command-line arguments

[mvolkmann]

How can I get the values of command-line arguments in Scryer Prolog?
In SWI-Prolog I use current_prolog_flag(argv, Args), but it appears that doesn't exist in Scryer.

[triska]

You can access them with argv/1 from the module '$toplevel'. Example:

$ scryer-prolog -- test
?- '$toplevel':argv(As).
   As = ["test"].

[pmoura]

Wondering why an explicit qualified call is required:

$ scryerlgt -- a b c
?- argv(L).
   false.
?- '$toplevel':argv(L).
   L = ["a","b","c","/opt/local/share ..."].

As we have:

:- module('$toplevel', [argv/1,
                        copy_term/3]).

And the module seems to be loaded automatically at startup.

[mvolkmann]

This is the first time I've encountered syntax like a $ inside single quotes followed by colon.
I'm referring to '$toplevel':argv(L).
Where can I read about that syntax?

[triska]

We can always ask Prolog what a term is, showing its canonical syntax with write_canonical/1:

?- write_canonical('$toplevel':argv(L)).
:('$toplevel',argv(_446))   true.

[mvolkmann]

Maybe I'm too new to Prolog for that to help me. I'm still wondering things like:

1. What is $toplevel and where can I read about in the Scryer docs?
2. What does the syntax :(thing1, thing2) mean?

[mthom]

1. $toplevel is completely unique to Scryer. It represents the toplevel, the environment that takes queries, prints solutions as goals, allows predicates to be defined and modules to be consulted, etc. It's not currently documented anywhere in the Scryer docs unfortunately.
2. It's used for module resolution: given a module named thing1, call its predicate thing2.

[mthom]

@pmoura is right, the qualification shouldn't be necessary. I don't think the toplevel module is consulted at startup, it's just loaded and the repl predicate is called internally. I'll look at fixing it tomorrow.

[i-am-neo]

Ok cool, thank you both. Because argv(As) yields a list of strings, what would you suggest to best convert a list of strings to atoms? Aka

maplist(atom_string, Atoms, ["a","b","c"]).

atom_string(apply) and maplist don't seem to exist in Scryer. Or maybe I'm missing something obvious.

[aarroyoc]

The predicate you're looking for is atom_chars/2. However, in Scryer we tend to use more strings than other Prolog systems, since they're so easy to work with by default.

[bakaq]

maplist/n is in the module library(lists). You can see the documentation for the libraries of Scryer here. You can use the search to find predicates that you know the name of but maybe aren't in the places you would expect them to be. Notice, however, that there are some predicates that are accessible but are not currently documented in this site (see issues #2020, #2030 and #2031).

[i-am-neo]

Thanks @aarroyoc @bakaq . The "bunch_of_rules" are written with atoms, thus looking to convert.

Or maybe the rules should be rewritten to take strings instead of atoms, one advantage that crossed my mind is regex. Any best practices to recommend in this case?

@triska like this?
scryer-prolog test4.pl -g main -- one two three on the command line?

[triska]

The Scryer-specific options should be stated before the file, the application-specific switches after the file, so better use:

scryer-prolog -g main test4.pl -- one two three 

I will contribute an explanation for this to the documentation.

Regarding strings and atoms: If you want to look "inside" the content of your elements, then strings definitely seem a lot more suitable: They can be conveniently and efficiently processed and analyzed with DCGs, list predicates, unification etc., and they can also be partially instantiated and can contain variables as elements, whereas atoms must always be ground terms.

[aarroyoc]

Ideally, a declarative approach should be used to manage command line arguments. There are some very popular libraries for that in Python (argparse) and Rust (clap), but I don't think there's an equivalent Prolog library. I would experiment with that :)

[pmoura]

Adding such a library to Logtalk as a portable library is in my TODO list. One example of such a library is:

https://www.swi-prolog.org/pldoc/man?section=optparse

[aarroyoc]

I made a prototype here: https://github.com/aarroyoc/args.pl

[UWN]

Does this mean that argv/1 is now a built-in? It would much better fit into a separate module, maybe together with related functionality.

[bakaq]

This would also make it easier to document (#2031), and I think this is a crucial predicate for people to know about.