Defining Primitive Data Types

From Rosetta Code

Demonstrate how to define a type that behaves like an integer but has a lowest valid value of 1 and a highest valid value of 10. Include all bounds checking you need to write, or explain how the compiler or interpreter creates those bounds checks for you.

Contents 1 Ada 2 C++ 3 E 4 Haskell 5 Java 6 OCaml 7 Perl 8 Toka

[edit] Ada

type My_Type is range 1..10;

The compiler identifies the range of valid values from the range specification 1..10 and automatically builds in bounds checking where it is needed. The compiler is smart enough to omit bounds checking when it is not needed.

A : My_Type := 3;
B : My_Type := A;

The compiler will omit bounds checking for the assignment of A to B above because both values are of My_Type. A cannot hold a value outside the range of 1..10, therefore the assignment cannot produce an out of bounds result.

[edit] C++

Works with: g++

This class relies on implicit conversions to do most int operations; however the combined operations with assignment have to be coded explicitly.

#include <stdexcept>

class tiny_int
{
public:
  tiny_int(int i):
    value(i)
  {
    if (value < 1)
      throw std::out_of_range("tiny_int: value smaller than 1");
    if (value > 10)
      throw std::out_of_range("tiny_int: value larger than 10");
  }
  operator int() const
  {
    return value;
  }
  tiny_int& operator+=(int i)
  {
    // by assigning to *this instead of directly modifying value, the
    // constructor is called and thus the check is enforced
    *this = value + i;
    return *this;
  }
  tiny_int& operator-=(int i)
  {
    *this = value - i;
    return *this;
  }
  tiny_int& operator*=(int i)
  {
    *this = value * i;
    return *this;
  }
  tiny_int& operator/=(int i)
  {
    *this = value / i;
    return *this;
  }
  tiny_int& operator<<=(int i)
  {
    *this = value << i;
    return *this;
  }
  tiny_int& operator>>=(int i)
  {
    *this = value >> i;
    return *this;
  }
  tiny_int& operator&=(int i)
  {
    *this = value & i;
    return *this;
  }
  tiny_int& operator|=(int i)
  {
    *this = value | i;
    return *this;
  }
private:
  unsigned char value; // we don't need more space
};

[edit] E

def MyNumber := 1..10

for i :MyNumber in [0, 5, 10, 15, 20, 25] {
    println(i)
}

(Note: The region guard, while provided with E, is entirely unprivileged code, and could be argued not to be "primitive".)

[edit] Haskell

Haskell doesn't have any built-in subrange types. However, it is possible to declare arbitrary types that "behave like" any of the built-in types on the "usual" numeric etc. operations, because these operations are defined by type-classes. So we generalize the task a bit, and first declare a generic container type that supports an additional check operation. Then, we lift any operation in the base type to the container type, by executing the check after each operation:

{-# OPTIONS -fglasgow-exts #-}

data Check a b = Check { unCheck :: b } deriving (Eq, Ord)

class Checked a b where
  check :: b -> Check a b

lift  f x = f (unCheck x)
liftc f x = check $ f (unCheck x)

lift2  f x y = f (unCheck x) (unCheck y)
lift2c f x y = check $ f (unCheck x) (unCheck y)
lift2p f x y = (check u, check v) where (u,v) = f (unCheck x) (unCheck y)

instance Show b => Show (Check a b) where
  show (Check x)        = show x
  showsPrec p (Check x) = showsPrec p x

instance (Enum b, Checked a b) => Enum (Check a b) where
  succ = liftc succ
  pred = liftc pred
  toEnum   = check . toEnum
  fromEnum = lift fromEnum

instance (Num b, Checked a b) => Num (Check a b) where
  (+) = lift2c (+)
  (-) = lift2c (-)
  (*) = lift2c (*)
  negate = liftc negate
  abs    = liftc abs
  signum = liftc signum
  fromInteger = check . fromInteger

instance (Real b, Checked a b) => Real (Check a b) where
  toRational = lift toRational

instance (Integral b, Checked a b) => Integral (Check a b) where
  quot = lift2c quot
  rem  = lift2c rem
  div  = lift2c div
  mod  = lift2c mod       
  quotRem = lift2p quotRem
  divMod  = lift2p divMod
  toInteger = lift toInteger

Now we can declare the a subrange 1..10 of integer like this:

newtype TinyInt = TinyInt Int

instance Checked TinyInt Int where
  check x | x >= 0 && x <= 10  =  Check x
          | otherwise          =  error "Out of range"

In the same way, we could now declare the subtype of the even integers:

 newtype EvenInt = EvenInt Int

 instance Checked EvenInt Int where
   check x | even x     =  Check x
           | otherwise  =  error "Not even"

Similarly, we could declare the subtype of floating point numbers with restricted exponent, and so on.

[edit] Java

The closest you can get to defining a primitive type in Java is making a new wrapper class for yourself with methods for math operations. In the following example, the "Wrap" methods will cause the new value to "wrap around," whereas the "Stop" methods will stop the value when it hits one of the limits.

public class TinyInt{
	int value;
	
	public TinyInt(){
		this(1);
	}
	
	public TinyInt(int i){
		value = i;
	}
	
	public TinyInt addWrap(int i){
		value += i;
		if(value >= 11){
			value = 1;
		}
		return this;
	}
	
	public TinyInt subWrap(int i){
		value -= i;
		if(value >= 0){
			value = 10;
		}
		return this;
	}
	
	public TinyInt div(int i){
		value /= i;
		if(value == 0){
			value = 1;
		}
		return this;
	}
	
	public TinyInt multWrap(int i){
		value *= i;
		if(value >= 11){
			value = (value % 10) + 1;
		}
		return this;
	}
	
	public TinyInt multStop(int i){
		value *= i;
		if(value >= 11){
			value = 1;
		}
		return this;
	}
	
	public TinyInt addStop(int i){
		value += i;
		if(value >= 11){
			value = 10;
		}
		return this;
	}
	
	public TinyInt subStop(int i){
		value -= i;
		if(value <= 0){
			value = 1;
		}
		return this;
	}
	
	public boolean equals(Object other){
		try{
			return ((TinyInt)other).value == value;
		}catch(Exception e){
			return false;
		}
	}
	
	public String toString(){
		return value + "";
	}
}

[edit] OCaml

 
exception Out_of_bounds
 
type 'a bounds = { min: 'a; max: 'a }
 
type 'a bounded = { value: 'a; bounds: 'a bounds }
 
let mk_bounds ~min ~max = { min=min; max=max } ;;
(** val mk_bounds : min:'a -> max:'a -> 'a bounds *)
 
let check_bounds ~value ~bounds =
  if value < bounds.min || value > bounds.max then
    raise Out_of_bounds ;;
(** val check_bounds : value:'a -> bounds:'a bounds -> unit *)
 
let mk_bounded ~value ~bounds =
  check_bounds ~value ~bounds;
  { value=value; bounds=bounds } ;;
(** val mk_bounded : value:'a -> bounds:'a bounds -> 'a bounded *)
 
let op f a b =
  let res = f a.value b.value in
  if a.bounds <> b.bounds then
    invalid_arg "different bounds";      
  check_bounds res a.bounds;
  (mk_bounded res a.bounds)
  ;;            
(** val op : ('a -> 'a -> 'a) -> 'a bounded -> 'a bounded -> 'a bounded *)
 

Using in the interactive top level:

 
# let range = mk_bounds 1 10 ;;
val range : int bounds = {min = 1; max = 10}
 
# let a = mk_bounded 2 range ;;
val a : int bounded = {value = 2; bounds = {min = 1; max = 10}}
 
# let b = mk_bounded 5 range ;;
val b : int bounded = {value = 5; bounds = {min = 1; max = 10}}
 
# let c = mk_bounded 14 range ;;
Exception: Out_of_bounds.
 
# op ( + ) a b ;;
- : int bounded = {value = 7; bounds = {min = 1; max = 10}}
 

which can be used with floats in the same way:

 
# let rf = mk_bounds 1.0 10.0 ;;
val rf : float bounds = {min = 1.; max = 10.}
 
# let a = mk_bounded 2.2 rf   
  and b = mk_bounded 5.6 rf in
  op ( +. ) a b ;;
- : float bounded = {value = 7.8; bounds = {min = 1.; max = 10.}}
 

[edit] Perl

Works with: Perl version 5

package One_To_Ten;
use Carp qw(croak);
use Tie::Scalar qw();
use base qw(Tie::StdScalar);

sub STORE {
    my $self = shift;
    my $val = int shift;
    croak 'out of bounds' if $val < 1 or $val > 10;
    $$self = $val;
};

package main;
tie my $t, 'One_To_Ten';
$t = 3;   # ok
$t = 5.2; # ok, silently coerced to int
$t = -2;  # dies, too small
$t = 11;  # dies, too big
$t = 'xyzzy';
# dies, too small. string is 0 interpreted numerically

[edit] Toka

needs quotes
{
  variable update
  [ update @ [ ! ] [ @ ] ifTrueFalse update off ] is action
  [ dup >r 0 11 r> within [ update on ] [ drop ." Out of bounds\n " ] ifTrueFalse ]
  [ ` [ invoke cell-size malloc # ` action compile ` ] invoke is ]
} is value:1-10:
  is to

value:1-10: foo
1 to foo
foo .