I'm reading the Dragon book. The beginning describes recursive descent parsers. Then it states that a predictive parser is the table version of a RD parser.
I also read up on LALR(1) parsers.
The differences between predictive and LALR parsers seem superficial to me. LALR seems to consume terminals quickly, but the predictive parser will do that too if the grammar is properly prepared (for example, by converting it to Greibach Normal Form.)
What am I missing? What is LALR's advantage?