What are "first-class" objects?
Question:
When are objects or something else said to be "first-class" in a given programming language, and why? In what way do they differ from languages where they are not?
When one says "everything is an object" (like in Python), do they indeed mean that "everything is first-class"?
Answers:
In short, it means there are no restrictions on the object’s use. It’s the same as
any other object.
A first class object is an entity that can be dynamically created, destroyed, passed to a function, returned as a value, and have all the rights as other variables in the programming language have.
Depending on the language, this can
imply:
- being expressible as an anonymous literal value
- being storable in variables
- being storable in data structures
- having an intrinsic identity (independent of any given name)
- being comparable for equality with other entities
- being passable as a parameter to a procedure/function
- being returnable as the result of a procedure/function
- being constructible at runtime
- being printable
- being readable
- being transmissible among distributed processes
- being storable outside running processes
In C++ functions themselves are not first class objects, however:
- You can override the ‘()’ operator making it possible to have an object function, which is first class.
- Function pointers are first class.
- boost bind, lambda and function do offer first class functions
In C++, classes are not first class objects but instances of those classes are. In Python both the classes and the objects are first class objects. (See this answer for more details about classes as objects).
Here is an example of Javascript first class functions:
// f: function that takes a number and returns a number
// deltaX: small positive number
// returns a function that is an approximate derivative of f
function makeDerivative( f, deltaX )
{
var deriv = function(x)
{
return ( f(x + deltaX) - f(x) )/ deltaX;
}
return deriv;
}
var cos = makeDerivative( Math.sin, 0.000001);
// cos(0) ~> 1
// cos(pi/2) ~> 0
Entities that are not first class objects are referred to as second-class objects. Functions in C++ are second class because they can’t be dynamically created.
Regarding the edit:
EDIT. When one says “everything is
an object” (like in Python), does he
indeed mean that “everything is
first-class”?
The term object can be used loosely and doesn’t imply being first class. And it would probably make more sense to call the whole concept ‘first class entities’. But in Python they do aim to make everything first class. I believe the intent of the person who made your statement meant first class.
“First class” means you can operate on them in the usual manner. Most of the times, this just means you can pass these first-class citizens as arguments to functions, or return them from functions.
This is self-evident for objects but not always so evident for functions, or even classes:
void f(int n) { return n * 2; }
void g(Action<int> a, int n) { return a(n); }
// Now call g and pass f:
g(f, 10); // = 20
This is an example in C# where functions actually aren’t first-class objects. The above code therefore uses a small workaround (namely a generic delegate called Action<>) to pass a function as an argument. Other languages, such as Ruby or Python, allow treating even classes and code blocks as normal variables (or in the case of Ruby, constants).
IMO this is one of those metaphors used to describe things in a natural language. The term is essentially used in context of describing functions as first class objects.
If you consider a object oriented language, we can impart various features to objects for eg: inheritance, class definition, ability to pass to other sections of code(method arguments), ability to store in a data structure etc. If we can do the same with an entity which is not normally considered as a object, like functions in the case of java script, such entities are considered to be first class objects.
First class essentially here means, not handled as second class (with degraded behaviour). Essentially the mocking is perfect or indistinguishable.
“When one says “everything is an object” (like in Python), does he indeed mean that “everything is first-class”?”
Yes.
Everything in Python is a proper object. Even things that are “primitive types” in other languages.
You find that an object like 2 actually has a fairly rich and sophisticated interface.
>>> dir(2)
['__abs__', '__add__', '__and__', '__class__', '__cmp__', '__coerce__', '__delattr__', '__div__', '__divmod__', '__doc__', '__float__', '__floordiv__', '__getattribute__', '__getnewargs__', '__hash__', '__hex__', '__index__', '__init__', '__int__', '__invert__', '__long__', '__lshift__', '__mod__', '__mul__', '__neg__', '__new__', '__nonzero__', '__oct__', '__or__', '__pos__', '__pow__', '__radd__', '__rand__', '__rdiv__', '__rdivmod__', '__reduce__', '__reduce_ex__', '__repr__', '__rfloordiv__', '__rlshift__', '__rmod__', '__rmul__', '__ror__', '__rpow__', '__rrshift__', '__rshift__', '__rsub__', '__rtruediv__', '__rxor__', '__setattr__', '__str__', '__sub__', '__truediv__', '__xor__']
Because everything’s a first-class object in Python, there are relatively few obscure special cases.
In Java, for example, there are primitive types (int, bool, double, char) that aren’t proper objects. That’s why Java has to introduce Integer, Boolean, Double, and Character as first-class types. This can be hard to teach to beginners — it isn’t obvious why both a primitive type and a class have to exist side-by-side.
It also means that an object’s class is — itself — an object. This is different from C++, where the classes don’t always have a distinct existence at run-time.
The type of 2 is the type 'int' object, which has methods, attributes, and a type.
>>> type(2)
<class 'int'>
The type of a built-in type like int is the type 'type' object. This has methods and attributes, also.
>>> type(type(2))
<class 'type'>
From a slide in Structure and Interpretation of Computer Programs, lecture 2A (1986), which in turns quotes Christopher Stracey:
The rights and privileges of first-class citizens:
- To be named by variables.
- To be passed as arguments to procedures.
- To be returned as values of procedures.
- To be incorporated into data structures
An easy way to understand is to ask: What things are not first class?
In Python, e.g. operators like + or * are not first class.
You can use the operators directly:
2+2
"foo"+"bar"
2*3
"bla"*3
But you can not assign operators to variables:
operator = +
operators = [+,*]
which means you can not pass operators as function arguments because…
… operators are not first class.
When are objects or something else said to be "first-class" in a given programming language, and why? In what way do they differ from languages where they are not?
When one says "everything is an object" (like in Python), do they indeed mean that "everything is first-class"?
In short, it means there are no restrictions on the object’s use. It’s the same as
any other object.
A first class object is an entity that can be dynamically created, destroyed, passed to a function, returned as a value, and have all the rights as other variables in the programming language have.
Depending on the language, this can
imply:
- being expressible as an anonymous literal value
- being storable in variables
- being storable in data structures
- having an intrinsic identity (independent of any given name)
- being comparable for equality with other entities
- being passable as a parameter to a procedure/function
- being returnable as the result of a procedure/function
- being constructible at runtime
- being printable
- being readable
- being transmissible among distributed processes
- being storable outside running processes
In C++ functions themselves are not first class objects, however:
- You can override the ‘()’ operator making it possible to have an object function, which is first class.
- Function pointers are first class.
- boost bind, lambda and function do offer first class functions
In C++, classes are not first class objects but instances of those classes are. In Python both the classes and the objects are first class objects. (See this answer for more details about classes as objects).
Here is an example of Javascript first class functions:
// f: function that takes a number and returns a number
// deltaX: small positive number
// returns a function that is an approximate derivative of f
function makeDerivative( f, deltaX )
{
var deriv = function(x)
{
return ( f(x + deltaX) - f(x) )/ deltaX;
}
return deriv;
}
var cos = makeDerivative( Math.sin, 0.000001);
// cos(0) ~> 1
// cos(pi/2) ~> 0
Entities that are not first class objects are referred to as second-class objects. Functions in C++ are second class because they can’t be dynamically created.
Regarding the edit:
EDIT. When one says “everything is
an object” (like in Python), does he
indeed mean that “everything is
first-class”?
The term object can be used loosely and doesn’t imply being first class. And it would probably make more sense to call the whole concept ‘first class entities’. But in Python they do aim to make everything first class. I believe the intent of the person who made your statement meant first class.
“First class” means you can operate on them in the usual manner. Most of the times, this just means you can pass these first-class citizens as arguments to functions, or return them from functions.
This is self-evident for objects but not always so evident for functions, or even classes:
void f(int n) { return n * 2; }
void g(Action<int> a, int n) { return a(n); }
// Now call g and pass f:
g(f, 10); // = 20
This is an example in C# where functions actually aren’t first-class objects. The above code therefore uses a small workaround (namely a generic delegate called Action<>) to pass a function as an argument. Other languages, such as Ruby or Python, allow treating even classes and code blocks as normal variables (or in the case of Ruby, constants).
IMO this is one of those metaphors used to describe things in a natural language. The term is essentially used in context of describing functions as first class objects.
If you consider a object oriented language, we can impart various features to objects for eg: inheritance, class definition, ability to pass to other sections of code(method arguments), ability to store in a data structure etc. If we can do the same with an entity which is not normally considered as a object, like functions in the case of java script, such entities are considered to be first class objects.
First class essentially here means, not handled as second class (with degraded behaviour). Essentially the mocking is perfect or indistinguishable.
“When one says “everything is an object” (like in Python), does he indeed mean that “everything is first-class”?”
Yes.
Everything in Python is a proper object. Even things that are “primitive types” in other languages.
You find that an object like 2 actually has a fairly rich and sophisticated interface.
>>> dir(2)
['__abs__', '__add__', '__and__', '__class__', '__cmp__', '__coerce__', '__delattr__', '__div__', '__divmod__', '__doc__', '__float__', '__floordiv__', '__getattribute__', '__getnewargs__', '__hash__', '__hex__', '__index__', '__init__', '__int__', '__invert__', '__long__', '__lshift__', '__mod__', '__mul__', '__neg__', '__new__', '__nonzero__', '__oct__', '__or__', '__pos__', '__pow__', '__radd__', '__rand__', '__rdiv__', '__rdivmod__', '__reduce__', '__reduce_ex__', '__repr__', '__rfloordiv__', '__rlshift__', '__rmod__', '__rmul__', '__ror__', '__rpow__', '__rrshift__', '__rshift__', '__rsub__', '__rtruediv__', '__rxor__', '__setattr__', '__str__', '__sub__', '__truediv__', '__xor__']
Because everything’s a first-class object in Python, there are relatively few obscure special cases.
In Java, for example, there are primitive types (int, bool, double, char) that aren’t proper objects. That’s why Java has to introduce Integer, Boolean, Double, and Character as first-class types. This can be hard to teach to beginners — it isn’t obvious why both a primitive type and a class have to exist side-by-side.
It also means that an object’s class is — itself — an object. This is different from C++, where the classes don’t always have a distinct existence at run-time.
The type of 2 is the type 'int' object, which has methods, attributes, and a type.
>>> type(2)
<class 'int'>
The type of a built-in type like int is the type 'type' object. This has methods and attributes, also.
>>> type(type(2))
<class 'type'>
From a slide in Structure and Interpretation of Computer Programs, lecture 2A (1986), which in turns quotes Christopher Stracey:
The rights and privileges of first-class citizens:
- To be named by variables.
- To be passed as arguments to procedures.
- To be returned as values of procedures.
- To be incorporated into data structures
An easy way to understand is to ask: What things are not first class?
In Python, e.g. operators like + or * are not first class.
You can use the operators directly:
2+2
"foo"+"bar"
2*3
"bla"*3
But you can not assign operators to variables:
operator = +
operators = [+,*]
which means you can not pass operators as function arguments because…
… operators are not first class.