Reading multiple terms from a string about trealla HOT 13 CLOSED

Here is a quick and rather untested ad hoc definition of term//1:

:- use_module(library(dcgs)).
:- use_module(library(charsio)).
:- use_module(library(lists)).

term(T) -->
        call(term_(T)).

term_(T, Cs0, Cs) :-
        append(Prefix, Cs, Cs0),
        catch(read_from_chars(Prefix, T), error(syntax_error(_),_), false).

Example:

?- phrase(("<",term(T),">"), "<  hello. >").
   T = hello
;  false.

Tested with Scryer Prolog.

from trealla.

A DCG nonterminal would be perfect. I'm currently using a DCG to look for delimiters (<? and ?> in my case 😉) and then calling read_from_chars/3, but I could use term//1 directly to do that quite elegantly.

Thanks for the workaround as well @triska!

from trealla.

In Trealla that is...

term(T) -->
        call(term_(T)).

term_(T, Cs0, Cs) :-
        append(Prefix, Cs, Cs0),
        catch(read_term_from_chars(T, Prefix), error(syntax_error(_),_), false).

?- phrase(("<",term(T),">"), "<  hello. >").
   T = hello
;  false.

EDIT: note this is no longer correct. it is.. catch(read_term_from_chars(Prefix,T), error(syntax_error(),), false).
EDIT2: or use the code directly as provided by Markus.

from trealla.

I think a DCG nonterminal term//1 would be useful for this. With it, one could use phrase/3 to read a term, and have the remainder available for further reasoning.

For example, one could read two terms with:

?- phrase((term(T1),term(T2),...), Cs).

or, using list differences, with:

?- phrase(term(T1), Cs0, Cs1), phrase(term(T2), Cs1, Cs2).

where Cs2 is the remainder after T1 and T2 are read from the list of characters Cs0.

from trealla.

By the way, as a workaround, until such a nonterminal is available, you can manually try every possible string prefix to read a term:

?- append(Prefix, Rest, Cs), read_from_chars(Prefix, T).

with suitable use of catch/3 to catch syntax errors, failing if Prefix cannot (yet) be parsed. This incurs quadratic overhead, which is unacceptable for longer strings, but may be acceptable for the present use case, and until a faster (linear) construct is available.

from trealla.

$ tpl
?- help(read_term_from_chars/2).
read_term_from_chars/2: read_term_from_chars(+chars,?term)
   true.

?- help(read_term_from_chars/3).
read_term_from_chars/3: read_term_from_chars(+chars,?term,+list)
   true.

from trealla.

Personally, I think a good building block would be read_from_chars_(T, Cs0, Cs), reading from Cs0 the term T, leaving remainder Cs. This can be used (internally, since it relies on the implementation dependent order of arguments) in DCGs, and we can also (portably) implement term//1 and read_term_from_chars/3 on top of it.

For example, with this building block, term//1 could be implemented as:

term(T) --> call(read_from_chars_(T)).

from trealla.

Sounds good.

from trealla.

from trealla.

Is this supposed to happen?

?- '$read_term_from_chars'(T,[],"abc.",Rest).
   T = abc, Rest = "abc.".

Puts my naïve DCG into an infinite loop.

from trealla.

from trealla.

This is working quite nicely now. Very quick and no problem with decimal points. Thanks a bunch!

Here it is in action: https://php.energy/moon.html

from trealla.

from trealla.