Symbol Tables

1
Symbol Tables

• Contents and Use
• Construction
• Implementing Scoping Rules

2
Symbol Table Design

• a matter of occult arts

3
Structure of a Compiler
Scanner
Parser
Semantic Analyzer
SYMBOL TABLE
Code Generator
Optimizer
4
Role of Symbol Tables

• The symbol table is one of the most important
  data structures in a compiler.
• Created during front end to hold information on
  programs objects
• IR refers to objects via pointers to the
  corresponding entries in symbol table
• This is done until near the end of compilation
• So symbol table needed for entire processing
• Contents may be modified as part of optimizations

5
Contents of Symbol Tables

• Symbol Tables save information on
• identifiers
• labels
• numerical values (constants)
• character strings
• Temporaries (compiler-generated variables)
• More efficient to store numerical values and
  character strings separately.
• There might be a separate label table

Symbol tables may be created on a per-procedure
basis. There may be separate tables for global
data.
6
Contents of Symbol Tables

• There are a variety of typical kinds of
  identifiers
• Scalar variables, arrays and structures
  (records), procedures and functions, defined
  constants, temporaries
• Some kinds of information stored with identifiers
• The name or value
• Data type
• Size and dimensionality information (for arrays)
• Other scoping and storage information
• Result types, parameters (formal arguments),
  prototypes

The information saved depends on the kind of
object, so need flexible format for entries in
symbol table
7
Type of Data in Symbol Table

• The entry for this array would require a complete
  description of its type, including the type of
  each of its elements.
• Array A 1 .. 100 of mytype
• Mytpe is a record
• username char string
• emailaddress char string
• acctdetails array ( 1 5) of integer
• usage pointer to array ( 112) of reals

8
Implementation of Symbol Tables

• Simple implementations, e.g. as an array, are too
  inefficient in practice
• Most common approach is to use open hashing
• Hash function based on identifier
• A number of good functions are known
• Symbol table entry usually points to another
  structure to save character strings (name of
  variable, char const)
• Enables same size for each entry

9
Role of Hash Function

• The symbol table is heavily accessed.
• Add or delete entries, modify or augment
  information associated with entry, find entry and
  enter reference into IR or other structure
• A good choice of hash function is critical for
  practice.
• Goal reasonably uniform distribution of entries.
  Not too many unused indices.
• A large table is more efficient, but takes more
  space, than a small table.

10
Scope

• A (syntactically defined) region within a
  program within which variables may be declared.
• E.g. procedure, statement block.
• Scoping units may be nested in some languages.

11
Scope of Variable Declaration

• Static Scope
• scope is determined from nesting relationships in
  source program, e.g. Pascal.
• Dynamic scope
• scope is determined by run time calling
  relationships. E.g. Lisp.

12
Lexical vs. Dynamic Scoping
This program calls (twice) show small What is
the output?

• program whichone (input , output )
• var r real
• procedure show
• begin write ( r 53 ) end
• procedure small
• var r real
• begin r 0.125 show end
• begin
• r 0.25
• show small writeln
  
• end

13
Lexical vs. Dynamic Scoping
This program calls (twice) show small What is
the output?

• program whichone (input , output )
• var r real
• procedure show
• begin write ( r 53 ) end
• procedure small
• var r real
• begin r 0.125 show end
• begin
• r 0.25
• show small writeln
  
• end

With lexical scoping, it is 0.250 0.250.
14
Lexical vs. Dynamic Scoping
This program calls (twice) show small What is
the output?

• program whichone (input , output )
• var r real
• procedure show
• begin write ( r 53 ) end
• procedure small
• var r real
• begin r 0.125 show end
• begin
• r 0.25
• show small writeln
  
• end

With lexical scoping, it is 0.250 0.250.
With dynamic scoping, it is 0.250 0.125.
15
The Procedure as a Name Space

• Details of scoping rules are language dependent
• C has global, static, local, and block scopes
  (Fortran-like)
• Blocks can be nested, procedures cannot
• Scheme has global, procedure-wide, and nested
  scopes
• Procedure scope (typically) contains formal
  parameters

Lexical scoping can be determined at compile
time.
16
Lifetime of Variable

• Lifetime is the time during execution of a
  program when a given variable first becomes
  visible to when it is last visible.
• Also called extent.
• The lifetime of a global variable covers the
  entire execution (unless it is temporarily
  superseded in some places).
• The lifetime of a local or automatic variable is
  usually an activation of the program unit within
  which it is declared.
• Some languages permit variables to be local to
  blocks also.

17
Lifetime and Scope Examples

• In C, scope of automatic variable is procedure or
  block where it is declared to end of program unit
• In PL/1 scope encompasses entire relevant program
  unit
• In Pascal, a variable in outermost scope is
  visible everywhere in program except where
  another variable with same name is declared (and
  in the routines it contains)
• Fortran has common blocks, static memory that is
  visible in all routines where it is declared.
  Scope encompasses all those routines.
• Local static variables in C and SAVEd variables
  in Fortran have global lifetime but are only
  visible in certain regions, so scope may be a
  file or a procedure
• Dynamic variables have a lifetime that extends
  from their point of allocation to the point(s) of
  their destruction.

18
Block Scoping of Variables

• Nested blocks
• If variable x is used in block B and declared in
  B, then that declaration is used.
• Else if x is declared in block C surrounding
  block B, but not in any other block within C and
  surrounding B, use declaration of x in block C.

Which x? y?
In general At point p, which declaration of x is
current? How can compiler keep track of this?
19
Scoping

• Compiler has to be sure that it is accessing the
  right variable
• Care must be taken that the object code uses data
  only within its scope
• Two variables with the same name and different
  scopes must both be entered into symbol table
• We will have to organize the symbol table setup
  to take account of the scope of data objects
• And make sure that different variables with the
  same name are assigned distinct storage
  locations.

20
Lexically-scoped Symbol Tables

• Many implementation schemes have been proposed
• Maintain separate symbol table for each scope
• A tree of local symbol tables, active ones on a
  stack
• If new scope is entered, create new symbol table.
• Or
• Number procedures (scoping units) during parsing.
  
• Compute these numbers when units are encountered.
• Save scope as part of entry
• Idea can also be used for handling blocks if they
  are scoping units
• A standard interface
• insert(name, level ) creates record for name at
  level
• lookup(name, level ) returns pointer or index

21
Use of Symbol Table

• Symbol table can be used to support layout of
  data in memory
• Save memory requirements for object
• Record memory location of object for code
  generation
• As register containing base address and offset
  from base address
• The symbolic register that variable is assigned
  to
• While doing so, assign memory to variables in
  order encountered
• Or sort on variable size (to ensure that as many
  variables as possible get small offsets)
• Or sort to maximize alignment of individual
  objects on appropriate boundaries

22
Some Typical Fields in Symbol Table