Re: Identification of root rule

> The first thing that occurs to me is that we could allow a declaration
for the start rule in the prolog.

Yes please.

Having an explicit label for the root rule makes it checkable without
having a mental model of how parsing works at the implementation level.
From this user's perspective, that's desirable even if that part of the
mental model would be easy to construct.

-- 
Graydon Saunders  | graydonish@fastmail.com
Þæs oferéode, ðisses swá mæg.
-- Deor  ("That passed, so may this.")

On Tue, Oct 17, 2023 at 9:50 AM Norm Tovey-Walsh <norm@saxonica.com> wrote:

> > On the other hand, I feel bad for the user; I think notations should
> > try to serve users, and not the other way round: usability first.
> > Which is why I did my original implementation that way.
>
> I like the simplicity of “the first rule”. I think it’s easier to
> explain than explaining that we work backwards to find the start rule.
> Also, I don’t think we currently forbid useless rules:
>
> S = 'a', B.
> B = 'b'.
> C = 'c'.
>
> so it’s possible to have grammars with more than one plausible start
> rule.
>
> Having said all that, I’m very sympathetic to the user’s issue.
>
> The first thing that occurs to me is that we could allow a declaration
> for the start rule in the prolog.
>
> > Anyway, it's a potential discussion point.
>
> Definitely.
>
>                                         Be seeing you,
>                                           norm
>
> --
> Norm Tovey-Walsh
> Saxonica
>