The following is a short description of important concepts used in the BETA language. Please note, that these descriptions are deliberately informal. The precise meanings of these terms must be found in [MMN 93].
A program execution is viewed as a physical model or representation of part of the world. Objects on the computer model phenomena in the world; attributes of objects model properties of phenomena.
A collection of objects. Some represent phenomena while others are simply part of the implementation.
Computer representation of a real world phenomenon. Its structure consists of attributes and actions.
Computer representation of a real world concept. Objects defined according to the pattern are called instances or pattern defined objects: A pattern is to its instances as a concept is to its phenomena.
An object representing a singular "one-of-a-kind" phenomenon - the object is not defined as an instance of some pattern.
The combined values of its measurable properties at some point in time.
A property which has a measurable value. The value may vary over time.
An object which is part of another object. Part objects are used to model part or aggregation hierarchies.
An autonomous self-contained object which is not a part object.
An attribute which is a reference to a separate object.
The actions of a phenomenon in a real world system often take place concurrently (i.e. in parallel) with those of other phenomena in the system. A single phenomenon normally alternates among its own actions.
An association or binding of a name to some entity. The syntactic construct used is the colon ":" as in, <name>: <entity>. For attributes of an object descriptor, these are sometimes referred to as the attribute name and attribute description, respectively.
A declaration binding a pattern name to an object descriptor, describing the structure of instances of the pattern. Pattern declarations serve as templates for generating objects having a given structure.
Declaration of a singular object binding the object name to the singular object description (an object descriptor).
An occurrence of an attribute's name in an object descriptor.
A reference from within a pattern's object descriptor to an attribute declared inside the same object descriptor.
Any attribute reference which is not local.
Used to describe the structure of objects and consists of a prefix part and a main part.
Part of object descriptor used to specify the superpattern of the descriptor. The prefix part is specified by a pattern name (or is empty).
Used to describe the additional structure of objects. Has the syntactic form(# E #) and consists of an attribute part and an action part.
Part of object descriptor used to describe the object's attributes. Consists of a list of attribute declarations.
Part of object descriptor used to describe the actions to be performed when the object is executed. Consists of three parts: enter-part, do-part, exit-part.
Part of action part describing the enter parameters.
Part of action part consisting of a list of imperatives.
Part of action part describing the exit parameters.
An object descriptor that can be compiled and executed.
An attribute that denotes an object. Reference attributes can be either static references or dynamic references.
A reference attribute that constantly denotes the same object. Such objects are often referred to as static objects. In cases where these objects are used to model part (or aggregation) hierarchies, they are referred to as part objects, that is, they are part of an enclosing object.
Used to define static reference attributes.
<name>: @<ptn.name or obj.descriptor>
A reference attribute that denotes a object. The reference is variable in that it may denote different objects over time. Initially it denotes NONE which represents "no object."
Used to define dynamic reference attributes.
<name>: ^<pattern name>
A repetition (or array) of object references referred to by a single name plus an index. The size of a repetition A is denoted by A.range. A refers to the first element in the repetition, A[A.range] to the last.
Name: [eval] @<ptn.name or obj.descriptor> Name: [eval] ^<ptn.name>
The size of the repetition can be dynamically extended by:
<number> -> A.extend
The pattern name appearing in a reference attribute declaration. It restricts the set of objects that can be denoted by the reference.
Used to denote attributes within an enclosing object.
Used to denote attributes within objects that are returned as the result of evaluations.
A reference attribute that denotes a pattern. The structure of the pattern is represented locally using a structure object. Such objects include a reference back to the object of which the pattern is an attribute. This reference is called the origin of the pattern.
Used to define a pattern.
<name>: <object descriptor>
Used to define pattern variable attributes. A pattern variable may denote different patterns during the execution. The qualification restricts the set of patterns which may be denoted by the pattern variable.
<name>: ##<pattern name>
Generally, a pattern used to model physical objects.
Generally, a pattern used to model action sequences.
A procedure pattern which computes and returns a value. Such patterns always have an exit part.
A pattern that is predefined within the BETA language. Examples are integer, real, boolean, and char. Relevant operations include: +, -, *, div, mod, and, or, not, true, false, =, <, >, <>, <=, >=.
Describes an action; executing the imperative causes the action. Imperatives appear in the do-part of an object. Kinds of imperatives include evaluations, reference assignments, dynamic object creation, and control structures.
An imperative that can cause state changes and may produce a value when executed.
An evaluation imperative that sets (changes) the value of an attribute.
3 -> I
An imperative used to change the value of a dynamic reference.
objRef -> dynObjRef
objRef may be any object reference but dynObjRef must be a dynamic object reference.
An imperative used to change the pattern denoted by a pattern variable.
ref## -> dynPatRef##
Ref may be the name of a pattern variable, the name of an object, or the name of a pattern but dynPatRef must be a dynamic pattern reference.
An evaluation imperative that causes several assignments.
3 -> I -> J
Imperatives used to create new dynamic objects.
&Pat or &Pat
True when two references denote objects that have the same state.
A = B
True when two references denote the same object.
A = B
True when two pattern references denote the same pattern.
A## = B##
Note that < and <= are also defined for pattern comparisons based on the inheritance hierarchy.
An evaluation imperative that causes invocation of a procedure pattern.
(arg1,arg2) -> &ProcPat
An evaluation imperative that causes invocation of a function pattern.
(arg1,arg2) -> &FuncPat -> result
An imperative that controls the flow of executions.
A control structure used to support iteration. A list of imperatives are executed repeatedly while an index steps from 1 up to the number of iterations.
(for Index: Range repeat Imperative-list
A control structure used to support selection. Based on evaluating a condition evaluation and comparing it to the values of a number of selection evaluations, one of a set of imperative-lists is executed.
(if E0 // E1 then I1 // E2 then I2 E // En then In else I if)
A control structure used to support boolean selection. Based on evaluating a condition evaluation and testing if it is true or false, one of two imperative-lists is executed.
(if E then I1 else I2 if)
A means of naming an imperative. References to the label (via jump imperatives) can be made from within the imperative.
L: (# ... do ... #)
Causes flow of control to "jump" to another location. A jump imperative is one of a Leave imperative or a Restart imperative.
Causes termination of the execution of a labelled imperative; execution resumes after the labelled imperative. This imperative can only appear within the labelled imperative.
Causes restarting of the execution of a labelled imperative, that is, jump is to the start of the imperative. Can only appear within the labelled imperative.
The nesting of one structure in another in the text of a program. In BETA, object descriptors and imperatives can be nested inside of other object descriptors and imperatives. It is the job of the programmer to use indentation to make such nesting visible to readers. In the following example, Deposit's object descriptor is nested inside of Account's.
Account: (# Deposit: (# E do E #); #);
An association of a name with some defining expression.
Recall that colon ":" always signals a declaration of some kind.
Any occurrence of a name in a program which is not a declaration. Note that this does not include keywords of the BETA syntax (e.g. if, for, repeat, do), but does include predefined pattern and attribute names (e.g. char, putInt, stream).
The part of the program text "covered" by the declaration, that is, where applications of the declared name refer to the given declaration. In BETA, the scope of a declaration is the object descriptor it appears in. The exception to this is that the declaration may be "hidden" by declarations of the same name in nested object descriptors or labelled imperatives. Note that the declared name can also be applied outside its object descriptor using remote access. We say that a name is local to the object descriptor in which it is declared and global to any nested object descriptors (for which it is not hidden).
A means of generating (and executing) a procedure object allocated as part of the enclosing object.
A -> P -> B
A -> P(# E #) -> B
This differs from dynamic generation, &P, in that the instance of P is generated only once rather than each time the imperative is executed. Note that inserted items should not be used to define recursive procedures. That is, an inserted instance of P may be specified in the action part of P.
A pattern P is a direct subpattern of Q if P extends (specialises) the definition of Q. Q is called the direct superpattern of P and instances of P are also instances of Q.
P: Q(# E #)
Q is called the prefix pattern (or simply prefix), while the contents of (# E #) is called the main-part of P. The prefix Q means that P's object descriptor inherits all of Q's declarations in addition to any new ones defined in P's main-part.
A pattern P is a subpattern of Q if it is either a direct subpattern of Q or a subpattern of a direct subpattern of Q. Likewise, Q is a superpattern of P if it is either a direct superpattern of P or a superpattern of the direct superpattern of P. A pattern can have at most one direct superpattern.
A pattern used only as a superpattern for other patterns, that is, it is not intended to be used to generate objects. If P is declared without the use of a superpattern, P: (# E #), then P is assumed to be a subpattern of the most general abstract superpattern, Object. Note that the basic patterns, Integer, Real, Boolean, Char and Real are not subpatterns of Object.
If R is a dynamic reference qualified by the pattern Q (i.e. R: ^Q) and Q is a superpattern of P, then instances of both P and Q can be assigned to R. However, only attributes of Q (and of superpatterns of Q) can be accessed using remote access through R. That is, if attribute A is declared in the main part of P, then the remote access R.A is illegal.
The use of a subpattern to extend the action part of a pattern. Action specialisation can involve any or all of the enter-part, exit-part and do-part. The enter and exit parts of instances of P (again, a subpattern of Q) consist of Q's enter and exit parameters together with those defined by P. Extending the do-part of Q requires the use of the inner imperative in Q's action part. Executing the do-part of an instance of P proceeds by executing Q's do-part and executing P's do-part each time inner is encountered.
Q: (# E do E inner E #); P: Q(# E do E #);
A pattern attribute V of a pattern Q is virtual if it is only partially defined in Q. That is, the definition of V can be extended in subpatterns of Q.
Q: (# V:< S #) Q: (# V:< S0(# E #) #) Q: (# V:< (# E #) #)
In the first of the three forms, we say that the virtual V is qualified by the pattern S, in the second and third forms, we say that V is directly qualified.
The means by which a virtual attribute V of a pattern Q is extended in a subpattern P of Q.
P: Q(# V::< S1 #) P: Q(# V::< S1(# E #) #) P: Q(# V::< (# E #) #)
S1, S1(# E #), or (# E #) is called the extended descriptor of V. If we're using either the first or second form, and if V is qualified by S in the pattern Q, then S1 must be a subpattern of S. In the case of the third form there are no constraints on Q's declaration of V. If X is an instance of P, then X.V specialises (that is, adds properties to) the definition of V in Q. Note that V is now a virtual pattern in P (as well as Q) and can continue to be further bound in subpatterns of P.
The means by which a virtual attribute V of a pattern Q is extended in a subpattern P of Q, and at the same time made non-virtual.
R: P(# V:: S2 #) R: P(# V:: S2(# E #) #) R: P(# V:: (# E #) #)
Final binding is identical to further binding, except that with final binding, V is no longer virtual.
|BETA Language Introduction||© 1994-2004 Mjølner Informatics||
[Modified: Monday October 23rd 2000 at 14:11]