Subprograms
ISBN 0-321-49362-1

• Introduction
• Fundamentals of Subprograms
• Design Issues for Subprograms
• Local Referencing Environments
• Parameter-Passing Methods
• Parameters That Are Subprograms
• Overloaded Subprograms
• Generic Subprograms
• Design Issues for Functions
• User-Defined Overloaded Operators
• Coroutines
Copyright © 2007 Addison-Wesley. All rights reserved.
1-2

• Two fundamental abstraction facilities
– Process abstraction
• Emphasized from early days
– Data abstraction
• Emphasized in the1980s
Copyright © 2007 Addison-Wesley. All rights reserved.
1-3

• Each subprogram has a single entry point
• The calling program is suspended during execution of the called subprogram
• Control always returns to the caller when the called subprogram’s execution terminates
Copyright © 2007 Addison-Wesley. All rights reserved.
1-4

• A subprogram definition abstraction.
describes the interface to and the actions of the subprogram
– In Python, function definitions are executable; in all other languages, they are non-executable. For example: if … def fun(…): else def fun(…):
• A
… subprogram call is an explicit request that the subprogram be executed.
• A subprogram header is the first part of the definition, including the name, the kind of subprogram, and the formal parameters. For example:
– Subroutine Adder ( parameters ) Fortran
– Procedure Adder (parameters) Ada
– def adder ( parameters ): Python
– void adder ( parameters) C
• The parameter profile its parameters.
(aka signature ) of a subprogram is the number, order, and types of
• The protocol is a subprogram’s parameter profile and, if it is a function, its return type.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-5

• Function declarations in C and C++ are often called prototypes
• A subprogram declaration provides the protocol, but not the body, of the subprogram
• A formal parameter subprogram is a dummy variable listed in the subprogram header and used in the
• An actual parameter represents a value or address used in the subprogram call statement
Copyright © 2007 Addison-Wesley. All rights reserved.
1-6

• Positional
– The binding of actual parameters to formal parameters is by position: the first actual parameter is bound to the first formal parameter and so forth
– Safe and effective
• Keyword
– The name of the formal parameter to which an actual parameter is to be bound is specified with the actual parameter
– Example call to a Python function:
Sumer( length = my_length, list = my_array, sum = my_sum )
–
–
Advantage : Parameters can appear in any order, thereby avoiding parameter correspondence errors
Disadvantage : User must know the formal parameter’s names
• Both Positional and Keyword
– Ada, Fortran95, and Python
– Example:
Sumer( my_length, sum = my_sum, list = my_array )
Copyright © 2007 Addison-Wesley. All rights reserved.
1-7

• In some languages, C++, Python, Ruby, Ada, and
PHP, formal parameters can be assigned a default values, if the actual parameter is not passed.
Python function example: def compute_pay ( income, exemptions = 1, tax_rate )
Function call example: pay = compute_pay(20000.0, tax_rate = 0.15)
– In C++, default parameters must appear last because parameters are positionally associated. For example:
Float compute_pay ( float income, float tax_rate, int exemptions = 1)
– To call : pay = compute_pay(20000.0, 0.15)
Copyright © 2007 Addison-Wesley. All rights reserved.
1-8

• Variable numbers of parameters
– C, C++, Perl and JavaScript allow the number of actual parameters to be fewer than the number of formal parameters.
• For example: printf function of C
– C# allows a variable number of parameters as long as they are of the same type. For example: public void DisplayList ( params int [] list ) { foreach (int next in list ) {
Console.WriteLine(“Next value {0}”, next);
}
}
Call examples: myObject.DisplayList(myList);// myList is an int array.
myObject.DisplayList(2, 4, 3 * x – 1, 17);
- Ruby and Python also support variable numbers of parameters. In Ruby, the actual parameters are sent as elements of a hash literal and the corresponding formal parameter is preceded by an asterisk.
- In Python, the actual is a list of values and the corresponding formal parameter is a name with an asterisk
Copyright © 2007 Addison-Wesley. All rights reserved.
1-9

• There are two categories of subprograms
–
–
Procedures are collection of statements that define parameterized computations
Functions structurally resemble procedures but are semantically modeled on mathematical functions
• They are expected to produce no side effects
• In practice, program functions have side effects
Copyright © 2007 Addison-Wesley. All rights reserved.
1-10

• Are local variables static or dynamic?
• Can subprogram definitions appear in other subprogram definitions?
• What parameter passing methods are provided?
• Are parameter types checked?
• If subprograms can be passed as parameters and subprograms can be nested, what is the referencing environment of a passed subprogram?
• Can subprograms be overloaded?
• Can subprogram be generic?
Copyright © 2007 Addison-Wesley. All rights reserved.
1-11

Are local variables static or dynamic?
• Local variables can be stack-dynamic
- Advantages
• Support for recursion
• Storage for locals is shared among some subprograms
– Disadvantages
• Allocation/de-allocation, initialization time
• Uses indirect addressing
• Subprograms cannot be history sensitive
• Local variables can be static
– Advantages and disadvantages are the opposite of those for stack-dynamic local variables
Copyright © 2007 Addison-Wesley. All rights reserved.
1-12

• Most languages implement local variables as stack-dynamic variables. Ada and methods of C++, C# and Java.
• Fortran95 allows the user to choose via the Save statement.
• In this PL/I example, Count is stack-dynamic due to the fact that the subroutine is marked as Recursive, whereas, Sum is so due to the keyword Save.
Recursive Subroutine Sub()
Integer :: Count
Save, Real :: Sum
…
End Subroutine Sub
Copyright © 2007 Addison-Wesley. All rights reserved.
1-13

• The default for local variables in C/C++ functions is stackdynamic, however, by using the keyword static in the declaration, they can be implemented as static variables.
– Example: int adder(int list[], int listlen) { static int sum = 0; int count; for (count = 0; count < listlen; count++) sum += list[count]; return sum;
}
Non-static version: int adder(int list[], int listlen) { int sum = 0; int count; for (count = 0; count < listlen; count++) sum += list[count]; return sum;
}
Copyright © 2007 Addison-Wesley. All rights reserved.
1-14

• This idea originated with Algol 60.
• Motivation was to create a hierarchy of both logic and scope.
• Combined with static scoping, this provides a highly structured way to grant access to nonlocal variables in enclosing subprograms.
• Allowed by Algol60, Algol 68, Pascal and Ada.
• Not allowed by C, C++ and Java.
• Allowed by JavaScript, Python and Ruby.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-15

• Semantic Models of Parameter Passing
– In mode
– Out mode
– Inout mode
• Conceptual Models of Transfer
– An actual value is copied to the caller, to the callee, or both ways (see above)
– An access path is transmitted, such as a simple pointer or reference.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-16

Three Semantic Models of Parameter Passing When Values are Copied.

Copyright © 2007 Addison-Wesley. All rights reserved.
1-18

• Implementation Models of Parameter
Passing
– Pass-by-value
– Pass-by-result
– Pass-by-value–result
– Pass-by-reference
– Pass-by-name
Copyright © 2007 Addison-Wesley. All rights reserved.
1-19

• The value of the actual parameter is used to initialize the corresponding formal parameter, which then acts as a local.
– Normally implemented by copying.
• Disadvantages : (if by physical move): additional storage is required (stored twice) and the actual move can be costly
(for large parameters).
– Can be implemented by transmitting an access path.
• Disadvantages : must write-protect the variable in the called subprogram and accesses cost more (indirect addressing).
Copyright © 2007 Addison-Wesley. All rights reserved.
1-20

• No value is transmitted to the subprogram.
• The corresponding formal parameter acts as a local variable, with its value transmitted to caller’s actual parameter when control is returned to the caller, by a physical move.
• Same advantages/disadvantages of pass-by-value.
• If values are returned by access path, then there is the problem of ensuring that the initial value of the actual paramete ris not used in the called subprogram. The difficulties of this lead to values returned by copy method.
• If values are returned by copy, then disadvantages are:
– Extra storage and overhead of the copy operation.
– Actual parameter collision.
– Time at which the return address is evaluated.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-21

• Actual parameter collision.
• Consider the call: sub(p1, p1); whichever formal parameter is copied to its corresponding actual parameter last will represent the current value of p1 in the caller.
• Example in C#.
void Fixer (out int x, out int y) { s = 17; y = 35;
}
… f.Fixer(out a, out a);
What is the value of a upon return from the function call.
• Depends upon the implementation. Which of the copies from formal to actual occurs last.
• Since this order may vary by different implementations, the results of the same program may also vary.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-22

• Problem of the implementation’s choice of time at which the return address is evaluated.
– Time of the call
– Time of the return.
– Example in C#.
void DoIt( out int x, int index) { x = 17; index = 42;
}
.. sub = 21; f.DoIt(list[sub], sub);
– Result if address of list[sub] is evaluated at the time of the call:
• Value of 17 will be returned to list[21]
– Result if address of list[sub] is evaluated at the time of the return:
• Value of 17 will be returned to list[42];
– Again, since implementations may vary, the results of the same program may also vary.
1-23

– Those of pass-by-result
– Those of pass-by-value
Copyright © 2007 Addison-Wesley. All rights reserved.
1-24

• Passes an access path.
• Also called pass-by-sharing
• Advantage:
– Passing process is efficient
– No duplicated storage
• Disadvantages
– Slower accesses (compared to pass-by-value) to formal parameters
– Potentials for unwanted changes to the actual parameter.
– Unwanted aliases (access broadened)
Copyright © 2007 Addison-Wesley. All rights reserved.
1-25

• Unwanted aliases examples
– Collisions between actual parameters. Example in C++ void fun (int &first, int &second)
Now consider call of: fun(total, total); first and second are now aliases.
– Collisions between array elements.
Consider call of: fun( list[i], list[j] );
If i and j are equal, then first and second are now aliases.
– Collisions between formal parameters and nonlocal variables that are visible. Example in C int * global; void main() {
… sub(global);
} void sub(int * param) {….}
Param and global are now aliases.

• Pascal provides for two:
– X: integer is a pass-by-value
– Var X: integer is a pass-by-reference
• C (from Algol 68)
– Pass-by-value only, but pass-by-reference is achieved by using pointers as parameters
• Actual parameter is &i in the call -- subA(&i);
• Formal parameter is *x in the function heading -- subA(int *x);
• C++
– Has a special pointer type called reference type for pass-by-reference, which are implicitly dereferenced.
• void fun (const int & p1, int p2, int &p3) {…} where, p1 and p3 are both implicitly dereferenced and p1 cannot be changed.
• Java
– All parameters are passed are passed by value. However, since you are passing a reference in the case of Object parameters, you are giving write-access to the object.
1-27

• Ada
– Three semantics modes of parameter transmission: in, out, in out; in is the default mode
– Formal parameters
• declared out can be assigned but not referenced
• declared in can be referenced but not assigned
• declared in out parameters can be referenced and assigned
– Example : Procedure Adder (A: in out Integer; B : in Integer; C : out Float)
• Fortran 95
- Parameters can be declared to be in, out, or inout mode
– Example: Subroutine Adder(A, B, C)
Integer, Intent(Inout) :: A
Integer, Intent(In) :: B
Integer, Intent(out) :: C
• C#
Default method: pass-by-value
– Pass-by-reference is specified by preceding both a formal parameter and its actual parameter with ref
– Example : void sumer (ref int oldSum, int newOne) {..} call is sumer(ref sum, newValue);
• PHP: very similar to C#
• Perl: all actual parameters are implicitly placed in a predefined array named
@_
• Python and Ruby use pass-by-assignment (all data values are objects)
Copyright © 2007 Addison-Wesley. All rights reserved.
1-28

• In most languages parameter transmissions takes place through the run-time stack.
– Pass-by-value actual parameters have their values copied into those stack locations which serve as storage for their corresponding formal parameters.
– Pass-by-result actual parameters are assigned the final value of their corresponding formal parameters when the subprogram terminates.
– Pass-by-value-result have their values copied into those stack locations which serve as storage for their corresponding formal parameters, where The formal parameters are used as local variables. When the subprogram terminates, the final values of these locals are copied back into the actual parameters as with pass-by-result.
– Pass-by-reference is the simplest to implement; only an address is placed in the stack.
• Subtle but fatal errors can occur with pass-by-reference and pass-by-value-result.
– A formal parameter corresponding to a constant can mistakenly be changed.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-29

Main calls sub with actual parameters: w, x, y and z
Which correspond to the actual parameters a, b, c and d in function sub.
void sub(int a, int b, int c, int d) with the following parameter passing mechanisms:
(by-value, by-result, by-value-result, by-reference)

• Subtle but fatal errors can occur with pass-by-reference and passby-value-result if implementations do not prevent their occurrence.
•
– A formal parameter corresponding to a constant can mistakenly be changed.
– Example:
• Hypothetical Subprogram uses pass-by-reference
Sub1( Integer: P1, Integer: P2) begin
P1 = 8
P2 = 5 end.
• Calling program (assume that b is 10
Sub1(a, 10); result = b * 10; // result will be 50 vs. 100
– Actually happened in Fortran IV
– C and C++ provide for the typing of formal parameters as pointers to constants to allow for the efficiency of pass-by-reference with the security of pass by value.
Even if the actual parameter is not a constant it is coerced to a constant.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-31

• Considered very important for reliability.
• No type checking required in FORTRAN 77 and original C.
• C89 the user can choose.
– Prototypes are used for this purpose, where
• formal parameters can also be defined as: double sin(double x)
{…}
– Vs.
double sin(x) double x;
{…}
• C99 and C++ all functions must be in prototype form.
• Always required in Pascal, FORTRAN 90, Java, and Ada.
• Relatively new languages Perl, JavaScript, and PHP do not require type checking.
• In Python and Ruby, variables do not have types (objects do), so parameter type checking is not possible.
Copyright © 2007 Addison-Wesley. All rights reserved.
1-32

• If a multidimensional array is passed to a subprogram and the subprogram is separately compiled, the compiler needs to know the declared size of that array to build the storage mapping function
Copyright © 2007 Addison-Wesley. All rights reserved.
1-33

• Programmer is required to include the declared sizes of all but the first subscript in the actual parameter
• Disallows writing flexible subprograms
• Solution: pass a pointer to the array and the sizes of the dimensions as other parameters; the user must include the storage mapping function in terms of the size parameters
Copyright © 2007 Addison-Wesley. All rights reserved.
1-34

• The actual parameter is, in effect, textually substituted for the corresponding formal parameter in all of its occurrences in the subprogram.
• Not part of any widely used language.
– Used at compile time by the macros in assembly languages
– Used for generic parameters of the generic programs of C++ and Ada.
• Allows flexibility in late binding
1-35

• Ada – not a problem
– Constrained arrays – size is part of the array’s type
– Unconstrained arrays - declared size is part of the object declaration
• Java and C#
– Similar to Ada
– Arrays are objects; they are all single-dimensioned, but the elements can be arrays
– Each array inherits a named constant ( length in Java,
Length in C#) that is set to the length of the array when the array object is created
Copyright © 2007 Addison-Wesley. All rights reserved.
1-36

– For single-dimension arrays, the subscript is irrelevant
– For multidimensional arrays, the sizes are sent as parameters and used in the declaration of the formal parameter, so those variables are used in the storage mapping function
Copyright © 2007 Addison-Wesley. All rights reserved.
1-37

• Two important considerations
– Efficiency
– One-way or two-way data transfer
• But the above considerations are in conflict
– Good programming suggest limited access to variables, which means one-way whenever possible
– But pass-by-reference is more efficient to pass structures of significant size
Copyright © 2007 Addison-Wesley. All rights reserved.
1-38

• It is sometimes convenient to pass subprogram names as parameters
• Issues:
1. Are parameter types checked?
2. What is the correct referencing environment for a subprogram that was sent as a parameter?
Copyright © 2007 Addison-Wesley. All rights reserved.
1-39

• C and C++: functions cannot be passed as parameters but pointers to functions can be passed and their types include the types of the parameters, so parameters can be type checked
• FORTRAN 95 type checks
• Ada does not allow subprogram parameters; an alternative is provided via Ada’s generic facility
• Java does not allow method names to be passed as parameters
Copyright © 2007 Addison-Wesley. All rights reserved.
1-40

•
•
•
: The environment of the call statement that enacts the passed subprogram
- Most natural for dynamic-scoped languages
: The environment of the definition of the passed subprogram
- Most natural for static-scoped languages
: The environment of the call statement that passed the subprogram
Copyright © 2007 Addison-Wesley. All rights reserved.
1-41

• An overloaded subprogram is one that has the same name as another subprogram in the same referencing environment
– Every version of an overloaded subprogram has a unique protocol
• C++, Java, C#, and Ada include predefined overloaded subprograms
• In Ada, the return type of an overloaded function can be used to disambiguate calls (thus two overloaded functions can have the same parameters)
• Ada, Java, C++, and C# allow users to write multiple versions of subprograms with the same name
Copyright © 2007 Addison-Wesley. All rights reserved.
1-42

• A generic or polymorphic subprogram different types on different activations takes parameters of
• Overloaded subprograms provide ad hoc polymorphism
• A subprogram that takes a generic parameter that is used in a type expression that describes the type of the parameters of the subprogram provides parametric polymorphism
- A cheap compile-time substitute for dynamic binding
• In Ada, generic subprograms are instantiated explicitly
• In C++, they are instantiated implicitly, when the subprogram is named in a call or when its address is taken with the & operator
Copyright © 2007 Addison-Wesley. All rights reserved.
1-43

• Java 5.0
- Differences between generics in Java 5.0 and those of C++ and Ada:
1. Generic parameters in Java 5.0 must be classes
2. Java 5.0 generic methods are instantiated just once as truly generic methods
3. Restrictions can be specified on the range of classes that can be passed to the generic method as generic parameters
4. Wildcard types of generic parameters
Copyright © 2007 Addison-Wesley. All rights reserved.
1-44

• C# 2005
- Supports generic methods that are similar to those of Java 5.0
- One difference: actual type parameters in a call can be omitted if the compiler can infer the unspecified type
Copyright © 2007 Addison-Wesley. All rights reserved.
1-45

template <class Type>
Type max(Type first, Type second) { return first > second ? first : second;
}
• The above template can be instantiated for any type for which operator > is defined int max (int first, int second) { return first > second? first : second;
}
Copyright © 2007 Addison-Wesley. All rights reserved.
1-46

• Are side effects allowed?
– Parameters should always be in-mode to reduce side effect (like Ada)
• What types of return values are allowed?
– Most imperative languages restrict the return types
– C allows any type except arrays and functions
– C++ is like C but also allows user-defined types
– Ada subprograms can return any type (but Ada subprograms are not types, so they cannot be returned)
– Java and C# methods can return any type (but because methods are not types, they cannot be returned)
– Python and Ruby treat methods as first-class objects, so they can be returned, as well as any other class
Copyright © 2007 Addison-Wesley. All rights reserved.
1-47

• Operators can be overloaded in Ada, C++,
Python, and Ruby
• An Ada example function "*" (A,B: in Vec_Type): return Integer is
Sum: Integer := 0; begin for Index in A' range loop
Sum := Sum + A(Index) * B(Index) end loop return sum; end "*";
… c = a * b; -- a, b, and c are of type Vec_Type
Copyright © 2007 Addison-Wesley. All rights reserved.
1-48

• A coroutine is a subprogram that has multiple entries and controls them itself
• Also called symmetric control: caller and called coroutines are on a more equal basis
• A coroutine call is named a resume
• The first resume of a coroutine is to its beginning, but subsequent calls enter at the point just after the last executed statement in the coroutine
• Coroutines repeatedly resume each other, possibly forever
• Coroutines provide quasi-concurrent execution interleaved, but not overlapped of program units (the coroutines); their execution is
Copyright © 2007 Addison-Wesley. All rights reserved.
1-49

Copyright © 2007 Addison-Wesley. All rights reserved.
1-50

Copyright © 2007 Addison-Wesley. All rights reserved.
1-51

Copyright © 2007 Addison-Wesley. All rights reserved.
1-52

• A subprogram definition describes the actions represented by the subprogram
• Subprograms can be either functions or procedures
• Local variables in subprograms can be stackdynamic or static
• Three models of parameter passing: in mode, out mode, and inout mode
• Some languages allow operator overloading
• Subprograms can be generic
• A coroutine is a special subprogram with multiple entries
Copyright © 2007 Addison-Wesley. All rights reserved.
1-55

• Ruby includes a number of iterator functions, which are often used to process the elements of arrays
• Iterators are implemented with blocks, which can also be defined by applications
• Blocks are attached methods calls; they can have parameters (in vertical bars); they are executed when the method executes a yield statement def fibonacci(last) first, second = 1, 1 while first <= last yield first first, second = second, first + second end end puts "Fibonacci numbers less than 100 are:" fibonacci(100) {|num| print num, " "} puts
Copyright © 2007 Addison-Wesley. All rights reserved.
1-56