Partial strings:

[UWN]

?- X = [a,b,c|_], X = [a,b,c|Y].
X = [a, b, c | _1], Y = _1.                             %%% OK
?- partial_string("abc",X), X = [a,b,c|Y].
X = [a, b, c | _], Y = _.                                 %%% ???

Why this difference?

[mthom]

In the second instance, Y points to the end of a singly allocated string buffer, which is not part of the heap, so there is no heap cell to refer to.

[UWN]

That does not make sense to me. The end must be a real variable.

[UWN]

On the heap, Xs0 = [a,b,c|Xs], Xs0 = [_|Xs1] would look like (byte-by-byte)"

a <---- Xs0 points here
b <---- Xs1 points here
c
\0\ % zero to terminate the actual string
\0\ % further zeroes for padding, if needed
Xs_b1 % four-byte address on a 32bit system (eight on a 64)
Xs_b2
Xs_b3
Xs_b4

[UWN]

See #24

[UWN]

chars_partialstring(Chs, Cs0,Cs) :-
    must_be(chars, Chs),
    can_be(chars, Cs0),
    can_be(chars, Cs),
   /* now the hard part */

[UWN]

Currently I get:

?- partial_string(_,_).
error: exception thrown: error(syntax_error(inadmissible_query_term), repl/0)

But this is not a syntax error. And further, "inadmissible_query_term" is what is found and not what is expected. See N226 for a lengthy analysis of what errors are about. TL;DR: They signal primarily what is expected and not what is found. In the long run, you will find yourself going through that table of error classification quite frequently.

An existence_error or a permission_error is preferable in this situation. OK, you can also signal a system_error.

[UWN]

Do you think you will be able to continue work on partial strings?

[mthom]

I do. Things are sluggish right now because I'm preparing to leave in two weeks.

[UWN]

How do you intend to proceed? It seems in spite of this, you have still these _-variables. There are no such variables in Prolog. The underscore is just a unique variable name. But the variable itself has an identity

[mthom]

I will change the _-variables to something else, probably using a specialized convention to denote that the variable points to the end of a character buffer.

[UWN]

Here is a related problem:

?- partial_string("abc", L), length(L, N), N = 4.
false.

Expected: success.

[UWN]

Yet,

?- partial_string("abc", L), L = [_,_,_,d|_],length(L,N), N = 4.
L = [a,b,c,d|_], N = 4, _180 = a, _189 = b, _192 = c, _196 = _.

[mthom]

I don't understand why the partial string "abc" should have 4 elements.

[UWN]

Because partial_string("abc",L) should be equivalent (but not internally equivalent) to L = [a,b,c|_]. And now,

?- L = [a,b,c|_], length(L, N), N = 4.

succeeds as well.

[UWN]

In other words:

?- partial_string("abc",L), length(L, N).
L = [a,b,c|_], N = 3.
% Expected: More answers
?- L = [a,b,c|_], length(L, N).
L = [a,b,c], N = 3, _181 = [] ;
L = [a,b,c,_10], N = 4, _181 = [_10] ;
L = [a,b,c,_10,_14], N = 5, _181 = [_10,_14] ;
L = [a,b,c,_10,_14,_18], N = 6, _181 = [_10,_14,_18] ...

( In any case: very nice ... )