For some authors (Laurence Tratt), full LR parsing (no LALR or SLR) is the best approach, because it is: fast, static-check, good error messages. But only if the parser generator is using new parsing tecniques that are able to reduce a lot the RAM used from an LR parser. LR parser of the past were not much practical, and instead LALR were used.
TODO study better It seems a new approch for incremental parsing during development. Its main objective is producing an AST.