Introduce shortcut syntax for optional bool in WHERE expr

[jongleb]

This PR adds new syntax:

WHERE A? AND B? OR C? which is a shortcut to WHERE @choice { None { TRUE } | Some (params) { A } } AND etc

And considers join_cond as not null by the default

Example:

DDL:

CREATE TABLE test19 (
  a INT NOT NULL,
  b INT NOT NULL
);
CREATE TABLE test20 (
  c INT NOT NULL,
  d INT NOT NULL,
  r TEXT NOT NULL
);

SQL:

SELECT *
FROM test19
LEFT JOIN test20 on test20.c = @test20a
WHERE { c = @choice2 }? 
GROUP BY b;

will be generated the following code:

let invoke_callback stmt =
      callback
        ~a:(T.get_column_Int stmt 0)
        ~b:(T.get_column_Int stmt 1)
        ~c:(T.get_column_Int_nullable stmt 2)
        ~d:(T.get_column_Int_nullable stmt 3)
        ~r:(T.get_column_Text_nullable stmt 4)
    in
    let set_params stmt =
      let p = T.start_params stmt (1 + (match choice2 with Some _ -> 1 | None -> 0)) in
      T.set_param_Int p test20a;
      begin match choice2 with
      | None -> ()
      | Some (choice2) ->
        T.set_param_Int p choice2;
      end;
      T.finish_params p
    in
    T.select db ("SELECT *\n\
FROM test19\n\
LEFT JOIN test20 on test20.c = ?\n\
WHERE " ^ (match choice2 with Some _ -> " c = " ^ "?" ^ " " | None -> " TRUE ") ^ " \n\
GROUP BY b") set_params invoke_callback

Also, it differs from the regular Choice in that, in this case, not a polymorphic variant is generated because for this case, it would lead to the useless matching from the regular option type to the poly version of the option in the user code.

[ygrek]

didn't look at the code

1. why is :: Type? here?
2. So it means if A? OR B? and both are None - then condition is true and everything is selected.
3. What is expected usage ratio for A! and A? One obviously can be expressed with the other (or via current manual unwrapping), so if one is very low use - i would prefer to not introduce extra syntax.

[jongleb]

why is :: Type? here?

Just shortcut
Remove?

So it means if A? OR B? and both are None - then condition is true and everything is selected.

Yes, something like with COALESCE (maybe a bad analogy), while the first is not null - type is always null

What is expected usage ratio for A! and A? One obviously can be expressed with the other (or via current manual unwrapping), so if one is very low use - i would prefer to not introduce extra syntax.

Ok I'll remove it

[ygrek]

ok, so keep this PR only about introducing automatic elimination for A? OR B?
I am also slightly uneasy about ::Type? precisely because it means different thing from WHERE A? which may be possibly confusing? Maybe it is worth to be more explicit with :: Maybe Type or some such or maybe I am overthinking

[jongleb]

@ygrek I removed the unnecessary stuff, made a couple of edits and allowed myself to merge, since this PR is already a month old, if there will any new comments, I will fix them in a separate PR