Libraries

[giovinazzo-kevin]

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

Modules are the individual Ergo code unit. A module exposes a list of public predicates that become available to all derived modules. A module might also reserve special semantics, such as in the case of attribute modules.

A library is a special type of module that is used to group together other modules under one package that is accessible to both C# and Ergo. Libraries are also the entry point for several engine extensions that only become available once the library is imported, thus providing a mechanism to implement the aforementioned semantics while limiting their scope.

Structure

A library is an ordinary Ergo module. From C#, a matching class is added to the interpreter (if defined). That class contains:

• A list of interpreter directives provided by that library
• A list of solver built-ins provided by that library
• An entry point for handling Ergo events

[giovinazzo-kevin]

personal notes:

• Move data sources/sinks to CSharp library

  • This has proven a bit difficult due to the fact that it makes more sense for those to be in the scope of the solver. Libraries are longer-lived.

• Move tabling facilities and predicate transformations to Meta or (new) Tabling library

• Move expansions to Lambda or (new) Expansions library

• Should shell commands and abstract parsers be supplied by libraries?

  • Pros: it looks neater that way, since there's only one entry point; it feels more symmetric having commands, directives and built-ins under the same umbrella.
  • Cons: parsers aren't bound to a scope, and neither are shell commands (though they could be)

• Make sure that built-ins and directives declared by libraries come from the same module as the library, or make them unqualified to begin with.

  • This is not necessary for directives, as they don't specify a module
  • This might be undesirable at best for builtins, and pointless at worst