Sounds like a good idea, as long as it does not break backwards compatibility.
Regards, Ruben On Thu, Dec 2, 2021 at 7:22 AM Jay Narale <[email protected]> wrote: > ＋1 re-writing to other types would be easier with this > > On Thu, Dec 2, 2021 at 16:18 Alessandro Solimando < > [email protected]> wrote: > > > Hi Jacques, > > I have faced the same issue recently and I think it's a good idea to > have a > > generic version of it and have RelShuttle be one of its concrete > > implementations. > > > > Best regards, > > Alessandro > > > > > > Il Gio 2 Dic 2021, 06:02 Jacques Nadeau <[email protected]> ha scritto: > > > > > Right now, RelNode's accept method and RelShuttle forces a user to > return > > > RelNodes. This makes tree traversal/conversion extra painful in main > > > circumstances (you want to have more specific relnode classes explicit, > > you > > > want to rewrite to non-relnodes, etc). I've worked around this for many > > > years but really think it should be resolved inside Calcite. I wanted > to > > > get some feedback on introducing a more generic pattern (that > RelShuttle > > > becomes a concrete variation of). If people are on board with the > change, > > > I'll pull it through the codebase. > > > > > > You can see the rough idea in this wip patch: > > > https://github.com/apache/calcite/pull/2625 > > > > > > Basically, introduce a new parameterized visitor that people can use. > > This > > > doesn't change the existing behavior of RelShuttle (it's now just a > > > specific version of this generic version) but it does allow other > rewrite > > > patterns. > > > > > > Thoughts? > > > > > > -- > Warm Regards, > > Jay Narale >
