Can those who know EDCG check that this was done correctly

Help!
, ,
1

Using SWI-Prolog (threaded, 64 bits, version 8.1.24) on Windows.

Being new to using EDCGs just want to make sure I am doing this correctly. The key piece of the code that I am not sure is the correct way to do it with EDCGs is the checking of the end of input. Since the input is a a list, checking for [] is done using

    A/dcg,          % Peek at head of hidden dcg list
    A \= [],

The code simply copies values from an input list (closed list) one value at a time to an output list (open list), but uses the named accumulator dcg instead of creating one.

:- module(edcg_example,
    [
        example/2
    ]).

:- use_module(library(edcg)).

edcg:acc_info(output, T, Out, In, Out=[T|In]). % reverse accumulator

edcg:pred_info(example,0,[dcg,output]).

example -->>
    A/dcg,          % Peek at head of hidden dcg list
    A \= [],
    [H]:dcg,
    [H]:output,
    example,
    !.
example -->> [].

example(Input,Output) :-
    example(Input,Rest,Out_hole_0,Out_hole),
    assertion( Rest == [] ),
    Out_hole = [],
    Output = Out_hole_0.

Listing of code after transformation

?- listing(edcg_example:_).

example(A, B, C, D) :-
    true,
    A\=[],
    A=[H|E],
    true,
    C=[H|F],
    true,
    example(E, B, F, D),
    !.
example(A, A, B, B).

example(Input, Output) :-
    example(Input, Rest, Out_hole_0, Out_hole),
    assertion(Rest==[]),
    Out_hole=[],
    Output=Out_hole_0.
true.

Test cases

:- begin_tests(eos_tests).

eos_test_case_generator(      [],     [] ).
eos_test_case_generator(     [1],    [1] ).
eos_test_case_generator(   [1,2],  [1,2] ).
eos_test_case_generator( [1,2,3],[1,2,3] ).

test(01,[forall(eos_test_case_generator(Input,Output))]) :-
    call_with_time_limit(1,example(Input,Output)).

:- end_tests(eos_tests).

Example run

?- run_tests.
% PL-Unit: eos_tests .... done
% All 4 tests passed
true.

If it looks good a simple yes will do.

EDIT

In doing some rubber duck debugging noticed in the listing

    A\=[],
    A=[H|E],

and so changed the source to

example -->>
    % A/dcg,          % Peek at head of hidden dcg list
    % A \= [],
    [H]:dcg,
    [H]:output,
    example,
    !.
example -->> [].

and the test still passed.

Was about to delete the entire post but figure it might be of value to someone someday.

Still we be nice to here some feedback.

2

EDCGs for this should be the same as DCGs; and I don’t understand why you would want to check for “end of input”.

If your DCG/EDCG is

foo --> [a], [b].

then an input of [a,b] will be detected properly.

I presume you’re using a variant of phrase/3 or call_dcg/3 to call your rules?

1 Like
3

No.

Just making sure I understand some of the basic concepts of how EDCGs work before I try them on the code doing the Hex Dump that has several accumulators.

4

There should be no need to check for end-of-input explicitly; that’s done implicitly by the rules.

After all, phrase/2 is basically (I’m doing this from memory):

phrase(Goal, L) :- call(Goal, L, []).

but there’s no reason it can’t be call(Goal, L, L2), L2=[], so there’s no guarantee that L2 is instantiated – L1, L2 form a difference list, with L2 being eventually instantiated to []. (Also, DCGs work on things other than lists; and with EDCGs that’s even more likely – for example, some of my accumulators are rbtrees.)

Also: https://www.metalevel.at/prolog/dcg (and there are other tutorials).

5

You meant:

phrase(GRBody, L) :- phrase(GRBody, L, []).
1 Like