Inter-language bridging is an important issue of scripting language design and implementation. Most of the popular languages such as Python, Perl, Tcl, and Ruby use a bridging approach based on wrappers that are written in the external language (usually C/C++) and serve as a glue layer between the languages. This allows a wide flexibility in defining the glue abstractions, but it requires the user to specify them on the level of the implementation language, and it therefore impairs the higher-level scripting process. In contrast, the first implementations of JPiccola, a scripting and composition language implemented in Java, use a generic bridging strategy based on information provided by Java’s runtime introspection facilities. This strategy makes accessing of external objects more lightweight, but it does not provide the necessary means of abstraction and leads to a very tight coupling between the two language levels. In this thesis, we present a new bridging strategy for Piccola that combines the advantages of the two approaches. We minimize the bridging functionality that is hardcoded in the virtual machine by making it a meta-aspect of the language Piccola.

Inter-language bridging is an important issue of scripting language design and implementation. Most of the popular languages such as Python, Perl, Tcl, and Ruby use a bridging approach based on wrappers that are written in the external language (usually C/C++) and serve as a glue layer between the languages. This allows a wide flexibility in defining the glue abstractions, but it requires the user to specify them on the level of the implementation language, and it therefore impairs the higher-level scripting process. In contrast, the first implementations of JPiccola, a scripting and composition language implemented in Java, use a generic bridging strategy based on information provided by Java’s runtime introspection facilities. This strategy makes accessing of external objects more lightweight, but it does not provide the necessary means of abstraction and leads to a very tight coupling between the two language levels. In this thesis, we present a new bridging strategy for Piccola that combines the advantages of the two approaches. We minimize the bridging functionality that is hardcoded in the virtual machine by making it a meta-aspect of the language Piccola. This allows

Scripting and composition languages offer high-level mechanisms to combineam compose services provided bya lower-level host progra -kB" la-kB"5 Inter-lark42q bridgingmechagkB1 ae therefore needed toma hostlatk4W( entitiesat services to akB"5(1kC5q of the scripting laiptinga vice versa Ma ypopula lalaC52 sucha Python, Perl,ar Ruby usea bridging aidging bagi on wrapperstha must be written or generakB in the hostlak414"" Otherlaerk4q2 like Jythonat Ka wa aB" a fixed bridgingstran gytha exploits reflect8S feaa1k provided by the hostlak41(11 Although both of theseasekBB"2( a usasek theya" cumbersomeau low-level. Inpa422"kCW , it ca be very difficult toa""( hostlatk42q services tocooperak searakCW with ahkq1(BkCW5 of the scripting laipting In this we presenta lightweight bridgingstrai gy for scripting ag compositionlapositi tha simplifies the tak ofakB5B41 host latk5"( services to theak4W"5qkCW level of the scriptinglai guaip Thisstra" gy uses introspection fa4"5(WkC of the hostlatkB54 toak54(Bk thewra115" process, while providinga hook for progra -kB-definedanedk41q5 of thegeneraB4 interfate Ameta52 vel bridgingladg is responsible forwra45q" aa unwra"2kC both host ao scriptinglaiptin entities so theyca sea1(kC2q coopera2q The bridgingstran gy employspa4q2( eva4""11k ofwra25(5 aa unwra(5kC opera(5kC toak42 veaqW(2kC22 performance.