On Sun, 1 Apr 2007, Hans Aberg wrote: > If there is a choice. And which variation of LR(1) would you prefer, if there > now is a difference: One that compacts the states and still has the same > problem as LALR(1) that when an error token comes by, some actions can be > performed before the error is being reported plus it cannot be used in > interactive scanners to report possible token completions,
I'm focusing on that one. > or one that does > not have those flaws, but takes up perhaps much more space? I'm thinking there could be an option to modify the definition of state compatibility so that Bison can generate either canonical or compact LR(1) tables. Assuming the compact LR(1) works out as I hope, canonical LR(1) would be interesting to me purely for an efficiency comparison. I'm very much still in the planning stages, so don't hold me to any of this. _______________________________________________ [email protected] http://lists.gnu.org/mailman/listinfo/help-bison
