Chapter 6 Interface or Type Relationships

In this chapter, two kinds of terminology are used for identical concepts to get better understanding (e.g., by both engineers and computer scientists). A short summary of the terms is given in the following table. The details are defined in the rest of this chapter.

If A is a subtype of B, then B is said to be a supertype of A.

A type can conceptually be viewed as a set of values. When we say that the variable x has the type Real, we mean that the value of x belongs to the set of values represented by the type Real, i.e., roughly the set of floating point numbers representable by Real, for the moment ignoring the fact that Real is also viewed as a class with certain attributes. Analogously, the variable b having Boolean type means that the value of b belongs to the set of values {false, true}. The built-in types Real, Integer, String, Boolean are considered to be distinct types.

The subtype relation between types is analogous to the subset relation between sets. A type A1 being a subtype of type A means that the set of values corresponding to type A1 is a subset of the set of values corresponding to type A.

The type Integer is not a subtype of Real in Modelica even though the set of primitive integer values is a subset of the primitive real values since there are some attributes of Real that are not part of Integer (section 4.9).

The concept of interface as defined in section 6.3 and used in this document is equivalent to the notion of type based on sets in the following sense:

An element is characterized by its interface defined by some attributes (section 6.3). The type of the element is the set of values having the same interface, i.e., the same attributes.

A subtype A1 in relation to another type A, means that the elements of the set corresponding to A1 is a subset of the set corresponding to A, characterized by the elements of that subset having additional properties.

[Example: A record R: record R Boolean b; Real x; end R;

Another record called R2: record R2 Boolean b; Real x; Real y; end R2;

An instance r: R r;

An instance r2: R2 r2;

The type R of r can be viewed as the set of all record values having the attributes defined by the interface of R, e.g., the infinite set {R(b=false, x=1.2), R(b=false, x=3.4), R(b=true, x=1.2), R(b=true, x=1.2, y=2), R(b=true, x=1.2, a=2), $\mathrm{\dots }$}. The statement that r has the type (or interface) R means that the value of r s to this infinite set.

The type R2 is a subtype of R since its instances fulfill the additional property of having the component Real y; in all its values.

]

Based on a flattened class or component we can construct an interface for that flattened class or component. The interface or type (the terms interface and type are equivalent and can be used interchangeably, and are different from inheritance interface and class type) is defined as the following information about the flattened element itself:

• •

  Whether it is replaceable or not.

• •

  Whether the class itself or the class of the component is transitively non-replaceable (section 6.3.1), and if not, the reference to the replaceable class it refers to.

• •

  Whether it is a component or a class.

• •

  Additional information about the element:

  • –

    The flow or stream prefix.

  • –

    Declared variability (constant, parameter, discrete).

  • –

    The prefixes input and output.

  • –

    The prefixes inner and/or outer.

  • –

    Whether the declaration is final, and in that case its semantics contents.

  • –

    Array sizes (if any).

  • –

    Condition of conditional components (if any).

  • –

    Which kind of specialized class.

  • –

    For an enumeration type or component of enumeration type the names of the enumeration literals in order.

  • –

    Whether it is a built-in type and the built-in type (RealType, IntegerType, StringType or BooleanType).

• •

  Only for an operator record class and classes derived from ExternalObject: the full name of the operator record base class (i.e., the one containing the operations), or the derived class. See chapter 14 and section 12.9.7.

  The following item does not apply for an operator record class or class derived from ExternalObject, since the type is already uniquely defined by the full name.

• •

  For each named public element of the class or component (including both local and inherited named elements) a tuple comprised of:

  • –

    Name of the element.

  • –

    Interface or type of the element.

    [This might have been modified by modifiers and is thus not necessarily identical to the interface of the original declaration.]

The corresponding constraining interface is constructed based on the constraining type (section 7.3.2) of the declaration (if replaceable – otherwise same as actual type) and with the constraining interface for the named elements.

In a class all references to elements of that class should be limited to their constraining interface.

[The constraining interface consists of only the public elements, and if the declaration is replaceable the element is limited to the constraining interface.]

[The public interface does not contain all of the information about the class or component. When using a class as a base class we also need protected elements, and for internal type-checking we need, e.g., import-elements. However, the information is sufficient for checking compatibility and for using the class to flatten components.]

An interface of a class or component A is compatible with an interface of a class or component B (or the constraining interface of B), or equivalently that the type of A is a subtype of the type of B, iff:

• •

  A is a class if and only if B is a class (and thus: A is a component if and only if B is a component).

• •

  If A has an operator record base class then B must also have one and it must be the same. If A does not have an operator record base class then B shall not have one. See chapter 14.

• •

  If A is derived from ExternalObject, then B must also be derived from ExternalObject and have the same full name. If A is not derived from ExternalObject then B shall not be derived from ExternalObject. See section 12.9.7.

• •

  If B is not replaceable then A shall not be replaceable.

• •

  If B is transitively non-replaceable then A must be transitively non-replaceable (section 6.3.1). For all elements of the inheritance interface of B there must exist a compatible element with the same name and visibility in the inheritance interface of A. The interface of A shall not contain any other elements.

  [We might even extend this to say that A and B should have the same contents, as in the additional restrictions below.]

• •

  If B is replaceable then for all elements of the component interface of B there must exist a plug-compatible element with the same name in the component interface of A.

• •

  If B is neither transitively non-replaceable nor replaceable then A must be linked to the same class, and for all elements of the component interface of B there must thus exist a plug-compatible element with the same name in the component interface of A.

• •

  Additional restrictions on the additional information. These elements should either match or have a natural total order:

  • –

    If B is a non-replaceable long class definition A must also be a long class definition.

  • –

    The flow or stream prefix should be matched for compatibility.

  • –

    Declared variability is ordered constant < parameter < discrete < continuous-time (Real without prefix), and A is only compatible with B if the declared variability in A is less than or equal the variability in B.

    [For a redeclaration of an element the variability prefix is as default inherited by the redeclaration (i.e., no need to repeat parameter when redeclaring a parameter).]

  • –

    The input and output prefixes must be matched. This ensures that the rules regarding inputs/outputs for matching connectors and (non-connector inputs) are preserved, as well as the restriction on blocks.

    [For a redeclaration of an element the input or output prefix is inherited from the original declaration.]

  • –

    The inner and/or outer prefixes should be matched.

    [For a redeclaration of an element the inner and/or outer prefixes are inherited from the original declaration (since it is not possible to have inner and/or outer as part of a redeclare).]

  • –

    If B is final A must also be final and have the same semantic contents.

  • –

    The number of array dimensions in A and B must be matched.

  • –

    Conditional components are only compatible with conditional components. The conditions must have equivalent contents (similar to array sizes, except there is no : for conditional components).

    [For a redeclaration of an element the conditional part is inherited from the original.]

  • –

    A function class is only compatible with a function class, a package class only compatible with a package class, a connector class only with a connector class, a model or block class only compatible with a model or block class, and a type or record class only compatible with a type or record class.

  • –

    If B is an enumeration type A must also be an enumeration type and vice versa. If B is an enumeration type not defined as (:) then A must have the same enumeration literals in the same order; if B is an enumeration type defined as (:) then there is no restriction on the enumeration type A.

  • –

    If B is a built-in type then A must also be of the same built-in type and vice versa.

[Intuitively, that the type A is a subtype of the type of B means that all important elements of B are be present in A.]

Plug-compatibility is a further restriction of compatibility (subtyping) defined in section 6.5, and further restricted for functions, see section 6.6. For a replaceable declaration or modifier the default class must be compatible with the constraining class.

For a modifier the following must apply:

• •

  The modified element should exist in the element being modified.

• •

  The modifier should be compatible with the element being modified, and in most cases also plug-compatible, section 6.5.

[If the original constraining flat class is legal (no references to unknown elements and no illegal use of class/component), and modifiers legal as above, then the resulting flat class will be legal (no references to unknown elements and no illegal use of class/component and compatible with original constraining class) and references refer to similar entities.]

[If a sub-component is redeclared, see section 7.3, it is impossible to connect to any new connector. A connector with input prefix must be connected to, and since one cannot connect across hierarchies, one should not be allowed to introduce such a connector at a level where a connection is not possible. Therefore all public components present in the interface A that are not present in B must be connected by default.]

Based on the above definitions, there are the following restrictions:

• •

  A redeclaration of an inherited top-level component must be compatible with (subtype of) the constraining interface of the element being redeclared.

• •

  In all other cases redeclarations must be plug-compatible with the constraining interface of the element being redeclared.

[The reason for the difference is that for an inherited top-level component it is possible to connect to the additional connectors, either in this class or in a derived class.

Example:

If a subcomponent is redeclared, it is impossible to connect to any new connector. Thus any new connectors must work without being connected, i.e., the default connection of flow variables. That fails for inputs (and expandable connectors may contain inputs). For parameters and non-connector inputs it would be possible to provide bindings in a derived class, but that would require hierarchical modifiers and it would be bad modeling practice that a hierarchical modifier must be used in order to make a model valid. A replaceable class might be used as the class for a sub-component, therefore plug-compatibility is required not only for replaceable sub-components, but also for replaceable classes.]

[Functions may be called with either named or positional arguments, and thus both the name and order is significant. If a function is redeclared, see section 7.3, any new arguments must have defaults (and be at the end) in order to preserve the meaning of existing calls.]

Based on the above definition the following restriction holds:

• •

  The interface of a redeclared function must be function-compatible with or a function subtype of the constraining interface of the function being redeclared.

Note that variability of function calls, see section 3.8.1, cannot be determined using just the interface of a function, as the variabilities of default argument expressions are not expressed by the interface. Hence a function redeclaration being function-compatible does not ensure that function calls will fulfill variability requirements, and tools must therefore check variability requirements separately.

[Example: Demonstrating a redeclaration using a function-compatible function

Note: PointMassGravity is not function-compatible with GravityInterface (no default for m), but sunGravity inside PlanetSimulation is function-compatible with GravityInterface.

The constant failed in planet1, will violate variability constraints, whereas it will work in planet2. The call gravity(frame_a.r0) will work in both of them.]

Certain expressions consist of an operator applied to two or more subexpressions (A and B). This includes:

• •

  if-expressions, e.g., if x then A else B.

• •

  Array expressions, e.g., {A, B}

• •

  Binary operators if both operands are of simple types, e.g., A + B. Binary operators for other types are only defined for operator records, section 14.5, and do not necessarily require that the operands are type compatible with each other.

If the subexpressions satisfy the following restrictions they are called type compatible expressions. Otherwise the expression is illegal. The type of the full expression (e.g., if x then A else B) is also defined below.

• •

  If A is a record expression, B must also be a record expression with the same named elements. In an expression that is not an array expression those elements must be type compatible. In an array expression the two records may contain elements with different sizes, but apart from that they must be type compatible. That generates a heterogenous array of records, see chapter 10. The type of the full expression is a record comprised of named elements that are type compatible with the corresponding named elements of both A and B.

• •

  The rules for array expressions depend on the operation (the rules for binary operators are given in section 10.6 and for array concatenation in section 10.4.2). The rules for the remaining case of if-expressions and array-expressions are:

  • –

    If A is an array expression then B must also be an array expression, and ndims(A) = ndims(B). The type of the full expression is an array expression with elements compatible with the elements of both A and B. If both size(A) and size(B) are known and size(A) = size(B) then this defines the size of the full expression, otherwise the size of the full expression is not known until the expression is about to be evaluated. In case of an if-expression the size of the full expression is defined based on the branch selected, and for other cases size(A) = size(B) must hold at this point.

  • –

    If A is a scalar expression of a simple type B must also be a scalar expression of a simple type.

• •

  If A is a Real expression then B must be a Real or Integer expression. The type of the full expression is Real, compare section 10.6.13, unless the operator is a relational operator (section 3.5) where the type of the full expression is Boolean.

• •

  If A is an Integer expression then B must be a Real or Integer expression. For exponentiation and division the type of the full expression is Real (even if both A and B are Integer) see section 10.6.7 and section 10.6.5, for relational operators the type of the full expression is Boolean. In other cases the type of the full expression is Real or Integer (same as B), compare section 10.6.13.

• •

  If A is a Boolean expression then B must be a Boolean expression and the type of the full expression is Boolean.

• •

  If A is a String expression then B must be a String expression and the type of the full expression is String, unless the operator is a relational operator (section 3.5) where the type of the full expression is Boolean.

• •

  If A is an enumeration expression then B must be an enumeration expression and the type of the full expression is enumeration expression, unless the operator is a relational operator (section 3.5) where the type of the full expression is Boolean. The enumeration expressions must be defined in terms of an enumeration type with the same enumeration literals in the same order.

• •

  For array and if-expressions, if A has an operator record base class then B must also have an operator record base class, and it must be the same, and otherwise neither A nor B may have an operator record base class. This is also the operator record base class for the full expression, e.g., for if (cond) then A else B.

• •

  If A is derived from ExternalObject then B must also be derived from ExternalObject and they must have the same full name; and otherwise neither A nor B may be derived from ExternalObject. The common full name also defines the type of the full expression, e.g., for if (cond) then A else B.