Overloaded operators
An overloaded operator can be used on more than one data type, or represents a different action depending on the context. For example, if your language lets you use "+" for adding numbers and concatenating strings, then one would say that the "+" operator is overloaded.
- Task
Demonstrate overloaded operators in your language, by showing the different types of data they can or cannot operate on, and the results of each operation.
6502 Assembly
Many commands have multiple addressing modes, which alter the way a command is executed. On the 6502 most of these are in fact different opcodes, using the same mnemonic.
LDA #$80 ;load the value 0x80 (decimal 128) into the accumulator.
LDA $80 ;load the value stored at zero page memory address $80
LDA $2080 ;load the value stored at absolute memory address $2080.
LDA $80,x ;load the value stored at memory address ($80+x).
LDA ($80,x) ;use the values stored at $80+x and $81+x as a 16-bit memory address to load from.
LDA ($80),y ;use the values stored at $80 and $81 as a 16-bit memory address to load from. Load from that address + y.
68000 Assembly
Most assemblers will interpret instructions such as MOVE
, ADD
, etc. according to the operand types provided.
MOVE.L D0,A0 ;assembles the same as MOVEA.L D0,A0
EOR.W #%1100000000000000,D5 ;assembles the same as EORI.W #%1100000000000000,D5
CMP.W myAddress,D4 ;assembles the same as CMPM.W myAddress,D4
If you use MOVE.W
with an address register as the destination, the CPU will sign-extend the value once it's loaded into the address register. In other words, a word that is $8000 or "more" will get padded to the left with Fs, whereas anything $7FFF or less will be padded with zeroes.
MOVE.W #$9001,A0 ;equivalent of MOVEA.L #$FFFF9001,A0
MOVE.W #$200,A1 ;equivalent of MOVEA.L #$00000200,A1
Ada
Many Ada standard libraries overload operators. The following examples are taken from the Ada Language Reference Manual describing operators for vector and matrix types:
type Real_Vector is array (Integer range <>) of Real'Base;
type Real_Matrix is array (Integer range <>, Integer range <>) of Real'Base;
-- Real_Vector arithmetic operations
function "+" (Right : Real_Vector) return Real_Vector;
function "-" (Right : Real_Vector) return Real_Vector;
function "abs" (Right : Real_Vector) return Real_Vector;
function "+" (Left, Right : Real_Vector) return Real_Vector;
function "-" (Left, Right : Real_Vector) return Real_Vector;
function "*" (Left, Right : Real_Vector) return Real'Base;
function "abs" (Right : Real_Vector) return Real'Base;
-- Real_Vector scaling operations
function "*" (Left : Real'Base; Right : Real_Vector)
return Real_Vector;
function "*" (Left : Real_Vector; Right : Real'Base)
return Real_Vector;
function "/" (Left : Real_Vector; Right : Real'Base)
return Real_Vector;
-- Real_Matrix arithmetic operations
function "+" (Right : Real_Matrix) return Real_Matrix;
function "-" (Right : Real_Matrix) return Real_Matrix;
function "abs" (Right : Real_Matrix) return Real_Matrix;
function Transpose (X : Real_Matrix) return Real_Matrix;
function "+" (Left, Right : Real_Matrix) return Real_Matrix;
function "-" (Left, Right : Real_Matrix) return Real_Matrix;
function "*" (Left, Right : Real_Matrix) return Real_Matrix;
function "*" (Left, Right : Real_Vector) return Real_Matrix;
function "*" (Left : Real_Vector; Right : Real_Matrix)
return Real_Vector;
function "*" (Left : Real_Matrix; Right : Real_Vector)
return Real_Vector;
-- Real_Matrix scaling operations
function "*" (Left : Real'Base; Right : Real_Matrix)
return Real_Matrix;
function "*" (Left : Real_Matrix; Right : Real'Base)
return Real_Matrix;
function "/" (Left : Real_Matrix; Right : Real'Base)
return Real_Matrix;
The following examples are from the Ada package Ada.Numerics.Generic_Complex_Types
function "+" (Right : Complex) return Complex;
function "-" (Right : Complex) return Complex;
function Conjugate (X : Complex) return Complex;
function "+" (Left, Right : Complex) return Complex;
function "-" (Left, Right : Complex) return Complex;
function "*" (Left, Right : Complex) return Complex;
function "/" (Left, Right : Complex) return Complex;
function "**" (Left : Complex; Right : Integer) return Complex;
function "+" (Right : Imaginary) return Imaginary;
function "-" (Right : Imaginary) return Imaginary;
function Conjugate (X : Imaginary) return Imaginary renames "-";
function "abs" (Right : Imaginary) return Real'Base;
function "+" (Left, Right : Imaginary) return Imaginary;
function "-" (Left, Right : Imaginary) return Imaginary;
function "*" (Left, Right : Imaginary) return Real'Base;
function "/" (Left, Right : Imaginary) return Real'Base;
function "**" (Left : Imaginary; Right : Integer) return Complex;
function "<" (Left, Right : Imaginary) return Boolean;
function "<=" (Left, Right : Imaginary) return Boolean;
function ">" (Left, Right : Imaginary) return Boolean;
function ">=" (Left, Right : Imaginary) return Boolean;
function "+" (Left : Complex; Right : Real'Base) return Complex;
function "+" (Left : Real'Base; Right : Complex) return Complex;
function "-" (Left : Complex; Right : Real'Base) return Complex;
function "-" (Left : Real'Base; Right : Complex) return Complex;
function "*" (Left : Complex; Right : Real'Base) return Complex;
function "*" (Left : Real'Base; Right : Complex) return Complex;
function "/" (Left : Complex; Right : Real'Base) return Complex;
function "/" (Left : Real'Base; Right : Complex) return Complex;
function "+" (Left : Complex; Right : Imaginary) return Complex;
function "+" (Left : Imaginary; Right : Complex) return Complex;
function "-" (Left : Complex; Right : Imaginary) return Complex;
function "-" (Left : Imaginary; Right : Complex) return Complex;
function "*" (Left : Complex; Right : Imaginary) return Complex;
function "*" (Left : Imaginary; Right : Complex) return Complex;
function "/" (Left : Complex; Right : Imaginary) return Complex;
function "/" (Left : Imaginary; Right : Complex) return Complex;
function "+" (Left : Imaginary; Right : Real'Base) return Complex;
function "+" (Left : Real'Base; Right : Imaginary) return Complex;
function "-" (Left : Imaginary; Right : Real'Base) return Complex;
function "-" (Left : Real'Base; Right : Imaginary) return Complex;
function "*" (Left : Imaginary; Right : Real'Base) return Imaginary;
function "*" (Left : Real'Base; Right : Imaginary) return Imaginary;
function "/" (Left : Imaginary; Right : Real'Base) return Imaginary;
function "/" (Left : Real'Base; Right : Imaginary) return Imaginary;
ALGOL 68
This overrides the standard integer + operator (as in the F# sample) and provides an overloaded TOSTRING operator.
Also, the + operator is overloaded to operate on an INT left-hand operand and a BOOL right-hand operand.
Though not shown here, it is also possible to change the priorities of existing dyadic operators.
For new dyadic operators, the priority must be specified (though some implementations provide a default).
In both cases this is done with a PRIO declaration.
BEGIN
# Algol 68 allows operator overloading, both of existing operators and new ones #
# Programmer defined operators can be a "bold word" (uppercase word) or a symbol #
# Symbolic operators can be one or two characters, optionally followed by := or #
# =:, =: can also be defined as an operator (Allowed in Algol 68G, possibly not #
# in other implementations) #
# the characters allowed in a symbolic operator depends on the implementation #
# but would include +, -, *, /, <, =, > #
# define a new TOSTRING operator and overload it #
OP TOSTRING = ( INT n )STRING:
whole( n, 0 ); # returns a string representation of n in the minimum width #
OP TOSTRING = ( BOOL bv )STRING: IF bv THEN "true" ELSE "false" FI;
# overide a standard operator #
INT a = 10, b = 11, c = 21;
BEGIN
OP + = ( INT na, INT nb )INT: na - nb;
# + between strings is a standard operator that does string concation #
print( ( TOSTRING a, " ""+"" ", TOSTRING b, " = ", TOSTRING ( a + b ), " = " ) );
print( ( TOSTRING c, "? ", TOSTRING ( ( a + b ) = c ), newline ) )
END;
# same print, with the stndard + #
print( ( TOSTRING a, " + ", TOSTRING b, " = ", TOSTRING ( a + b ), " = " ) );
print( ( TOSTRING c, "? ", TOSTRING ( ( a + b ) = c ), newline ) );
# overload + to allow a BOOL to be added to an INT #
OP + = ( INT nv, BOOL bv )INT: IF bv THEN nv + 1 ELSE nv FI;
print( ( TOSTRING a, " ""+"" ", TOSTRING ( a = 10 ), " = ", TOSTRING ( a + ( a = 10 ) ), newline ) )
END
- Output:
10 "+" 11 = -1 = 21? false 10 + 11 = 21 = 21? true 10 "+" true = 11
BQN
What would generally be called operators in other languages are the basic functions in BQN. Nearly all functions have overloads, and the monadic and dyadic cases are generally inter-related.
For example, ⌊
is Floor in the monadic case:
⌊4.5
4
But in the dyadic case, it becomes Minimum:
4 ⌊ 3
3
C++
Operator overloading is one of the classic features of C++, but truth be told, I am writing such code after 20 years.......
And yes, I know subtracting cuboids can give negative volumes. It's theoretically possible (remember volume or triple integrals ? ).
//Aamrun, 4th October 2021
#include <iostream>
using namespace std;
class Cuboid {
private:
double length;
double breadth;
double height;
public:
double getVolume(void) {
return length * breadth * height;
}
void setLength( double l ) {
length = l;
}
void setBreadth( double b ) {
breadth = b;
}
void setHeight( double h ) {
height = h;
}
Cuboid operator +(const Cuboid& c) {
Cuboid biggerCuboid;
biggerCuboid.length = this->length + c.length;
biggerCuboid.breadth = this->breadth + c.breadth;
biggerCuboid.height = this->height + c.height;
return biggerCuboid;
}
Cuboid operator -(const Cuboid& c) {
Cuboid smallerCuboid;
smallerCuboid.length = this->length - c.length;
smallerCuboid.breadth = this->breadth - c.breadth;
smallerCuboid.height = this->height - c.height;
return smallerCuboid;
}
};
int main() {
Cuboid c1;
Cuboid c2;
Cuboid c3;
double volume = 0.0;
c1.setLength(6.0);
c1.setBreadth(7.0);
c1.setHeight(5.0);
c2.setLength(12.0);
c2.setBreadth(13.0);
c2.setHeight(10.0);
volume = c1.getVolume();
std::cout << "Volume of 1st cuboid : " << volume <<endl;
volume = c2.getVolume();
std::cout << "Volume of 2nd cuboid : " << volume <<endl;
//Adding the two cuboids
c3 = c1 + c2;
volume = c3.getVolume();
std::cout << "Volume of 3rd cuboid after adding : " << volume <<endl;
//Subtracting the two cuboids
c3 = c1 - c2;
volume = c3.getVolume();
std::cout << "Volume of 3rd cuboid after subtracting : " << volume <<endl;
return 0;
}
- Output:
Volume of 1st cuboid : 210 Volume of 2nd cuboid : 1560 Volume of 3rd cuboid after adding : 5400 Volume of 3rd cuboid after subtracting : -180
DuckDB
To understand polymorphism in DuckDB, it is helpful to remember that an operator or function can be applied successfully even if there is no restriction that matches exactly. This is because of implicit type-casting rules.
For example, the infix operator '+' has many restrictions, such as:
+(INTEGER, DATE) +(INTEGER, INTEGER) +(VARCHAR, VARCHAR) +(INTERVAL, INTERVAL)
but there is none for `+(FLOAT, INTEGER)`, and yet `1::FLOAT + 2` evaluates to 3.0 because the integer 2 is implicitly cast to a FLOAT.
It should also be noted that just because one type can be cast to another does not mean that that will automatically be done.
Another important point concerns the semantics of `NULL`: the fact that expressions such as `1::INTEGER + NULL` do not raise an error does not mean that `NULL` is regarded as an integer, or that there is a special retriction of '+' that allows `NULL` to return `NULL`. Rather, `NULL` just represents the absence of a value. This is in contrast to JSON, for example, where `null` is regarded as a distinct value. Since DuckDB supports `NULL` as well as JSON's `null`, the following table may be helpful:
D select 'null' as s, 'null'::JSON as j, NULL as sql, s = j, s = sql, j=j, sql=sql; ┌─────────┬──────┬───────┬─────────┬─────────────┬─────────┬─────────────────┐ │ s │ j │ sql │ (s = j) │ (s = "sql") │ (j = j) │ ("sql" = "sql") │ │ varchar │ json │ int32 │ boolean │ boolean │ boolean │ boolean │ ├─────────┼──────┼───────┼─────────┼─────────────┼─────────┼─────────────────┤ │ null │ null │ │ true │ │ true │ │ └─────────┴──────┴───────┴─────────┴─────────────┴─────────┴─────────────────┘
Notice that even though `s` and `j` have different types and are thus not identical, `s=j` evaluates to true because of implicit casting rules, as mentioned above.
Symbolic Operators
The comparison operators ('=', '!=', '<>', '<', '<=', '>', '>=') can be used symmetrically for any type, and the implicit type casting rules mean they can generally be used whenever the LHS and RHS are type-compatible.
Note in particular:
D select 1=true, 0=false, 2=true; ┌────────────────────────────┬────────────────────────────┬────────────────────────────┐ │ (1 = CAST('t' AS BOOLEAN)) │ (0 = CAST('f' AS BOOLEAN)) │ (2 = CAST('t' AS BOOLEAN)) │ │ boolean │ boolean │ boolean │ ├────────────────────────────┼────────────────────────────┼────────────────────────────┤ │ true │ true │ false │ └────────────────────────────┴────────────────────────────┴────────────────────────────┘
Other symbolic operators also have a wide berth. In brief:
- the modulo, bitwise, negation and factorial operators work only on integral data types;
- the other mathematical operators are available for all numeric data types;
- text-related operators (such as `~` and `!~` ) are defined for VARCHAR strings.
Logical Operators
The logical operators AND, OR and NOT have boolean signatures, but for the reasons already outlined above, they can be used with numeric and string arguments.
Concatenation
The `||` operator is primarily intended for concatenating strings (including bitstrings and BLOBs) to each other, for concatenating lists to each other, and for concatenating same-type arrays to each other, but once again the implicit type conversion rules allow a wide but not entirely predictable berth, as illustrated by the following typescript:
D select [1,2]::INTEGER[2] || [1,2]::FLOAT[2] as "success"; ┌──────────────────────┐ │ success │ │ float[] │ ├──────────────────────┤ │ [1.0, 2.0, 1.0, 2.0] │ └──────────────────────┘ D select [1,2]::INTEGER[2] || [3,4]; Binder Error: Cannot concatenate types INTEGER[2] and INTEGER[]
IN
The IN operator is applicable whenever the LHS value is type-compatible with the items specified by the RHS. The RHS could be a list, an array, a tuple (unnamed struct), or a column of any type.
Needless to say, `select NULL in (NULL) is NULL;` evaluates to true.
IS and IS NOT
`IS` and `IS NOT` handle NULL specially. Their type signatures can be summarized as follows:
IS(ANY, NULL) IS(BOOLEAN, BOOLEAN) IS(ANY, DISTINCT FROM ANY)
IS(ANY, NOT NULL) IS(ANY, NOT DISTINCT FROM ANY)
Array Indexing ([ ... ])
DuckDB supports the array indexing operator for retrieving elements by index and works with VARCHAR strings, lists, arrays, tuples, structs, and maps, as well as JSON objects and JSON arrays.
-> and -->
JSON support includes the infix operators `->` and `->>` with signatures characterized by:
op(JSON, BIGINT) -> JSON op(JSON, VARCHAR) -> JSON op(JSON, VARCHAR[]) -> JSON[]
F#
For those who complain that they can't follow my F# examples perhaps if I do the following it will help them.
// Overloaded operators. Nigel Galloway: September 16th., 2021
let (+) (n:int) (g:int) = n-g
printfn "%d" (23+7)
- Output:
16
FreeBASIC
Operators can be overloaded by default, so the Overload
keyword is not needed when declaring custom operators. At least one of the operator's parameters must be of a user-defined type (after all, operators with built-in type parameters are already defined).
The following example overloads the member operators Cast
(Cast) and *=
(Multiply And Assign) for objects of a user-defined type.
Type Rational
As Integer numerator, denominator
Declare Operator Cast () As Double
Declare Operator Cast () As String
Declare Operator *= (Byref rhs As Rational)
End Type
Operator Rational.cast () As Double
Return numerator / denominator
End Operator
Operator Rational.cast () As String
Return numerator & "/" & denominator
End Operator
Operator Rational.*= (Byref rhs As Rational)
numerator *= rhs.numerator
denominator *= rhs.denominator
End Operator
Dim As Rational r1 = (2, 3), r2 = (3, 4)
r1 *= r2
Dim As Double d = r1
Print r1, d
jq
Works with gojq, the Go implementation of jq
Many of jq's built-in operators are "overloaded" in the sense that they can be used on more than one built-in jq data type, these being: "null", "boolean", "string", "object" and "array".
The prime example of an overloaded operator in jq is `+`, which is defined on:
null x ANY # additive zero ANY x null # additive zero number x number # addition array x array # concatenation object x object # coalesence
Note that `+` is symmetric except for its restriction to object x object, as illustrated by:
{"a":1} + {"a": 2} #=> {"a": 2}
{"a":2} + {"a": 1} #=> {"a": 1}
Most of the other operators that are usually thought of as "arithmetic" are also overloaded, notably:
-: array x array # e.g. [1,2,1] - [1] #=> [2] *: string x number # e.g. "a" * 3 #=> "aaa" /: string x string # e.g. "a/b/c" / "/"' #=> ["a","b","c"]
The comparison operators (<, <=, ==, >=, >) are defined for all JSON entities and thus can be thought of as being overloaded, but this is only because jq defines a total order on JSON entities.
The comparison operators can also be used on non-JSON entities as well, e.g.
0 < infinite #=> true
nan < 0 #=> true
The logical operators (`and`, `or`, `not`) are also defined for all JSON entities, their logic being based on the idea that the only "falsey" values are `false` and `null`.
Whether a function (meaning a given name/arity pair) is "overloaded" or not depends entirely on its definition, it being understood that jq functions with the same name but different arities can have entirely unrelated definitions.
`length/0` is defined on all JSON entities except `true` and `false`. Note that it is defined as the absolute value on JSON numbers, and that:
nan|length #=> null
It is also worth pointing out that a single name/arity function can have multiple definitions within a single program, but normal scoping rules apply so that in any one context, only one definition is directly accessible. The functionality of the "outer" definition, however, can be accessed indirectly, as illustrated by the following contrived example:
def foo:
def outer_length: length;
def length: outer_length | tostring;
[outer_length, length];
"x" | foo #=> [1, "1"]
Julia
Most operators in Julia's base syntax are in fact just syntactic sugar for function calls. In particular, the symbols:
* / ÷ % & ⋅ ∘ × ∩ ∧ ⊗ ⊘ ⊙ ⊚ ⊛ ⊠ ⊡ ⊓ ∗ ∙ ∤ ⅋ ≀ ⊼ ⋄ ⋆ ⋇ ⋉ ⋊ ⋋ ⋌ ⋏ ⋒ ⟑ ⦸ ⦼ ⦾ ⦿ ⧶ ⧷ ⨇ ⨰ ⨱ ⨲ ⨳ ⨴ ⨵ ⨶ ⨷ ⨸ ⨻
⨼ ⨽ ⩀ ⩃ ⩄ ⩋ ⩍ ⩎ ⩑ ⩓ ⩕ ⩘ ⩚ ⩜ ⩞ ⩟ ⩠ ⫛ ⊍ ▷ ⨝ ⟕ ⟖ ⟗
are parsed in the same precedence as the multiplication operator function *, and the symbols:
+ - ⊕ ⊖ ⊞ ⊟ ∪ ∨ ⊔ ± ∓ ∔ ∸ ≏ ⊎ ⊻ ⊽ ⋎ ⋓ ⧺ ⧻ ⨈ ⨢ ⨣ ⨤ ⨥ ⨦ ⨧ ⨨ ⨩ ⨪ ⨫ ⨬ ⨭ ⨮ ⨹ ⨺ ⩁ ⩂ ⩅ ⩊ ⩌ ⩏ ⩐ ⩒ ⩔ ⩖ ⩗ ⩛ ⩝ ⩡ ⩢ ⩣
are parsed as infix operators with the same precedence as +. There are many other operator symbols that can be used as prefix or as infix operators once defined as a function in Julia.
As a language, much of Julia is organized around the concept of multiple dispatch. Because the language dispatches function calls according to the types of the function arguments, even the base arithmetic operators are in fact overloaded operators in Julia.
For example,
2 * 3
is sent to the function *(x::Int, y::Int)::Int, whereas
2.0 * 3.0
is dispatched to the overloaded function *(x::Float64, y::Float64)::Float64. Similarly, string concatenation in Julia is with * rather than +, so "hello " * "world" is dispatched to the overloaded function *(x::String, y::String)::String, and other types such as matrices also have arithmetic operators overloaded in base Julia:
julia> x = [1 2; 3 4]; y = [50 60; 70 80]; x + y
2×2 Matrix{Int64}:
51 62
73 84
Users may define their own overloaded functions in similar ways, whether or not such operators are already used in base Julia. In general, it is considered bad practice, and as "type piracy", to define a user overloaded operator which dispatches on the same types for which base Julia has already defined the same function. Instead, user defined types can be best made to have analogous operators defined for the new type so as to leverage existing code made for analogous base types. This can allow generic functions to use new types in efficient and constructive ways.
A simple example
The code below is just given as a simplistic example, since in practice the body of such simple one-liner functions would most likely be used without the overloading syntax.
import Base.-
""" overload - operator on vectors to return new vector from which all == subelem element are removed """
-(vec, subelem) where T = [elem for elem in vec if elem != subelem]
""" overload - operator on strings to return new string from which all == char c are removed """
-(s::String, c::Char) = String([ch for ch in s if ch != c])
@show [2, 3, 4, 3, 1, 7] - 3 # [2, 3, 4, 3, 1, 7] - 3 = [2, 4, 1, 7]
@show "world" - 'o' # "world" - 'o' = "wrld"
Mathematica /Wolfram Language
Define a custom vector object and define how plus works on these objects:
vec[{a_, b_}] + vec[{c_, d_}] ^:= vec[{a + c, b + d}]
vec[{4, 7}] + vec[{9, 3}]
- Output:
vec[{13, 10}]
Nim
Nim allows overloading of operators. There is no restrictions regarding types of arguments when overloading an operator. For instance, we may define a vector type and addition of vectors:
type Vector = tuple[x, y, z: float]
func `+`(a, b: Vector): Vector = (a.x + b.x, a.y + b.y, a.z + b.z)
echo (1.0, 2.0, 3.0) + (4.0, 5.0, 6.0) # print (x: 5.0, y: 7.0, z: 9.0)
The list of predefined operators with their precedence can be found here: https://rosettacode.org/wiki/Operator_precedence#Nim
Nim allows also user defined operators which must be composed using the following characters:
= + - * / < > @ $ ~ & % | ! ? ^ . : \
For instance, we may define an operator ^^
the following way:
func `^^`(a, b: int): int = a * a + b * b
To determine the precedence of user-defined operators, Nim defines a set of rules:
Unary operators always bind stronger than any binary operator: $a + b
is ($a) + b
and not $(a + b)
.
If an unary operator's first character is @
it is a sigil-like operator which binds stronger than a primarySuffix: @x.abc
is parsed as (@x).abc
whereas $x.abc
is parsed as $(x.abc)
.
For binary operators that are not keywords, the precedence is determined by the following rules:
Operators ending in either ->
, ~>
or =>
are called arrow like, and have the lowest precedence of all operators.
If the operator ends with =
and its first character is none of <
, >
, !
, =
, ~
, ?
, it is an assignment operator which has the second-lowest precedence.
Otherwise, precedence is determined by the first character.
Perl
See 'perldoc overload' for perl's overload capabilities. This example defines a class(package) that represent non-negative numbers as a string of 1's and overloads the basic math operators so that they can be used on members of that class(package). Also see 'Zeckendorf arithmetic' where overloading is used on Zeckendorf numbers.
use v5.36;
package Ones;
use overload qw("" asstring + add - subtract * multiply / divide);
sub new ( $class, $value ) { bless \('1' x $value), ref $class || $class }
sub asstring ($self, $other, $) { $$self }
sub asdecimal ($self, $other, $) { length $$self }
sub add ($self, $other, $) { bless \($$self . $$other), ref $self }
sub subtract ($self, $other, $) { bless \($$self =~ s/$$other//r), ref $self }
sub multiply ($self, $other, $) { bless \($$self =~ s/1/$$other/gr), ref $self }
sub divide ($self, $other, $) { $self->new( $$self =~ s/$$other/$$other/g ) }
package main;
my($x,$y,$z) = ( Ones->new(15), Ones->new(4) );
$z = $x + $y; say "$x + $y = $z";
$z = $x - $y; say "$x - $y = $z";
$z = $x * $y; say "$x * $y = $z";
$z = $x / $y; say "$x / $y = $z";
- Output:
111111111111111 + 1111 = 1111111111111111111 111111111111111 - 1111 = 11111111111 111111111111111 * 1111 = 111111111111111111111111111111111111111111111111111111111111 111111111111111 / 1111 = 111
Phix
Phix does not allow operator overloading and it is not possible to define new operators.
(Fairly weak arguments for, pretty strong against, any few minutes saved typing something in the first time are almost always lost the first time it needs to be maintained, if you want my opinion)
The standard arithmetic operators accept (mixed) integer and floating point values without casting.
The relational operators accept integer, float, string, and sequence values.
The logical operators only accept atoms, however there are 40-something sq_xxx() builtins that can be used
to perform all the builtin operations on any mix of integer, float, string, or sequence values.
Subscripts and concatenation work equivalently on strings and sequences, and in fact concatenation on integers and floats.
Any parameter can be integer, float, string, or sequence if it is declared as an object.
For example printf() can accept [a file number, format string and] a single atom or a sequence of objects,
it being wise to wrap lone strings in {} to ensure you get the whole thing not just the first letter.
Inline assembly mnemonics have multiple implicit addressing modes as per the standard intel syntax.
printf(1,"%g\n",3.5 + 3) -- 6.5 printf(1,"%t\n",3.5 > 3) -- true printf(1,"%t\n","a" = "a") -- true printf(1,"%t\n",{1} = {2}) -- false printf(1,"%V\n",{{1} & {2}}) -- {1,2} printf(1,"%V\n",{1 & 2.3}) -- {1,2.3} printf(1,"%V\n",{"a" & "b"}) -- "ab" printf(1,"%V\n",{"AB"[2] & {1,2}[1]}) -- {66,1} integer i #ilASM{ lea eax,[i] mov [eax],ebx mov [i],ebx mov [i],0 mov eax,ebx mov eax,1 -- etc }
Raku
While it is very easy to overload operators in Raku, it isn't really common... at least, not in the traditional sense. Or it's extremely common... It depends on how you view it.
First off, do not confuse "symbol reuse" with operator overloading. Multiplication * and exponentiation ** operators both use an asterisk, but one is not an overload of the other. A single asterisk is not the same as two. In fact the term * (whatever) also exists, but differs from both. The parser is smart enough to know when a term or an operator is expected and will select the correct one (or warn if they are not in the correct position).
For example:
1, 2, 1, * ** * * * … *
is a perfectly cromulent sequence definition in Raku. A little odd perhaps, but completely sensible. (It's the sequence starting with givens 1,2,1, with each term after the value of the term 3 terms back, raised to the power of the term two terms back, multiplied by the value one term back, continuing for some undefined number of terms. - Whatever to the whatever times whatever until whatever.)
One of the founding principles of Raku is that: "Different things should look different". It follows that "Similar things should look similar".
To pick out one tiny example: Adding numbery things together shouldn't be easily confusable with concatenating strings. Instead, Raku has the "concatenation" operator: ~ for joining stringy things together.
Using a numeric-ish operator implies that you want a numeric-ish answer... so Raku will try very hard to give you what you ask for, no matter what operands you pass it.
Raku operators have multiple candidates to try to fulfil your request and will try to coerce the operands to a sensible value.
Addition:
say 3 + 5; # Int plus Int
say 3.0 + 0.5e1; # Rat plus Num
say '3' + 5; # Str plus Int
say 3 + '5'; # Int plus Str
say '3' + '5'; # Str plus Str
say '3.0' + '0.5e1'; # Str plus Str
say (2, 3, 4) + [5, 6]; # List plus Array
+ is a numeric operator so every thing is evaluated numerically if possible
- Output:
8 8 8 8 8 8 5 # a list or array evaluated numerically returns the number of elements
Concatenation:
say 3 ~ 5; # Int concatenate Int
say 3.0 ~ 0.5e1; # Rat concatenate Num
say '3' ~ 5; # Str concatenate Int
say 3 ~ '5'; # Int concatenate Str
say '3' ~ '5'; # Str concatenate Str
say '3.0' ~ '0.5e1'; # Str concatenate Str
say (2, 3, 4) ~ [5, 6]; # List concatenate Array
~ is a Stringy operator so everything is evaluated as a string (numerics are evaluated numerically then coerced to a string).
- Output:
35 35 35 35 35 3.00.5e1 2 3 45 6 # default stringification, then concatenate
There is nothing preventing you from overloading or overriding existing
operators. Raku firmly believes in not putting pointless restrictions on
what you can and can not do. Why make it hard to do the "wrong" thing when
we make it so easy to do it right?
There is no real impetus to "overload" existing operators to do different things, it is very easy to add new operators in Raku, and nearly any Unicode character or combination may used to define it. They may be infix, prefix, postfix, (or post-circumfix!) The precedence, associativity and arity are all easily defined. An operator at heart is just a subroutine with funny calling conventions.
Borrowed from the Nimber arithmetic task:
New operators, defined in place. Arity is two (almost all infix operators take two arguments), precedence is set equivalent to similar existing operators, default (right) associativity. The second, ⊗, actually uses itself to define itself.
sub infix:<⊕> (Int $x, Int $y) is equiv(&infix:<+>) { $x +^ $y }
sub infix:<⊗> (Int $x, Int $y) is equiv(&infix:<×>) {
return $x × $y if so $x|$y < 2;
my $h = exp $x.lsb, 2;
return $h ⊗ $y ⊕ (($x ⊕ $h) ⊗ $y) if $x > $h;
return $y ⊗ $x if $y.lsb < $y.msb;
return $x × $y unless my $comp = $x.lsb +& $y.lsb;
$h = exp $comp.lsb, 2;
(($x +> $h) ⊗ ($y +> $h)) ⊗ (3 +< ($h - 1))
}
say 123 ⊗ 456;
- Output:
31562
Base Raku has 27 different operator precedence levels for built-ins. You could theoretically give a new operator an absolute numeric precedence but it would be difficult to predict exactly what the relative precedence would be. Instead, precedence is set by setting a relative precedence; either equivalent to an existing operator, or, by setting it tighter(higher) or looser(lower) precedence than an existing operator. When tighter or looser precedence is specified, a whole new precedence level is created squeezed in between the named level and its immediate successor (predecessor). The task Exponentiation with infix operators in (or operating on) the base demonstrates three different operators that nominally do the same thing, but may yield different results due to differing precedence levels.
That's all well and good, but suppose you have a new class, say, a Line class, and
you want to be able to do arithmetic on Lines. No need to override the built
in arithmetic operators, just add a multi candidate to do the right thing. A multi
allows adding a new definition of the operator without disturbing the existing ones.
Very, very basic Line class:
class Line {
has @.start;
has @.end;
}
# New infix + multi to add two Lines together, for some bogus definition of add
multi infix:<+> (Line $x, Line $y) {
Line.new(
:start(
sqrt($x.start[0]² + $y.start[0]²),
sqrt($x.start[1]² + $y.start[1]²)
),
:end(
sqrt($x.end[0]² + $y.end[0]²),
sqrt($x.end[1]² + $y.end[1]²)
)
)
}
# In operation:
say Line.new(:start(-4,7), :end(5,0)) + Line.new(:start(1,1), :end(2,3));
- Output:
Line.new(start => [4.123105625617661e0, 7.0710678118654755e0], end => [5.385164807134504e0, 3e0])
To be fair, all of this easy power in a bad programmers hands can lead to incomprehensible code... but bad programmers can be bad in any language.
REXX
A lot of the examples were taken from the Raku examples.
The REXX language has the "normal" (as say, compared with PL/I) overloading of:
- the prefix operators (+ and -) which are shared with the addition and subtraction operators,
- the multiplication operator (*) is "shared" with the exponentiation operator (**),
- the "or" operator (|) is "shared" with the concatenation operator (||),
- the "and" operator (&) is "shared" with the "XOR" (eXclusive OR) operator (&&), and
- the "negation" operator (\) is "shared" with the "not" logical comparison operator, as in: if a\=b then ...
Note that some REXXes may also have other characters (glyphs) for the negation operator (not) such as: ^ and/or ¬ glyphs.
/*REXX pgm shows overloading of some operators: prefix/addition/subtraction/concatenate.*/
say '──positive prefix──'
say +5 /* positive prefix integer */
say + 5 /* positive prefix integer */
say ++6 /* positive prefix integer */
say ++ 6 /* positive prefix integer */
say +++7 /* positive prefix integer */
say +++ 7 /* positive prefix integer */
say + + + + 8 /* positive prefix integer */
say + (9) /* positive prefix integer */
say '──negative prefix──'
say -1 /* negative prefix integer */
say - 1 /* negative prefix integer */
say --2 /* negative prefix integer */
say -- 2 /* negative prefix integer */
say ---3 /* negative prefix integer */
say --- 3 /* negative prefix integer */
say - - - - 4 /* negative prefix integer */
say - (9) /* negative prefix integer */
say '───addition───'
say 3 + 5 /* integer plus integer */
say 3 + (5) /* integer plus integer */
say 3.0 + 0.5e1 /* rational plus number */
say '3' + 5 /* string plus integer */
say 3 + ' 5 ' /* integer plus string */
say 3 + '5' /* integer plus string */
say '3' + '5' /* string plus string */
say '3' + "5" /* string plus string */
say '3.0' + '0.5e1' /* string plus string */
say '──subtraction──'
say 3 - 5 /* integer minus integer */
say 3 - (5) /* integer minus integer */
say 3.0 - 0.5e1 /* rational minus number */
say '3' - 5 /* string minus integer */
say 3 - '5' /* integer minus string */
say 3 - ' 5 ' /* integer minus string */
say '3' - '5' /* string minus string */
say '3' - "5" /* string minus string */
say '3.0' - '0.5e1' /* string minus string */
say '──concatenation──'
say 3 || 5 /* integer concatenated integer */
say 3 || (5) /* integer concatenated integer */
say 3.0 || 0.5e1 /* rational concatenated number */
say '3' || 5 /* string concatenated integer */
say 3 || '5' /* integer concatenated string */
say '3' || '5' /* string concatenated string */
say "3" || "5" /* string concatenated string */
say '3.0' | | '0.5e1' /* string concatenated string */
say 3 || ' 5 '. /* integer concatenated strings */
say '────abutment────'
say 3 5 /* integer abutted integer */
say 3 (5) /* integer abutted integer */
say 3.0 0.5e1 /* rational abutted number */
say '3' 5 /* string abutted integer */
say 3 '5' /* integer abutted string */
say '3' '5' /* string abutted string */
say "3" "5" /* string abutted string */
say 3 ' 5 '. /* integer abutted strings */
say '──multiplication──'
say 3 * 5 /* integer multiplied integer */
say 3 * (5) /* integer multiplied integer */
say 3.0 * 0.5e1 /* rational multiplied number */
say '3' * 5 /* string multiplied integer */
say 3 * '5' /* integer multiplied string */
say '3' * '5' /* string multiplied string */
say "3" * "5" /* string multiplied string */
say '3.0' * '0.5e1' /* string multiplied string */
say '──exponentation──'
say 3 ** 5 /* integer exponetiated integer */
say 3 ** (5) /* integer exponetiated integer */
say 3 * * 5 /* integer exponetiated integer */
say 3.0 ** 0.5e1 /* rational exponetiated number */
say '3' ** 5 /* string exponetiated integer */
say 3 ** '5' /* integer exponetiated string */
say '3' ** '5' /* string exponetiated string */
say "3" ** "5" /* string exponetiated string */
say '3.0' ** '0.5e1' /* string exponetiated string */
say '────division────'
say 3 / 5 /* integer divided integer */
say 3 / (5) /* integer divided integer */
say 3.0 / 0.5e1 /* rational divided number */
say '3' / 5 /* string divided integer */
say 3 / '5' /* integer divided string */
say '3' / '5' /* string divided string */
say "3" / "5" /* string divided string */
say '3.0' / '0.5e1' /* string divided string */
say '─────not────'
say \0 /* (not) invert binary */
say \1 /* (not) invert binary */
say \ 1 /* (not) invert binary */
say \ (0) /* (not) invert binary */
say \ 1 /* (not) invert binary */
say \ (0) /* (not) invert binary */
say \\ 0 /* (not) (not) invert binary */
say \ \ 1 /* (not) (not) invert binary */
say '─────or─────'
say 0 | 0 /* binary OR'ed binary */
say 0 | 1 /* binary OR'ed binary */
say '0' | "1" /* binary OR'ed binary */
say '1' | 0 /* binary OR'ed binary */
say '1' | (0) /* binary OR'ed binary */
say '─────and────'
say 0 & 0 /* binary AND'ed binary */
say 0 & 1 /* binary AND'ed binary */
say '0' & "1" /* binary AND'ed binary */
say '1' & 0 /* binary AND'ed binary */
say '1' & (0) /* binary AND'ed binary */
say '─────XOR────'
say 0 && 0 /* binary XOR'ed binary */
say 0 && 1 /* binary XOR'ed binary */
say '0' && "1" /* binary XOR'ed binary */
say '1' && 0 /* binary XOR'ed binary */
say '1' && (0) /* binary XOR'ed binary */
exit 0 /*stick a fork in it, we're all done. */
- output when using the internal default input:
──positive prefix── 5 5 6 6 7 7 8 9 ──negative prefix── -1 -1 2 2 -3 -3 4 -9 ───addition─── 8 8 8.0 8 8 8 8 8 8.0 ──subtraction── -2 -2 -2.0 -2 -2 -2 -2 -2 -2.0 ──concatenation── 35 35 3.00.5E1 35 35 35 35 3.00.5e1 3 5 . ────abutment──── 3 5 3 5 3.0 0.5E1 3 5 3 5 3 5 3 5 3 5 . ──multiplication── 15 15 15.0 15 15 15 15 15.0 ──exponentation── 243 243 243 243 243 243 243 243 243 ────division──── 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 ─────not──── 1 0 0 1 0 1 0 1 ─────or───── 0 1 1 1 1 ─────and──── 0 0 0 0 0 ─────XOR──── 0 1 1 1 1
Wren
All of Wren's operators can be overloaded except: &&, ||, ?: and =. It is not possible to create new operators from scratch.
When an operator is overloaded it retains the same arity, precedence and associativity as it has when used in its 'natural' sense.
The standard library contains several instances of overloading the + and * operators which are demonstrated below.
Otherwise, operator overloading can be used without restriction in user defined classes.
However, whilst it is very useful for classes representing mathematical objects, it should otherwise be used sparingly as code can become unreadable if it is used inappropriately.
import "./date" for Date
var s1 = "Rosetta "
var s2 = "code"
var s3 = s1 + s2 // + operator used to concatenate two strings
System.print("s3 = %(s3)")
var s4 = "a" * 20 // * operator used to provide string repetition
System.print("s4 = %(s4)")
var l1 = [1, 2, 3] + [4] // + operator used to concatenate two lists
System.print("l1 = %(l1)")
var l2 = ["a"] * 8 // * operator used to create a new list by repeating another
System.print("l2 = %(l2)")
// the user defined class Date overloads the - operator to provide the interval between two dates
var d1 = Date.new(2021, 9, 11)
var d2 = Date.new(2021, 9, 13)
var i1 = (d2 - d1).days
System.print("i1 = %(i1) days")
- Output:
s3 = Rosetta code s4 = aaaaaaaaaaaaaaaaaaaa l1 = [1, 2, 3, 4] l2 = [a, a, a, a, a, a, a, a] i1 = 2 days