COMS W4115
Programming Languages and Translators
Lecture 12: Bottom-Up Parsing
October 19, 2009

Lecture Outline

1. Bottom-up parsing
2. LR(1) parsing
3. Constructing a simple LR(1) parser
4. DFA for viable prefixes
5. Parsing action conflicts

1. Review

• FIRST and FOLLOW
• How to construct a predictive parsing table
• LL(1) grammars

2. Bottom-up Parsing

• Bottom-up parsing can be viewed as trying to find a rightmost derivation in reverse for an input string.
• A handle is the rightmost substring in right-sentential form that matches the body of a production and whose reduction by that production represents one step in the reverse of the rightmost derivation.
• Consider the grammar G:

      (1) S → S ( S )
      (2) S → ε
   

• The handles in a rightmost derivation for the input string ( ) ( ):

      S ⇒ S ( S )      // handle is S ( S )

        ⇒ S ( )        // handle is the empty string between ( )

        ⇒ S ( S ) ( )  // handle is S ( S )

        ⇒ S ( ) ( )    // handle is the empty string between first ( )

        ⇒ ( ) ( )      // handle is the empty string prefix
  

• Shift-reduce parsing is a form of bottom-up parsing in which we shift terminal symbols of the string to be parsed onto a stack until a handle appears on top of the stack. We then replace the handle by the nonterminal symbol on the left-hand side of the associated production (this is a "reduce" action). We keep repeating this process until we have reduced the input string to the start symbol of the grammar. This process simulates the reverse of a rightmost derivation for the input string. Thus, we can think of shift-reduce parsing as "handle pruning."

3. LR(1) Parsing

• Model of an LR(1) parser (Fig. 4.35).
• "L" means left-to-right scanning of the input, the "R" means constructing a rightmost derivation in reverse, and the "1" means one symbol of lookahead in making parsing decisions.
• LR parsing table for G:



State	Action			Goto
	(	)	$	S
0	r2	r2	r2	1
1	s2		acc
2	r2	r2	r2	3
3	s2	s4
4	r1	r1	r1

r2 means reduce the handle on top of the stack by production (2) S → ε.
s2 means shift the input symbol on the stack and then push state 2 on top of the stack.
acc means accept and stop parsing.
Goto[0,S] = 1 means push state 1 on top of the stack after reducing a handle to the nonterminal S in state 0.
A blank entry means report a syntax error.


• Moves made by an LR(1) parser on input ( ) ( ) [Alg. 4.44].

        Stack      Input           Action
        0          ()()$    reduce by production (2) S → ε and then push state 1 on the stack
        0S1        ()()$    shift input symbol ( on the stack and then push state 2 on the stack
        0S1(2       )()$    reduce by (2) S → ε and push state 3
        0S1(2S3     )()$    shift ( and push state 2
        0S1(2S3)4    ()$    reduce by (1) S → S(S) and push state 1
        0S1          ()$    shift ( and push state 2
        0S1(2         )$    reduce by (2) S → ε and push state 3
        0S1(2S3       )$    shift ) and push state 4
        0S1(2S3)4      $    reduce by (1) S → S(S) and push state 1
        0S1            $    accept       
   

• Note that an LR parser is a shift-reduce parser that traces out a rightmost derivation in reverse.

4. Constructing a Simple LR(1) Parsing Table for a Grammar

• An LR(0) item of a grammar is a production of the grammar with a dot at some position of the right side. E.g., S → ·S(S), S → S·(S), or S → S(S)·.
• We will use two functions to construct the sets of items for a grammar:
• closure(I), where I is a set of items, is the set of items constructed by the following two rules:
1. Initially, put every item in I into closure(I).
2. If A → α·Bβ is in closure(I) and B → γ is a production, then add the item B → ·β to closure(I) if it is not already there. Keep repeating this step until no more new items can be added to I.
• goto(I, X), where I is a set of items and X is a grammar symbol, is the closure of the set of all items A → αX·β where A → α·Xβ is in I.
• An augmented grammar G' is one to which we have added a new starting production S' → S where S is the start symbol of the given grammar G. Reducing by the new starting production signals acceptance of the input string being parsed. We will always augment a grammar when we construct an SLR parsing table for it.
• The sets-of-items construction
  • Input: An augmented grammar G'.
  • Output: C, the canonical collection of sets of LR(0) items for G'.
  • Method:

  
     I0 = closure({[S' → ·S]});
     C = {I0};
     repeat
       for each set of items I in C and grammar symbol X such that
         goto(I,X) is not empty and not in C do
           add goto(I,X) to C;
     until no more sets of items can be added to C;
     

• Example: Given the augmented grammar G'

  S' → S
  S  → S(S)
  S  → ε
  

C, the canonical collection of sets of LR(0) items for G', is

  I0: S' → ·S
      S → ·S(S)
      S → ·

  I1: S' → S·
      S → S·(S)

  I2: S → S(·S)
      S → ·S(S)
      S → ·

  I3: S → S(S·)
      S → S·(S)

  I4: S' → S(S)·
  

• Algorithm to construct the SLR(1) parsing table from C, the canonical collection of sets of LR(0) items for an augmented grammar G'
• Input: C = {I0, I1, ... , In}.
• Output: The SLR parsing table functions action and goto.
• Method:
• State i and its action and goto functions are constructed from Ii as follows:
• If item [A→ α·aβ] is in Ii and goto(Ii, a) = Ij, then add "shift j" to action[i, a]. Here a is a terminal.
• If item [A → α·] is in Ii, then add "reduce A → α" to action[i, a] for all a in FOLLOW(A). Here A cannot be S'.
• If item [S' → S·] is in Ii, then add "accept" to action[i, $].
• If goto(Ii, A) = Ij, then in the parsing table set goto[i, A] = j.
• The initial state of the parser is constructed from the set of items containing [S' → ·S].
• Notes:
• If each parsing table entry has at most one action, then the grammar is said to be SLR(1). If any entry has more than one action, then the algorithm fails to produce a parser.
• All undefined entries are made error.
• Example: the LR parsing table above is an SLR(1) parsing table for the balanced-parentheses grammar.

5. DFA for Viable Prefixes

• A viable prefix is a prefix of a right sentential form that does not continue past the right end of the rightmost handle of that sentential form.
• The shift and goto functions of the canonical collection of sets of LR(0) items for a grammar G define a DFA that recognizes the viable prefixes of G.
• An item [A → β·γ] is valid for a viable prefix αβ if there is a rightmost derivation from S' to αAw to αβγw.

6. Parsing Action Conflicts

• If a grammar G is not SLR(1), the SLR parsing table construction produces one or more multiply defined entries in the parsing table action function.
• These entries are either shift-reduce conflicts or reduce-reduce conflicts.

7. Reading

• ALSU, Sects. 4.5, 4.6.