On Thu, Sep 5, 2019 at 10:15 AM Jeff Davis <pg...@j-davis.com> wrote: > > On Sun, 2019-09-01 at 06:26 -0700, Paul A Jungwirth wrote: > > @+ and @- and @* (I dunno why but I kind of like it. We already have > > @> and <@.) > > I think I like this proposal best; it reminds me of perl. Though some > might say that's an argument against it.
Thanks Jeff, it's my favorite too. :-) Strangely it feels the hardest to justify. Right now I have + and - and * implemented but I'll change them to @+ and @- and @* so that I can support `range R range = multirange` too. Btw is there any reason to send a "preview" patch with my current progress, since we're starting a new commit fest? Here is what I have left to do: - Change those three operators. - Write range_intersect_agg. (range_agg is done but needs some tests before I commit it.) - Write documentation. - Add multiranges to resolve_generic_type, and figure out how to test that (see the other thread about two latent range-related bugs there). - Rebase on current master. (It should be just a few weeks behind right now.) - Run pgindent to make sure I'm conforming to whitespace/style guidelines. - Split it up into a few separate patch files. Right now I'm planning to do all that before sending a patch. I'm happy to send something something in-progress too, but I don't want to waste any reviewers' time. If folks want an early peak though let me know. (You can also find my messy progress at https://github.com/pjungwir/postgresql/tree/multirange) Also here are some other items that won't be in my next patch, but should probably be done (maybe by someone else but I'm happy to figure it out too) before this is really committed: - typanalyze - selectivity - gist support - spgist support If anyone would like to help with those, let me know. :-) Yours, Paul
