 796 129 <+48660796129> > [image: Polidea] <https://www.polidea.com/> >
