WG211/M20Reichenbach

Joint work with Alexandru Dura, Hampus Balldin, Emma Söderberg

Datalog and its relatives have emerged as useful tools for concisely expressing highly efficient program analyses. However, to build a practical program analyses in Datalog today, writing Datalog code alone is not enough: we must also hand-code "fact extractors" that translate the source language into database tables that are palatable to Datalog engines like Soufflé.

MetaDL unifies both of these components by making syntactic patterns over the source language available within Datalog rules. Our curent implementation allows us to analyse Datalog within Datalog, but (time permitting) I will also discuss ongoing work on support for other languages as well as for program transformation.