user:this, user:that ? hooks?

[jjtolton]

I'm sure this is one of those things that should be obvious but I see not documentation for -- what is the significance of prefixing user: onto some of these predicates? Whereas for other things, such as attribute_goals//, I see no such prefix.

https://github.com/mthom/scryer-prolog/blob/2f47343cb5feb09d4950dcc05dd29ff2b1a2443d/src/lib/dcgs.pl#L257C1-L276C1

I've seen:

• user:term_expansion
• user:goal_expansion
• user:exception

Is user a convention? Is it significant? Is it configurable? Does prefixing module symbols onto other predicates have other effects? If so, is load order significant, i.e.,

attribute_goals// is called in library('$project_atts') and verify_attrs// is called in library('$atts'), it seems.

mod:foo(A) :- A=foo.
:- use_module(mod, [foo/1]).

vs

:- use_module(mod, [foo/1]).
mod:foo(A) :- A = foo.

Or is this trick something that only works for user:*.

By the way, I'm not trying to get hung up on what these functions do, (will save that for a different post), I'm trying to understand the user: prefix and why term_expansion and goal_expansion are defined with a user: prefix and attribute_goals// and verify_attrs//, which seem to use the same "technique", are not.

What would we call these things, hooks?

[triska]

Term and goal expansions defined in the user module are meant to pertain to all future clauses that are compiled. A library such as library(clpz) defines such expansions in user: for them to affect also programs that load the library, whereas module-local expansions and also verify_attributes/3, attribute_goals//1 etc. are confined to the module in which they are defined.

[jjtolton]

Interesting. And I see one example of defining a "hook" for a module other than user: for dif:attribute_goals//. Fascinating.

[triska]

This only invokes a nonterminal from a different module. Still an interesting catch! I think this specific case may be more elegantly solvable with copy_term/3, instead of invoking internal predicates of library(dif).

[bakaq]

Good idea! When I wrote those tests I wasn't aware of copy_term/3, it's definitely the better way to do it.