Hi Thierry, On 10/9/2014 4:52 PM, Thierry Goubier wrote: > > I'm currently on a medium to long term quest to learn in depth / maybe > improve the AST automatic generation in SmaCC; > it should give me a core foundation (a common API for the ASTs for all > langages that SmaCC can parse) to build upon > interactive visualisations IDEs for the programming languages I'm getting > involved in: yesterday C, now R, tomorrow (I > hope) Fortran and Python. But this is more about static analysis than runtime > behavior analysis, unless I turn to > VectorFabrics underlying tech... Hum, this one is ... interesting. It opens > new ideas. Thanks a lot, this is really > great! > > Thierry
By coincidence (or maybe not :) ), I have also fiddled with SmaCC (and, funny thing, I am also interested in static analysis :) ). I ported it to VisualAge, maybe we should exchange notes. I am still not happy with the result, for the previous version I had optimized it to use a VA-specific primitive (Character>>testBit: - very useful little primitive, the VA scanner also uses it (I suspect they implemented it specifically for that)) instead of is... methods, and I did not get to do it yet for the current version. This is actually not just an optimization, I consider it a bug-fix as well - one cannot rely on every method in Character that happens to start with "is" to be fast. Plus you don't know in which environment you are using SmaCC and what kind of extensions people have implemented there. I know that I hit some really bad such methods (now I don't remember if it was ours or some third-party extension). Plus, you are generating a parser that relies on methods that may not be there at runtime. Anyway, I also have some ideas for what I think would be some improvements for SmaCC. Too many ideas, too little time (: Florin