Types, Bindings and Parameters in a Persistent Environment

Abstract

Our experience of persistent data in PS-algol and surveys we have conducted have led us to identify various requirements on the stores embodied in the design of a programming language. This leads to a search for constructs that handle activities currently not supported. The requirements identified for persistent data include: the rights to transience or longevity are completely independent of the type of data; the data types should include some form of inheritance to model specialisation; and the types and bindings used should allow system evolution on an incremental and localised basis. Currently we are engaged in the design of a new language, Napier, to succeed PS-algol. This paper presents some aspects of the design of the binding and type checking we are considering. These particularly address the second and third requirements above. The third requirement is equivalent to the requirement for data independence: schema editing without having to recode, recompile or relink all application programs, a requirement identified by the database community in the early 1970s. It is addressed in our design by controlled and localised incremental binding and type checking, in the context of a structure we call name spaces. In compliance with the first requirement, we do not constrain its use to long term data, which implies that it may also be used to allow parts of a program to evolve, and may be used to construct flexible interfaces within a program. We are able to show that these name spaces, structures, records, abstract data types and relations may all be put into a consistent framework described in terms of associating a formal store with an actual store, similar to parameter mechanisms. We also show that this enables, at least in principle, many activities previously dealt with outside the programming language, to be supported from within it.