Thesis on type inference for object-oriented languages available

I am pleased to announce the publication of my thesis:


                       Andreas V. Hense
                 Universit"at des Saarlandes, 1994

        (Supervisors: Prof. Dr. G. Smolka, Prof. Dr. R. Wilhelm)

Object-oriented programming languages enjoy an ever growing
popularity.  These languages promise an increased productivity by
favorizing code reuse.  Object-oriented languages undoubtedly have the
reputation of leading to main- tainable and expandable systems.
However, they are not regarded as very safe in the sense that
object-oriented programs are free of errors.  They are well- suited
for rapid prototyping, not for end products. A major part of the
errors in a dynamically typed language are type errors, in
object-oriented terminol- ogy they translate by "message not
understood".  Such errors cannot always be eliminated by testing.
Static type inference analyzes programs at compile time and guarantees
their absence at runtime. Whereas monomorphic type dis- ciplines
restrain reusability, polymorphic type inference provides the
necessary flexibility.

Despite its apparent urgency, the problem of inferring types for
object- oriented languages has still not been solved satisfactorily.
This work is con- cerned with the type inference for a complete
object-oriented language, thus showing the multiple facets of the
problem.  The type inferencer is of direct practical use but the work
shows the limitations of the chosen approach as well.  It is explored
what can be done with pure inference, i.e. without having to rely on
type declarations. Pure type inference can be used for existing
dynamically typed languages and can be regarded as an optional

The object-oriented language was translated into an intermediate
language by an intelligent translation function.  Thus, the most
promising of the exist- ing type inference results on small calculi
could be chosen and adapted.  The polymorphic type discipline of this
work comprises extensible record types, re- cursive types, and
imperative types. Since existing approaches were developed further,
the correctness was proven using order-sorted logic. Order-sorted
logic is used for the first time in this context. Its increased
expressivity benefits the formulation of the type inference algorithm.

The thesis is available as a book


author:     Andreas V. Hense

publisher:  1994, Pirrot Verlag
            Trierer Str. 7
            D-66125 Saarbruecken-Dudweiler
            ISBN 3-930714-00-0

price:      DM 30      (approx. 20 US $)