Re: Calling a parser from a parser

[Frank Heckenbach]

Hans Aberg wrote:

> There is in fact a situation where one writes a parser within a parser, and
> that is the case when the inner parser is dynamic in a way that it cannot be
> handled by the outer parser. This happens, for example, when the grammar has
> a large number of precedence levels (as in say Pascal): One can then let the
> outer parser just collect the tokens into a sequence, and let another parser
> sort out the semantics by the use of the precedence levels. In a language
> with just a few precedence levels, like Haskell, which only has ten, one can
> do it either way: Use this method, or write a .y section for each precedence
> level.

I really don't know what's it with you and Pascal, but just FYI,
Pascal has only 5 levels of precedence:

  parentheses
  exponentiating-operator
  multiplying-operator
  adding-operator
  relational-operator

Some dialects may have one or two more, but that's still quite a bit
less than C (which has 15 according to gcc's grammar) or Haskell (10
according to you).

Apart from that, why do you think that a simple parser couldn't
handle even many more levels of precedence? In fact, that's one of
the strengths of LALR(1), and the canonical examplel are of this
kind.

Frank

-- 
Frank Heckenbach, address@hidden
http://fjf.gnu.de/
GnuPG and PGP keys: http://fjf.gnu.de/plan (7977168E)