Symbol Table Management

7/28/2019 Symbol Table Management

1/7


jcsites.juniata.edu/faculty/rhodes/lt/sytbmgmt.htm 1/7

Symbol Table Managementlast updated 5/11/07

Screen Mode

Central part of compiler

The symbol table is accessed by most phases of a compiler, beginning with the lexicalanalysis to optimization.

The symbol table carries the collected information about each named object in the

program to other phases of the compiler. As soon as a named token is found,depending on the token type, a call to the symbol table routines must be made.

Tokens:

Keywords of the language, special characters/punctuation, user defined identifiersused for different meanings, constants

Identifiers:Identifiers are user defined

variables, memory location labelsparameter namesnames of subroutines (functions or procedures) which are associated with entrypoints of the routines' codearray or structure names

field names; file names; enumerated type valuesconstant names; type names

Need a global structure to store identifiers and their associated meanings (attributes)for future reference. This structure provides more information that the grammar andsyntax directed translation of the language would not normally be able to carry.

Reserved words

These may be stored here too. You have to look up an identifier and reserved wordshave the same syntax as an identifier.


2/7



Scoping rules

The symbol table must be able to handle the scoping rules of the language.

Symbol Table Requirements

fast lookup . The parser of a compiler is linear in speed. The table lookup needs to beas fast as possible.

flexible in structure. The entries must contain all necessary information, dependingon usage of identifier

efficient use of space . This will require runtime allocation (dynamic) of the space.

handle characteristics of language (e.g., scoping, implicit declaration)

Scoping requires handling entry of a local block and exit.Block exit requires removal or hiding of the entries of the block.

Program components

Declarative statements - define identifiers and their attributes

Imperative statements - uses identifiers assuming their attributes

Languages and their explicit declarative statements

Pascal, Ada VAR statement

C, C++, Java, PL/Istatements beginning with atype

FORTRAN TYPE, DIMENSION statementCOBOL DATA DIVISION


3/7



Implicit Declarations

Some languages have implicit declarations

FORTRAN or PL/I SUM = SUM + I (first encounter assumes declaration with defaultattributes)

APL, BASIC, LISP (most interpreted languages)

Actions on declarations

Declarative statements generally do not translate to (or directly associated with) anyexecutable code. They may allocate space at designated times, however.

Symbols can be monitored, especially in languages that require declaration statements

FIRST ENCOUNTER SUBSEQUENTENCOUNTER CONCLUSION

Declaration Reference (in imperativestmt)Continue--this is the

expected case

Declaration DeclarationMultiple declarations error unless appropriate within

scoping rulesDeclaration None Unused warning message

Reference (in imperativestatement) NA Undefined error

String Management

How to efficiently store names of identifiers


4/7



1. Restrict length of identifier

FORTRAN 77 and earlier limited variables to 1-6 characters (4 bytes packedencoding)Pascal had 1-10 characters in initial designAdvantages : structure consists of simple fixed sized string array; easy lookup

Disadvantages : limits programmer to choose effective names; may waste space

2. Separate String Space

Advantages : allows unlimited identifier names; doesn't waste spaceDisadvantages : extra memory reference for access

3. Dynamic allocation of strings

Advantages : don't need predefined space for strings

Disadvantages : complexity

Extend to other parts of the symbol table

Name Searching

Functions:

Declaration of name- enter new name into table- return error if already there- add new attributes as found

Use- expect name to be found- return position of entry in table

Entry of new scope - allow new declarations or redefined declarations

Exit of scope - delete entries

Goals:


5/7



1. efficiency of declaration entry, insertion of new name2. efficiency of retrieval, lookup3. sorted list of symbol table dump

Most important would be #2

Approaches:

n is the number of entries in the symbol table for the analysis below

n' is the number of entries in the current [local] block

1. Linear access

Assuming error free source fileTime(declaration) = k or O(1) [k is some constant]Time (reference) kn/2 or O(n)

To check for errorsTime(declaration) ~= kn'/2Time(reference) ~= kn/2

Time(block entry) = O(1)Time(block exit) = O(n') or O(1) if display usedTime(sort) = O(n log n)

2. Binary search access

data structure same as linear tableas new name is entered, insert into block to maintain sorted order on lookup use binary searchTime(declaration) = O(n')Time(ref) = O(log n')Time(sort) = 0 (already in order)Time(entry) = O(1)Time(exit) = O(1)


6/7



3. Tree access

- assumes a more random encounter of namesB A C P C D B Q C D E- entries of symbol table are organized into tree structure- have forest of trees - each tree constitutes a block

Average Time(decl) = O(log n)Average Time (ref) = O(log n)Worst case Time(decl) = T(ref) = O(n)Time(sort) = 0Time(entry) = O(1)Time(exit) = O(n')

4. Hash tables

use a hash function on name to find location in symbol tablesymbol table is organized like a table of stacks since some identifiers in differentblock will collideTime(declaration) = O(1) or O(n')Time(reference) = O(b) where b is the chain depth

Time(entry) = O(1)Time(exit) = O(n) or O(n') if second link connects block's ids

Single pass vs multipass compiler:

single pass can discard the entries while multipass will need to retain temporarily


7/7



inaccessible entries

Symbol Table Management

Documents

Transcript of Symbol Table Management