At Tue, 14 May 2019 18:58:14 +1200, Thomas Munro <thomas.mu...@gmail.com> wrote in <CA+hUKG+ojKTKw=aG6QU=vmpmc8sq7nm4ah7fk1e+g1yngcv...@mail.gmail.com> > On Tue, May 14, 2019 at 6:18 PM Kyotaro HORIGUCHI > <horiguchi.kyot...@lab.ntt.co.jp> wrote: > > I played with this a bit and found that "... (attr=[tab]" (no > > space between "r" and "=") complets with '='. Isn't it annoying? > > > > Only "UPDATE hoge SET a=[tab]" behaves the same way among > > existing completions. > > Hmm. True. Here's one way to fix that.
Thanks. That's what was in my mind. Some definition item names are induced from some current states (e.g. "CREATE TYPE name AS RANGE (" => "SUBTYPE = ") but I think it's too much. COLLATE is not suggested with possible collations but I think suggesting it is not so useful. PASSEDBYVALUE is suggested with '=', which is different from documented syntax but I don't think that's not such a problem for those who spell this command out. # By the way, collatable and preferred are boolean which behaves # the same way with passedbyvalue. Is there any intention in the # difference in the documentation? The completion lists contain all possible words correctly (I think "analyse" is an implicit synonym.). As the result, I find it perfect. regards. -- Kyotaro Horiguchi NTT Open Source Software Center