OberonX

Roberto Morelli

Diploma Thesis Winter 1996
Supervisor: P.Reali
Institute for Computer Systems, ETH Zürich


Objectives

The OberonX programming language is an experimantal further development of Oberon towards the Oberon-XSC pioneered in Karlsruhe. OberonX holds new constructs, which improve the Oberon abstraction capabilities. The programming of methematical software should be made easier by the introduction of new language extensions. A new array concept is implemented, which introduces a new semantic and syntactic definition, allowing lower bounds, semi-dynamic and truly dynamic arrays in the language. Overloaded operators, in combiantion with complex return values, increase the flexibility of the language by removing the former restrictions.

Results

Arrays

The most relevant features regard the possibility of indexing intervals specified by lower bounds and lengths, of allowing open dynamic components inside static ones, of using expressions to define the interval. The concept of pointers bounded to arrays is no longer valid : the compiler itself decides if an array block is to be considered static, semi-dynamic, dynamic or open: this allows the definition of truly dynamic arrays. The possibility of asymmetric multidimensional matrices declaration is also provided. Array structures are treated as whole entities with consequence on assignments, parameter passing and returs values. Te new array semantic results to be different from the original one and requires a different point of wiev for the programmer.

Complex Return Values

The return capability of Oberon functions has been extended to all the type structures: no restrictions are imposed. This will increase the flexibility and abstraction of the language when treating complex data types. Nested function calls with complex return types are also allowed.

Operators

The use of a set of terminal symbols belonging to the operator class, allows the declaration of functions usable as infix operators in expression context. Overloading of compiler bounded and user defined operators is therefore allowed. A method to resolve operator identification based on the argument type signatures has been developed. Imported operator qualification is not obligatory, but it can be necessary to avoid operator clashes, whose presence does not allow the program to be compiled. In case of the absence of a well suited operator, a method based on operan type distance computation allows the identification of the best operator available, allowing the use of every Oberon type construct. In case of approximated operator matching, type conversion for the basic data types is automatically executed.


[ Department | Institute | Student Projects ]
ETH Zürich: Institut für Computer Systeme, ETH Zentrum, CH-8092 Zürich
Comments to Patrik Reali <reali@inf.ethz.ch>
April 4, 1997