If we are going this way, adding shortcuts on symbols (and I'm not convinced that we should), we might as well do it for all the symbols....

That would be a breaking change, so I don't think that's worth it. On the contrary we have a real use case for the semicolon and it doesn't break anything

Do we really need to introduce a new token? Whitespace is allowed between the keyword and "%", so it would be better if it would as well between ";" and "%"; or does it introduce some parsing problems?

It doesn't seem to introduce any conflict.
Did [%, [@@@, etc introduce parsing problems?

how is it different from begin%foo e1; e2 end

@bobzhang, it's the same, just shorter. Basically the first form is nicer than the second:

(* first form *)
f x ;%bind
g y ;%bind
return ()

(* second form *)
begin%bind
  f x;
  begin%bind
    g y;
    return
  end
end

You could probably extend the interpretation of %bind to all immediately nested sequences, allowing to write:

begin%bind
  f x;
  g y;
  return ()
end

Well, you'd need to interpret all the let as well, as we often mix let ... in and ;, and that would be like the haskell do-notation. In practice, in real code we often have a mix of let, let%bind and ; at the same level so having a block bind would introduce more visual nesting

@diml, can you also do this { %ext a = 3; b = 4}, it is very similar to your extension here, I have a use case for it too, thanks!

Is anyone strongly opposed to this?

As the change is trivial and only effects what syntax ppx writers can use, I'll probably merge it in a couple of days if no one objects.

Just noticed this! This is great, especially if we adopt shortcuts like ;%b rather than using 'bind'. Still a little messy looking, but not too bad.

@bluddy I started a patch for Lwt's ppx here which we ended up closing due to a mixture of backwards-compatibility concerns and generally unpleasant-looking syntax with ;%lwt. The implementation was fairly straightforward though and should be simple enough to experiment with in other contexts.

I agree that ;%lwt isn't too appealing. ;%l is a little better, but really, I'd much prefer it if ;% was possible. Is there anything that theoretically prevents an empty string extension? In cases where brevity is desired, this could really be a game changer, because it's almost like a simple operator.