This is a natural generalization of SLR Parser Actions.

Let K be parser state at the end of stack and t current token:

• if there is a transition for non-empty
  • shift: push t onto stack
  • if t is EOF, then accept
• reduce: if item K then
  • pop p from stack
  • push X onto stack
• if none possible, report syntax error
• if both possible, shift-reduce conflict: grammar not LR(1)

Theory of parsing, from page 372

Note, among other properties that if a language can be parsed by a deterministic pushdown automaton, then there is a grammar for this language that can be parsed by an LR(1) automaton. (Such grammar may be much bigger than the alternative non-deterministic grammar.)