Re: [DISCUSS] CAST legacy behaviour

2022-03-09 Thread Francesco Guardiani
Hi all, As I see this thread has consensus, here is the issue and PR to disable the legacy behavior by default: https://issues.apache.org/jira/browse/FLINK-26551 https://github.com/apache/flink/pull/19020 We target to merge it before the release of 1.15, unless there are any objections. FG On Tu

Re: [DISCUSS] CAST legacy behaviour

2022-03-01 Thread Marios Trivyzas
Indeed, if we manage to use the configuration from *flink-conf.yaml* down the stack, it would be easy for everyone to configure a "system-wide" legacy cast behaviour. Best regards, Marios On Tue, Mar 1, 2022 at 2:52 PM Timo Walther wrote: > +1 > > Thanks for bringing up this discussion one more

Re: [DISCUSS] CAST legacy behaviour

2022-03-01 Thread Timo Walther
+1 Thanks for bringing up this discussion one more time Marios. I strongly support enabling the new behavior in 1.15. It definitely has implications on existing users, but as Seth said, thinking about the upcoming upgrade story we need to make sure that at least the core/basic operations are

Re: [DISCUSS] CAST legacy behaviour

2022-02-28 Thread Seth Wiesman
+1 Especially as SQL upgrades are right around the corner, it makes sense to get our defaults right. Seth On Mon, Feb 28, 2022 at 7:14 AM Martijn Visser wrote: > +1 for setting this option to disabled by default. I believe failures > should be brought forward as soon as possible, so they can b

Re: [DISCUSS] CAST legacy behaviour

2022-02-28 Thread Martijn Visser
+1 for setting this option to disabled by default. I believe failures should be brought forward as soon as possible, so they can be fixed as fast as possible. It will also be less confusing for new users. Last but not least, I believe the impact on existing users will be minimal (since it can be ch

Re: [DISCUSS] CAST legacy behaviour

2022-02-22 Thread Marios Trivyzas
Thanks Francesco, The two arguments you posted, further strengthen the need to make it DISABLED by default. On Tue, Feb 22, 2022 at 12:10 PM Francesco Guardiani < france...@ververica.com> wrote: > Hi all, > I'm +1 with what everything you said Marios. > > I'm gonna add another argument on top of

Re: [DISCUSS] CAST legacy behaviour

2022-02-22 Thread Francesco Guardiani
Hi all, I'm +1 with what everything you said Marios. I'm gonna add another argument on top of that: the "legacy-cast-behavior" has also a broken type inference, leading to incorrect results or further errors down in the pipeline[1]. For example, take this: SELECT COALESCE(CAST('a' AS INT), 0) ...