Skip to content

Constraints Aspect¶ ⧉

The constraints aspect defines constraints for the concepts of the language.

General

You can create new constraints using the concept ConceptConstraints. The "new" experimental constraint rules still might not support all the features of the old approach.

How do you hide inapplicable operands in dot expressions?

Dot expressions: hiding inapplicable operands ⧉ (Specific Languages' blog)

Are there any useful helper classes?

  • ModelConstraints ⧉: This class can calculate the constraint/scope of a specific concept/node programmatically.

Properties¶ ⧉

You can use property constraints to add setters, and getters and check if a new property value is valid. The getters are especially useful for setting properties based on other properties. Example: A wrapper concept that referencable a node can use the name of the reference as its name. For a reference r, you can implement INamedConcept ⧉ in your concept and set the property getter of name to r.name.

How can you set default values for properties?

Use the language com.mbeddr.mpsutil.propertydefault from mbeddr platform ⧉.

The property value isn't valid 🔰

Specific Languages Blog: Invalid property values ⧉

References¶ ⧉

You normally add a node reference constraint to set the scope. The inline scopes and inherited scopes both have advantages and disadvantages. You normally use the inline scopes where you can extract the scope into a utility class when you want to reuse it. The inherited scope is often used for references that inherit the referenced concept's scope. It makes finding the original scope declaration harder, though.

How can you emulate references?

Pseudo-references: emulating references ⧉ (Specific Languages' blog)

How do you check for an unresolved reference programmatically?

1
2
3
4
5
private static boolean isValidReference(reference ref) {
    if (ref == null ) return false; }
    SNode target = SNodeOperations.getTargetNodeSilently(ref);
    target != null && target.getModel() != null;
}

answered by: @kbirken

Scopes¶ ⧉

How do you filter scopes? Which scopes are available?

FilteringScope takes an existing scope and can do the filtering. Using reference scope | MPS documentation ⧉ lists some other scope implementations.

What's the best way to wrap a list of nodes into a Scope?

I have a list of nodes and want to wrap them into a scope. How do you do that?

You're probably looking for ListScope.forResolvableElements(sequence<node<>> elements).

forResolvableElements creates list scopes and also implements getName(child) like forNamedElements, yet returning the resolveInfo if the node is an IResolveInfo, the name if it is an INamedConcept ⧉ or else calls getPresentation().

forNamedElements

I'm unsure about the use-case of ListScope.forNamedElements(sequence<node<>> elements), but it behaves unexpected in some cases:

Note that elements with a blank name are not part of the scope created by ListScope.forNamedElements(sequence<node<>> elements). They'll get silently omitted.

Note that elements that are not an instance of INamedConcept ⧉, will make forNamedElements throw an exception.

contributed by: @abstraktor

How can you combine scopes?

Use the class CompositeScope ⧉.

What's the difference between parent scope and super.getScope?

A parent scope asks your node or up in the hierarchy for the scope until it finds a scope provider. super.getScope will use the implementation from the super-concept.

What does the resolve attribute in the XML files is used for? It often shows up in the diff tool or when running migrations.

The resolve attribute helps to bind references again. Assume you have a Base Language class that declares a variable myVar that you refer to. Now you delete the variable, which destroys the reference to that variable. Then you create a parameter called myVar. After pressing F5, the variable reference is bound to the parameter again.

This rebinding also works after restarting MPS because the reference text string is stored. Now, If you change the name of myVar to myParam, MPS automatically updates the projection. It also updates the reference in the same model, not everywhere (because that would be a global search). Note: You can rename it everywhere if you don't just type over the name of myVar but use rename refactoring (F6) . If you had a reference to myVar in some unrelated model, maybe even in a different module, resolveInfo would not be updated.

MPS updates the resolveInfo at some point, most of the time when it rewrites the model. That might happen during migration. You can also force it by choosing main menu → File → Save All.

contributed by: @enikao

Comments