Last modified: Feb 24, 2014
Dynamic binding is a key factor in allowing different classes to respond to the same message with different methods.
Binding is a term used in programming languages to denote the association of information with a symbol. It’s a very general term with many applications.
a = 2; is a binding of a value to a variable.
String s; is a binding of a type (String) to the variable name (s).
Binding Function Calls to Bodies
In OOP, we are particularly interested in the
binding of a function body (method) to a function call.
Given the following:
a = foo(b);
When is the decision made as to what code will be executed for this call to foo?
Compile-Time Binding
a = foo(b);
In traditionally compiled languages (FORTRAN, PASCAL, C, etc), the decision is made at compile-time.
Decision is immutable
If this statement is inside a loop, the same code will be invoked for foo each time.
Compile-time binding is cheap – there’s very little execution-time overhead.
Run-Time Binding
a = foo(b);
In traditionally interpreted languages (LISP, BASIC, etc), the decision is made at run-time.
Decision is often mutable
If this statement is inside a loop, different code may be invoked for foo each time.
Run-time binding can be expensive (high execution-time overhead) because it suggests that some sort of decision or lookup is done at each call.
Dynamic Binding = the Happy Medium?
OOPLs typically feature dynamic binding, an “intermediate” choice in which
the choice of method is made from a relatively small list of options
that list is determined at compile time
the final choice made at run-time
the options that make up that list are organized according to the inheritance hierarchy
Dynamic Binding and Programming Languages
In Java, all function calls are resolved by dynamic binding.
In C++, we can choose between compile-time and dynamic binding.
Virtual functions
A non-inherited function member is subject to dynamic binding if its declaration is preceded by the word virtual.
An inherited function member is subject to dynamic binding if that member in the base class is subject to dynamic binding.
Using the word virtual in subclasses is optional (but recommended).
Dynamic Binding in C++: virtual functions
Declaring a function as virtual gives programmers permission to call it via dynamic binding.
x.foo(), where x is an object, is bound at compile time
x.foo(), where x is a reference, is bound at run-time (dynamic).
x->{foo()}, where x is a pointer, is bound at run-time (dynamic).
An Animal Inheritance Hierarchy
For this example, we will introduce a simple hierarchy.
class Animal {
public:
virtual String eats() {return "???";}
String name() {return "Animal";}
};
We begin with the base class, Animal, which has two functions.
Plant Eaters
Now we introduce a subclass of Animal that overrides both those functions.
class Herbivore: public Animal {
public:
virtual String eats() {return "plants";}
String name() {return "Herbivore";}
};
That function is already virtual because it was declared that way in the base class.
But listing the virtual in the subclass is a nice reminder to the reader.
Cud-Chewers
Now we introduce a subclass of that class.
class Ruminants: public Herbivore {
public:
virtual String eats() {return "grass";}
String name() {return "Ruminant";}
};
Meat Eaters
And another subclass of the original base class.
class Carnivore: public Animal {
public:
virtual String eats() {return "meat";}
String name() {return "Carnivore";}
};
Output Function
It will also be useful in this example to have a simple utility function to print a pair of strings.
void show (String s1, String s2) {
cout << s1 << " " << s2 << endl;
}
Let’s Make Some Calls
Animal a, *paa, *pah, *par;
Herbivore h, *phh;
Ruminant r;
paa = &a; phh = &h; pah = &h; par = &r;
show(a.name(), a.eats()); // AHRC ?pgm
show(paa->name(), paa->eats()); // AHRC ?pgm
show(h.name(), h.eats); // AHRC ?pgm
show(phh->name(), phh->eats()); // AHRC ?pgm
show(pah->name(), pah->eats()); // AHRC ?pgm
show(par->name(), par->eats()); //AHRC ?pgm
Note the variety of variables we are using.
We have three actual objects, a, h, and r, which are of type Animal, Herbivore, and Ruminant, respectively.
The second letter of each pointer variable name indicates its data type.
Thus, pa, pah, and par are all of type Animal*. ph has type Herbivore*.
These pointer variables are all assigned the address of one of our three actual objects. (The unary prefix operator & in C++ is the “address-of” operator.)
The third letter in each pointer variable’s name indicates what type of object it actually points to. Thus paa is of type Animal* and actually points to an Animal object, a.
pah, on the other hand, is of type Animal* but actually points to an Herbivore object, h. (This is possible because of subtyping - we can substitute a subtype object into a context where the supertype object is expected.)
What’s the output?
Animal a, *paa, *pah, *par;
Herbivore h, *phh;
Ruminant r;
paa = &a; phh = &h; pah = &h; par = &r;
show(a.name(), a.eats()); // AHRC ?pgm
show(paa->name(), paa->eats()); // AHRC ?pgm
show(h.name(), h.eats); // AHRC ?pgm
show(phh->name(), phh->eats()); // AHRC ?pgm
show(pah->name(), pah->eats()); // AHRC ?pgm
show(par->name(), par->eats()); //AHRC ?pgm
Question: What will be the output of the various show calls?
Animal a, *paa, *pah, *par;
Herbivore h, *phh;
Ruminant r;
paa = &a; phh = &h; pah = &h; par = &r;
show(a.name(), a.eats());<:>>>:> // Animal ??? ➊
show(paa->name(), paa->eats());<:>>>:> // Animal ??? ➋
show(h.name(), h.eats);<:>>>:> // Herbivore plants ➌
show(phh->name(), phh->eats());<:>>>:> // Herbivore plants ➍
show(pah->name(), pah->eats());<:>>>:> // Animal plants ➎
show(par->name(), par->eats());<:>>>:> //Animal grass ➏
Dynamic binding lets us write application code for the superclass that can be applied to the subclasses, taking advantage of the subclasses’ different methods.
Collections of Pointers/References to a Base Class
Suppose we have an inheritance hierarchy:
and that we have a collection of pointers or references to the BaseClass.
The Key Pattern to All OOP
Then this code:
BaseClass* x;
for (each x in collection) {
x->virtualFunction(...);
}
uses dynamic binding to apply subclass-appropriate behavior to each element of a collection.
Each time around the loop, we extract a pointer from the collection.
But when we call virtualFunction through that
pointer, the runtime system uses the data type of the thing pointed to
determine which function body to invoke.
Study this pattern. Once you understand this, you have grasped the
essence of OOP!
There are lots of variations on this pattern. We can use almost any data structure for the collection.
Example: arrays of Animals
Animal** animals = new Animal*[numberOfAnimals];
⋮
for (int i = 0; i < numberOfAnimals; ++i)
cout << animals[i]->name() << " "
<< animals[i]->eats() << endl;
Example: Linked Lists of Animals (C++)
struct ListNode {
Animal* data;
ListNode* next;
};
ListNode* head; // start of list
⋮
for (ListNode* current = head; current != 0; current = current->next)
cout << current->data->name() << " "
<< current->data->eats() << endl;
Example: vector of Animals
vector<Animal*> animals;
⋮
for (int i = 0; i < animals.size(); ++i)
cout << animals[i]->name() << " "
<< animals[i]->eats() << endl;
Example: Trees of Animals
struct TreeNode {
Animal* data;
TreeNode* leftChild;
TreeNode* rightChild;
};
TreeNode* root;
void printTree (const TreeNode* t)
{
if (t != 0) {
printTree(t->leftChild);
cout << t->data->name() << " "
<< t->data->eats() << endl;
printTree(t->rightChild);
}
}
⋮
printTree(root);
Continuing our earlier example:
Every Cell holds a Value.
Every Value can be rendered into a string of a given max width. (See the render function in
#ifndef VALUE_H
#define VALUE_H
#include <string>
#include <typeinfo>
//
// Represents a value that might be obtained for some spreadsheet cell
// when its formula was evaluated.
//
// Values may come in many forms. At the very least, we can expect that
// our spreadsheet will support numeric and string values, and will
// probably need an "error" or "invalid" value type as well. Later we may
// want to add addiitonal value kinds, such as currency or dates.
//
class Value
{
public:
virtual ~Value() {}
virtual std::string render (unsigned maxWidth) const = 0;
// Produce a string denoting this value such that the
// string's length() <= maxWidth (assuming maxWidth > 0)
// If maxWidth==0, then the output string may be arbitrarily long.
// This function is intended to supply the text for display in the
// cells of a spreadsheet.
virtual Value* clone() const = 0;
// make a copy of this value
protected:
virtual bool isEqual (const Value& v) const = 0;
//pre: typeid(*this) == typeid(v)
// Returns true iff this value is equal to v, using a comparison
// appropriate to the kind of value.
friend bool operator== (const Value&, const Value&);
};
inline
bool operator== (const Value& left, const Value& right)
{
return (typeid(left) == typeid(right))
&& left.isEqual(right);
}
#endif
Pairs of Values can be compared for equality
Numeric, String, and Error values are some of the possible Values
Displaying a Cell
Here is the code to draw a spreadsheet on the screen.
void NCursesSpreadSheetView::redraw() const
{
drawColumnLabels();
drawRowLabels();
CellRange shown = showing();
for (CellName cn = shown.first();
shown.more(cn); cn = shown.next(cn))
drawCell(cn);
}
We have a loop that goes through the collection of cell names, invoking drawCell on each one.
drawCell
void NCursesSpreadSheetView::drawCell
(CellName name) const
{
string cellValue;
Cell* c = sheet.getCell(name);
const Value* v = c->getValue();
if (v != 0)
{
cellValue = v->render(theColWidth);
}
centerStringInWidth (cellValue,
theColWidth);
// . . . show cellValue on screen . . .
}
render()
Now render in value.h is virtual, and various bodies implementing it can be found in classes like
std::string NumericValue::render (unsigned maxWidth) const
// Produce a string denoting this value such that the
// string's length() <= maxWidth (assuming maxWidth > 0)
// If maxWidth==0, then the output string may be arbitrarily long.
// This function is intended to supply the text for display in the
// cells of a spreadsheet.
{
char buffer[256];
for (char precision = '6'; precision > '0'; --precision)
{
if (maxWidth > 0)
{
sprintf (buffer, "%.1u", maxWidth);
}
else
buffer[0] = 0;
string format = string("%") + buffer + "." + precision + "g";
int width = sprintf (buffer, format.c_str(), d);
if (maxWidth == 0 || width <= maxWidth)
{
string result = buffer;
result.erase(0, result.find_first_not_of(" "));
return result;
}
}
return string(maxWidth, '*');
}
, NumericValue,
std::string StringValue::render (unsigned maxWidth) const
// Produce a string denoting this value such that the
// string's length() <= maxWidth (assuming maxWidth > 0)
// If maxWidth==0, then the output string may be arbitrarily long.
// This function is intended to supply the text for display in the
// cells of a spreadsheet.
{
if (maxWidth == 0 || maxWidth > s.length())
return s;
else
return s.substr(0, maxWidth);
}
, StringValue, and
std::string ErrorValue::render (unsigned maxWidth) const
// Produce a string denoting this value such that the
// string's length() <= maxWidth (assuming maxWidth > 0)
// If maxWidth==0, then the output string may be arbitrarily long.
// This function is intended to supply the text for display in the
// cells of a spreadsheet.
{
string s = theValueKindName;
if (maxWidth == 0 || maxWidth > s.length())
return s;
else
return s.substr(0, maxWidth);
}
, ErrorValue, .
So that render call will be resolved by dynamic binding, sending us to the proper function body depending on just what kind of value is actually stored in the cell.
Combine that with the loop in the redraw function, and we have a loop going through a collection of pointers (the value pointers inside the cells inside the spreadsheet) and using each pointer to invoke a virtual function.
const Value* Cell::evaluateFormula()
{
Value* newValue = (theFormula == 0)
? new StringValue()
: theFormula->evaluate(theSheet);
if (theValue != 0 && *newValue == *theValue)
delete newValue;
else
{
delete theValue;
theValue = newValue;
notifyObservers();
}
return theValue;
}
After evaluating a formula in a spreadsheet cell we check to see if the value obtained is equal to theValue already stored in that cell.
If the values are equal, we simply discard the newly computed value. We don’t need it.
But if they are not equal, we need to save the new value in place of the old one and trigger the re-evaluation of any cells that mention this one in their formulas.
operator==
Look at the implementation of operator== in value.h.
bool NumericValue::isEqual (const Value& v) const
//pre: valueKind() == v.valueKind()
// Returns true iff this value is equal to v, using a comparison
// appropriate to the kind of value.
{
const NumericValue& vv = dynamic_cast<const NumericValue&>(v);
return d == vv.d;
}
,
bool StringValue::isEqual (const Value& v) const
//pre: valueKind() == v.valueKind()
// Returns true iff this value is equal to v, using a comparison
// appropriate to the kind of value.
{
const StringValue& vv = dynamic_cast<const StringValue&>(v);
return s == vv.s;
}
, and
bool ErrorValue::isEqual (const Value& v) const
//pre: valueKind() == v.valueKind()
// Returns true iff this value is equal to v, using a comparison
// appropriate to the kind of value.
{
return false;
}
.