[m-users.] A predicate taking a lists of predicates as input?

[Anders Lundstedt]

Thanks for the answers!


I think what I am trying to do is not possible in Mercury. For parsing
purposes, I want to be able to do something like the following. Given a list of
DCG rules r₁, ..., rₙ, each either det or semidet, I want to be able to
write the rule

r --> ?([r₁, ..., rₙ]).

(Even better would to be to be able to skip the brackets.)

This rule r should have the expected parsing behavior: it should try to apply
each of r₁, ..., rₙ, in that order, and consume tokens if and only if this
succeeds, and succeed as a rule no matter what.

To be able to write exactly as in the above definition of r, each of r₁, …, rₙ
must be of the same type. Thus some of the r₋ cannot be det while some of the
r₋ are semidet. I guess one solution would be to construct a discriminated
union type such that instead of ‘r₋’ one writes, for example, either
‘c_det(r₋)’ or ‘c_semidet(r₋)’. That would be quite verbose, though.


Tomas By <tomas at basun.net> wrote:

> So `?' applies a list of rules, and then you are nesting them?

Yes.

> If you really want to use DCG, ie experiment with the grammar and have
> something that is easy to work with, then just declare everything
> nondet and be done with it.

I think it will suffice to declare each rule as semidet.

> Then it might be better if you try to explain what it is all about.

I have written a parser in Mercury, DCG-style, for a context-sensitive
language. (I do not know if the grammar formalism used to define the language
has a name. It is a parsing expression grammar with added features from affix
grammars.) I am currently trying to simplify the parser by using the well-known
?, * and + parser operators. The language or its Mercury-written parser is not
yet public, but my implementation of the operators are:

https://github.com/anderslundstedt/mercury-experiments/tree/master/parsing-operators

(Thanks to everyone on this mailing list who has helped me! Further input of
course appreciated, but for my present purposes probably not needed.)



Best,

Anders