1. COP 4620 / 5625 Programming Language Translation / Compiler Writing Fall 2003 Lecture 5, 09/25/2003
Prof. Roy Levow

2. LL(1) Parser Generation
Roadmap
LL(1) parsing
LL(1) conflicts as an asset
LL(1) conflicts as a liability
The LL(1) push-down automaton
LL(1) error handling
LL gen

3. LL(1) Parsing
To determine choice of production, we use FIRST and FOLLOW sets
FIRST set
Contains tokens that can occur as the start of a production or the tail of a production
FOLLOW set
Contains tokens that can follow a nullable prefix in a production

9. FOLLOW Set

10. FIRST & FOLLOW Sets for Sample Grammar

11. LL(1) Conflicts First/First conflict First/Follow conflict
Two alternatives share some first choice
First/Follow conflict
There is a non-null alternative that shares a choice with the follow for a nullable alternative
Two nullable alternatives
We do not consider Follow/Follow conflicts since they will be seen also as one of the other conflict cases

12. Conditions to be LL(1) No First/First conflicts
No First/Follow conflicts
No multiple nullable alternatives
If we can remove all conflicts, we may be able to produce a LL(1) grammar from one that is not

13. Making a Grammar LL(1) Left-factoring
Factor out the common prefix into a separate production
Example:
term -> IDENT | IDENT ‘[‘ expr ‘]’ | …
Becomes
term -> IDENT (‘[‘ expr ‘]’ )? | …

14. Making a Grammar LL(1).2 Substitution
Replace non-terminal on RHS with its
Example (indirect First/First conflict)
term -> IDENT | indexed | …
indexed -> IDENT ‘[‘ expr ‘]’
Substitute and then left factor

15. Making a Grammar LL(1).3 Left Recursion Removal
Direct: a -> a b
Indirect: a -> b c; b -> a d
Hidden: a -> c a; c -> ε
In many cases, left recursion can be replaced by right recursion either by using a standard transformation or by studying the construct and developing an alternative

16. Correcting Semantic Effects
Restructuring a grammar may change the implied semantics, interfering with translation
Can sometimes resolve the problem by adding “marker rules” that are transformed along with the original grammar
See p.131

17. Automatic Conflict Resolution
Always choose first alternative
Use LL(2), look ahead an extra token
Dynamic resolution
Add decision rule to first alternative
Example
else-tail-option -> %if (1) ‘else’ stmt | ε

18. Table Driven LL(1) Parsing
Uses a push-down automaton
Finite State Automaton with a stack
Start with start symbol on stack
When processing a non-terminal
Pop non-terminal
Push selected RHS in reverse order
When processing a terminal
Match with input

19. Table Driven LL(1) Parsing.2

20. LL(1) PDA Operation

21. LL(1) Error Handling Concerns Possible approaches
Avoid corrupt syntax tree
Avoid infinite loops
Possible approaches
Insert needed token
May loop
Skip input until match
Pop items from stack until match
May produce corrupt syntax tree

22. LL(1) Error Handling.2
Acceptable set method is a combination of skipping input and popping items from stack
Discard input until acceptable continuation token is found
Resynchronize parser based on token found
Acceptable sets are restart points for each production
Set used is union for pending productions

23. LLgen Parser generator from Amsterdam Compiler Kit
Tanenbaum et al., 1983