Answer: (iii) Canonical LR

Explanation

Parsers can be ordered by the size of the grammar class they can handle (their power):

• Operator-precedence parsing: only works on operator grammars (no $\epsilon$, no two adjacent nonterminals) — the least general.
• SLR(1): uses the LR(0) items but decides reduce actions using the broad $\mathrm{FOLLOW}$ set of the left-hand nonterminal, ignoring the specific context of each item → weaker.
• LALR(1): merges LR(1) items that share the same core (same item, different lookahead). More powerful than SLR(1).
• Canonical LR / LR(1): keeps every lookahead separate, producing the most states and therefore accepting the largest set of grammars among the given options.

Hence the most powerful parser is:

$$\boxed{\text{(iii) Canonical LR (LR(1))}}$$

Order of power (weakest → strongest):
Operator-precedence < SLR(1) < LALR(1) < Canonical LR(1)