COMS W4115
Programming Languages and Translators
Lecture 13: Parsing Action Conflicts
October 21, 2009
Lecture Outline
- Review
- Parsing action conflicts
- Resolving shift/reduce conflicts
- Using Yacc to generate LALR(1) parsers
- Using Yacc with ambiguous grammars
- Error recovery in Yacc
- Reading
1. Review
- LR(1) parsing
- Constructing simple LR(1) parsers
- DFA for viable prefixes
2. Parsing Action Conflicts
- If a grammar 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.
3. Resolving Shift/Reduce Conflicts
- Example. Given the augmented grammar G'
(0) E' → E
(1) E → E+E
(2) E → E*E
(3) E → id
FOLLOW(E) = { +, *, $}
Note that this grammar does not specify the relative precedence
of the + and * operators
nor their associativities.
C, the canonical collection of sets of LR(0) items for G', is
I0: E' → ·E
E → ·E+E
E → ·E*E
E → ·id
I1: E' → E·
E → E·+E
E → E·*E
I2: E → id·
I3: E → E+·E
E → ·E+E
E → ·E*E
E → ·id
I4: E → E*·E
E → ·E+E
E → ·E*E
E → ·id
I5: E → E+E·
E → E·+E
E → E·*E
I6: E → E*E·
E → E·+E
E → E·*E
SLR(1) parsing table for G':
State |
Action |
Goto |
id |
+ |
* |
$ |
S |
0 |
s2 |
|
|
|
1 |
1 |
|
s3 |
s4 |
acc |
|
2 |
|
r3 |
r3 |
r3 |
|
3 |
s2 |
|
|
|
5 |
4 |
s2 |
|
|
|
6 |
5 |
|
s3
r1 |
s4
r1 |
r1 |
|
6 |
|
s3
r2 |
s4
r2 |
r2 |
|
Notes
- There is a shift/reduce conflict in
action[5,+] between
"shift 3" and "reduce by E → E+E"
because the associativity of the operator + is not defined
by the grammar. This conflict can be resolved in favor of
"reduce by E → E+E" if we want +
to be left associative.
- There is a shift/reduce conflict in
action[5,*]
between "shift 4" and "reduce by E → E+E"
because the relative precedence of the operators + and
* is not defined by the grammar. This conflict can be
resolved in favor of "shift 4" if we want *
to have higher precedence than +.
- Analogously, there is a shift/reduce conflict in
action[6,+]
between "shift 3" and "reduce by E → E*E"
because the relative precedence of the operators + and
* is not defined by the grammar. This conflict can be
resolved in favor of "reduce by E → E*E" if we want
* to have higher precedence than +.
- There is a shift/reduce conflict in
action[6,*] between
"shift 4" and "reduce by E → E*E"
because the associativity of the operator * is not defined
by the grammar. This conflict can be resolved in favor of
"reduce by E → E*E" if we want * to be
left associative.
4. Using Yacc to Generate LALR(1) Parsers
- Consider the yacc program
expr1.y:
%token id
%%
E : E '+' E
| E '*' E
| id
;
Invoking yacc -v expr1.y, we can see the kernels of the
sets of items for this grammar in the yacc output file y.output.
The parsing action conflicts are shown for states 5 and 6.
state 0
$accept : _E $end
id shift 2
. error
E goto 1
state 1
$accept : E_$end
E : E_+ E
E : E_* E
$end accept
+ shift 3
* shift 4
. error
state 2
E : id_ (3)
. reduce 3
state 3
E : E +_E
id shift 2
. error
E goto 5
state 4
E : E *_E
id shift 2
. error
E goto 6
5: shift/reduce conflict (shift 3, red'n 1) on +
5: shift/reduce conflict (shift 4, red'n 1) on *
state 5
E : E_+ E
E : E + E_ (1)
E : E_* E
+ shift 3
* shift 4
. reduce 1
6: shift/reduce conflict (shift 3, red'n 2) on +
6: shift/reduce conflict (shift 4, red'n 2) on *
state 6
E : E_+ E
E : E_* E
E : E * E_ (2)
+ shift 3
* shift 4
. reduce 2
5/127 terminals, 1/600 nonterminals
4/300 grammar rules, 7/1000 states
4 shift/reduce, 0 reduce/reduce conflicts reported
4/601 working sets used
memory: states,etc. 17/2000, parser 2/4000
3/3001 distinct lookahead sets
0 extra closures
9 shift entries, 1 exceptions
3 goto entries
0 entries saved by goto default
Optimizer space used: input 25/2000, output 9/4000
9 table entries, 3 zero
maximum spread: 257, maximum offset: 257
5. Using Yacc with Ambiguous Grammars
- We can specify the associativities and relative precedence
of the
+ and * operators in the
declarations sections of a yacc program.
- Consider the yacc program
expr2.y:
- The statement
%left '+' makes + left associative.
- The statement
%left '*' makes * left associative.
- Since the statement for
+ comes before the statement for
*, + has lower precedence than *.
%token id
%left '+'
%left '*'
%%
E : E '+' E
| E '*' E
| id
;
Invoking yacc -v expr2.y, we can now see that
no shift-reduce conflicts have been generated in
the yacc output file y.output.
state 0
$accept : _E $end
id shift 2
. error
E goto 1
state 1
$accept : E_$end
E : E_+ E
E : E_* E
$end accept
+ shift 3
* shift 4
. error
state 2
E : id_ (3)
. reduce 3
state 3
E : E +_E
id shift 2
. error
E goto 5
state 4
E : E *_E
id shift 2
. error
E goto 6
state 5
E : E_+ E
E : E + E_ (1)
E : E_* E
* shift 4
. reduce 1
state 6
E : E_+ E
E : E_* E
E : E * E_ (2)
. reduce 2
5/127 terminals, 1/600 nonterminals
4/300 grammar rules, 7/1000 states
0 shift/reduce, 0 reduce/reduce conflicts reported
4/601 working sets used
memory: states,etc. 17/2000, parser 2/4000
3/3001 distinct lookahead sets
0 extra closures
6 shift entries, 1 exceptions
3 goto entries
0 entries saved by goto default
Optimizer space used: input 19/2000, output 10/4000
10 table entries, 3 zero
maximum spread: 257, maximum offset: 257
Unless otherwise instructed, Yacc resolves all parsing
action conflicts using the following two rules:
- A reduce/reduce conflict is resolved by choosing the
conflicting production listed first in the Yacc
specification.
- A shift/reduce conflict is resolved in favor of shift.
Note that this rule correctly resolves the shift/reduce conflict
arising from the dangling-else ambiguity.
6. Error Recovery in Yacc
- In Yacc, error recovery can be performed with error productions.
- To process errors at the level of a nonterminal
A,
add an error production A → error α
where error is a Yacc reserved word, and α is a string
of grammar symbols, possibly empty. Yacc will generate a parser
including this production.
- If Yacc encounters an error during parsing, it pops symbols from its stack
until it finds the topmost state on its stack whose underlying set of items
includes an item of the form
-
A → . error α
The parser then shifts the special token error onto the stack
as though it saw the token error on the input.
- If α is not empty, Yacc skips ahead on the input looking for a
substring that can be "reduced" to α on the stack. Now the parser
will have
error α on top of its stack, which it will
reduce to A. The parser then resumes normal parsing.
- Example: See Fig. 4.61, p. 296. This Yacc specification contains
a translation rule
lines : error '\n' { yyerror("reenter previous line:"); yyerrok; }
On encountering an error, the parser starts popping symbols from
its stack until it encounters a state with a shift action on error.
The parser shifts the token error on to the stack and
then skips ahead in the input until it finds a newline which it shifts
onto the stack. The parser reduces error '\n' to
lines and emits the error message "reenter previous line:".
The special Yacc routine yyerrok resets the parser to its
normal mode of operation.
7. Reading
- ALSU, Sects. 4.8 and 4.9
- The first line in the Lex specification in Fig. 4.60 (p.295)
should not have an
e after the two
backslashes:
number [0-9]+\.?|[0-9]*\.[0-9]+
aho@cs.columbia.edu