Tight Integration of Java and Petri Nets

Olaf Kummer Universität Hamburg, Fa hberei h Informatik Vogt-Kölln-Straÿe 30, D-22527 Hamburg kummer informatik.uni-hamburg.de Abstra t In this paper we investigate how the Java programming language an be integrated in an obje t-oriented Petri net formalism. It be omes apparent how a arefully designed uni ation algorithm supports this goal, despite the fa t that Java itself is not based on uni ation. Some dedi ated algorithms for a Petri net simulator are proposed. 1 Introdu tion The basi net formalism dis ussed in this paper are referen e nets. In referen e nets, a net an be instantiated arbitrarily often and referen es to net instan es an be used as tokens in other net instan es. For details see [6℄. Referen e nets use a variant of Java as their ins ription language. See [2℄ for the original Java spe i ation. This paper an be seen as a ompanion to the earlier work [4℄, where the referen e net formalism and the large s ale ar hite ture of the Petri net simulator Renew are des ribed. Here we will dis uss some further aspe ts that were only very brie y sket hed in [4℄. 2 Tuples The Java language has no support for tuples. The only ase where tuples would ome in handy in Java would be multi-value returns. However, these o ur not very often and an be substituted by the return of a omplex obje t that aggregates the return values. In other ases, when we have multiple values that should be pro essed together, we an simply use multiple variables that hold one value ea h. But for Petri nets, tuples are more important. It is dangerous to store related values in multiple pla es, be ause, if multiple tokens have to be put there, tokens for di erent data sets an be mixed. Tuples solve this problem. Java's syntax o ers two opportunities for adding tuples. Either a tuple is written (1,2,3) or [1,2,3℄. The rst form has the disadvantage that it is impossible to write 1-tuples, be ause they ould be mistaken for grouping operations, e.g. "string"+(true|false)would be ambiguous. Hen e it was de ided to use the square bra ket notation. However, when we want to support lists as well as tuples, we have to ll both synta ti ni hes. In that ase, 1-tuples might be onsidered dispensable, but lists of length 1 would be mandatory, so that the de ision would have been the other way round.