Ad-hoc buffers?

[asmaloney]

In ccm you can just declare a buffer and start using it.

class FooModel(ACTR):
  foo=Buffer()
  ...

According to vanilla ACT-R, buffers may only be added through modules.

When the model (sic) is defined it may have any number of buffers associated with it. This is the only way to add new buffers to the system – they do not exist independently of a module. The buffers for a module will provide the interface for other modules to interact with the new module.

From a theory standpoint is this restriction necessary/useful?

Should amod allow creation/addition of ad-hoc buffers like ccm does? Should it provide a special module to create them?

[asmaloney]

I created a special gactar module called extra_buffers to handle this. They are created by adding the extra_buffers module to config and listing them like this:

~~ config ~~

modules {
    extra_buffers{ foo{} }
}

Still not clear on the theory question, but all three implementations do have a path for adding them.

In ccm you just declare it and start using it:

class FooModel(ACTR):
  foo=Buffer()
  ...

In pyactr, you use the unfortunately-named ACTRModel.set_goal like this:

foo = pyactr_topdown.set_goal('foo')

This means "create a goal-like buffer and set it in the model", not "set the goal to 'foo'".

In vanilla, you create a module for each of them like this:

(define-module foo (foo) nil
	:version "1.0"
	:documentation "Extra buffer: foo"
	:query goal-style-query
	:request goal-style-request
	:buffer-mod goal-style-mod-request)

Merged in PR #217.