Behavior protocols in classes

[lars-reimann]

Is your feature request related to a problem? Please describe.
Not all chains of functions calls and attribute accesses are valid.

Describe the solution you'd like

• Specify behavior protocols
• Repetition (0-n or 1-n)
• Option
• Alternative
• Sequence
• Syntax: protocol { f* g+ h? (i|j) k } - here f to k are references to callables and attributes (meaning reading accesses)
• Can refer to instance methods and instance attributes the next surrounding class (cannot access anything in outer classes if nested)
• Only makes sense in classes since it describes the lifecycle of a single instance
• At most 1 per class
• Must only be used in stub files (that's a given if it can only be used in classes)

Semantics

• If a class has a protocol only call sequences that conform exactly to the protocol are allowed. There must be no unmatched call prefix or suffix and nothing unmatched in-between.
• Protocol can be overridden by subclasses. However, the overriding protocol must conform to the overridden protocol. A class can only specify additionally, when its own methods/attributes can be used.
• If a class Sub has no own protocol, but a superclass Super has, instances of Sub must adhere to the protocol defined in Super, when using features of Super. Methods defined in Sub, however, can be used freely.
• If a class has no own protocol and neither of its superclasses does, all features can be mixed freely.

Tasks

• Scoping (can point to subterms and functions/attributes defined in the class or inherited from a superclass)
• Validation: referenced functions and attributes must not be static
• Validation: no duplicate names for subterms
• Update Prolog converter
• Creators
• Accessors
• Actual validation that calls conform to the protocol

[lars-reimann]

First batch of tasks done in Simple-ML/Simple-ML#124.

[lars-reimann]

We won't work on this further until we find more use-cases. The feature is not worth pursuing just to ensure the fit is called before predict or transform. See also #518.