Re: [Caml-list] Menhir grammar with sequences delimited by same token

[Dario Teixeira <dario.teixeira@nleyten.com>]

Hi,

> Jacques-Henri (or anybody else) please correct me if I am wrong.
> 
> Sometimes one way of dealing with such problems is to recognize (at the
> syntactic stage) a
> language that is a bit bigger than the desired one and then to check, 
> at
> the semantic level, that the real constraints are satisfied.

You make a good point, but in this case the reverse actually applies.
Trees such as "[Bold [Bold Text]]" are valid according to the AST,
and are in fact possible with other markups.

Such trees may be nonsensical in practice, but are also harmless,
and the "AST sanitiser" (a component that checks for semantic
errors in the AST) does not presently complain about them.

The longer revised grammar (the one with inline_sans_bold et al)
accepts only a *subset* of the valid ASTs, because it excludes
such improbable (but valid!) derivations.  Therefore, this is
a case where the parser deliberately recognises a language that
is *smaller* than that accepted by the AST.

Best regards,
Dario Teixeira