An object in AutoHotkey is an abstract data structure which provides three basic functions:
An object reference is a pointer or "handle" to a particular object. Like strings and numbers, object references can be stored in variables, passed to or returned from functions and stored in objects. After copying a reference from one variable to another as in x := y
, both variables refer to the same object.
IsObject can be used to determine if a value is an object:
Result := IsObject(expression)
There are currently three primary types of objects:
Create an array:
Array := [Item1, Item2, ..., ItemN] Array := Array(Item1, Item2, ..., ItemN)
Retrieve an item:
Value := Array[Index]
Assign an item:
Array[Index] := Value
Append an item:
Array.Insert(Value)
Insert one or more items at a given index:
Array.Insert(Index, Value, Value2, ...)
Remove an item:
RemovedValue := Array.Remove(Index)
If the array is not empty, MinIndex and MaxIndex return the lowest and highest index currently in use in the array. Since the lowest index is nearly always 1, MaxIndex usually returns the number of items. Looping through an array's contents can be done either by index or with a For-loop. For example:
array := ["one", "two", "three"] ; Iterate from 1 to the number of items: Loop % array.MaxIndex() MsgBox % array[A_Index] ; Enumerate the array's contents: For index, value in array MsgBox % "Item " index " is '" value "'"
An associative array is an object which contains a collection of unique keys and a collection of values, where each key is associated with one value. Keys can be strings, integers or objects, while values can be of any type. An associative array can be created as follows:
Array := {KeyA: ValueA, KeyB: ValueB, ..., KeyZ: ValueZ} Array := Object("KeyA", ValueA, "KeyB", ValueB, ..., "KeyZ", ValueZ)
Using the {key:value}
notation, quote marks are optional for keys which consist only of word characters. Any expression can be used as a key, but to use a variable as a key, it must be enclosed in parentheses. For example, {(KeyVar): Value}
and {GetKey(): Value}
are both valid.
Retrieve an item:
Value := Array[Key]
Assign an item:
Array[Key] := Value
Remove an item:
RemovedValue := Array.Remove(Key)
Enumerating items:
array := {ten: 10, twenty: 20, thirty: 30} For key, value in array MsgBox %key% = %value%
Associative arrays can be sparsely populated - that is, {1:"a",1000:"b"}
contains only two key-value pairs, not 1000.
By now, you may have noticed that associative arrays use very similar syntax to simple arrays. In fact, they're both the same thing in v1.x. However, treating []
as a simple linear array helps to keep its role clear, and improves the chance of your script working with a future version of AutoHotkey, which might change the implementation.
Retrieve a property:
Value := Object.Property
Set a property:
Object.Property := Value
Call a method:
ReturnValue := Object.Method(Parameters)
Call a method with a computed method name:
ReturnValue := Object[MethodName](Parameters)
Some properties of COM objects and user-defined objects can accept parameters:
Value := Object.Property[Parameters] Object.Property[Parameters] := Value
Related: Object, File Object, Func Object, COM object
Known limitation:
x.y[z]()
is treated as x["y", z]()
, which is not supported. As a workaround, `(x.y)[z]()
evaluates x.y
first, then uses the result as the target of the method call. Escaping the open-parenthesis (`(
) allows the expression to be used at the beginning of a line, where the parenthesis would otherwise begin a continuation section. Note that x.y[z].()
does not have this limitation since it is evaluated the same as (x.y[z]).()
.Array syntax (brackets) and object syntax (dots) can be used interchangeably.
Additionally, object references can themselves be used in expressions:
= == != <>
, they are considered equal only if both values are references to the same object.if obj
, !obj
or obj ? x : y
.&
address-of operator. This uniquely identifies the object from the point of its creation to the moment its last reference is released.If an object is used in any context where an object is not expected, it is treated as an empty string. For example, MsgBox %object%
shows an empty MsgBox and object + 1
yields an empty string. Do not rely on this behaviour as it may change.
When a method-call is followed immediately by an assignment operator, it is equivalent to setting a property with parameters. For example, the following are equivalent:
obj.item(x) := y obj.item[x] := y
Compound assignments such as x.y += 1
and --arr[1]
are supported.
Some limitations apply to which values can be used as keys in objects created with []
, {}
or the new
operator:
x[0x10]
, x[16]
and x[00016]
are equivalent. This also applies to numeric strings which don't have a decimal point.x[1]
and x["1"]
are not equivalent. Additionally, if a quoted literal string is concatenated with another value (as in "0x" x
), the result is treated as purely non-numeric. However, this does not apply to variables, so x[1]
and x[y:="1"]
are equivalent. This issue will be resolved in v2, so scripts should avoid using quoted numeric literals as keys.0+1.0
or Sqrt(y)
) are forced into the current float format. For consistency and clarity, scripts should avoid using floating-point literals as keys.If the variable func contains a function name, the function can be called one of two ways: %func%()
or func.()
. However, this requires the function name to be resolved each time, which is inefficient if the function is called more than once. To improve performance, the script can retrieve a reference to the function and store it for later use:
Func := Func("MyFunc")
To call a function by reference, the following syntax must be used:
RetVal := Func.(Params)
For details about additional properties of function references, see Func Object.
AutoHotkey supports "multi-dimensional" arrays by transparently storing arrays inside other arrays. For example, a table could be represented as an array of rows, where each row is itself an array of columns. In that case, the content of column y
of row x
can be set using either of the methods below:
table[x][y] := content ; A table[x, y] := content ; B
If table[x]
does not exist, A and B differ in two ways:
table[x]
.table
's base defines meta-functions, they are invoked as follows:
table.base.__Get(table, x)[y] := content ; A
table.base.__Set(table, x, y, content) ; B
Consequently, B allows the object to define custom behaviour for the overall assignment.This behaviour only applies to script-created objects, not more specialized types of objects such as COM objects or COM arrays.
An array of functions is simply an array containing function names or references. For example:
array := [Func("FirstFunc"), Func("SecondFunc")] ; Call each function, passing "foo" as a parameter: Loop 2 array[A_Index].("foo") ; Call each function, implicitly passing the array itself as a parameter: Loop 2 array[A_Index]() FirstFunc(param) { MsgBox % A_ThisFunc ": " (IsObject(param) ? "object" : param) } SecondFunc(param) { MsgBox % A_ThisFunc ": " (IsObject(param) ? "object" : param) }
For backward-compatibility, the second form will not pass array as a parameter if array[A_Index]
contains a function name instead of a function reference. However, if array[A_Index]
is inherited from array.base[A_Index]
, array will be passed as a parameter.
Objects in AutoHotkey are prototype-based rather than class-based. That is, an object can inherit properties and methods from its prototype or base
object, but do not need to have a pre-defined structure. Properties and methods can also be added to (or removed from) an object or any of the objects it derives from at any time. However, AutoHotkey emulates classes by translating class definitions into ordinary objects. For more complex or specialized situations, the base object can override the standard behaviour by defining meta-functions.
To create an object derived from another object, scripts can assign a base
or use the new
keyword:
baseObject := {foo: "bar"} obj1 := Object(), obj1.base := baseObject obj2 := {base: baseObject} obj3 := new baseObject MsgBox % obj1.foo " " obj2.foo " " obj3.foo
Prototype or base
objects are constructed and manipulated the same as any other object. For example, an ordinary object with one property and one method might be constructed like this:
; Create an object. thing := {} ; Store a value. thing.foo := "bar" ; Create a method by storing a function reference. thing.test := Func("thing_test") ; Call the method. thing.test() thing_test(this) { MsgBox % this.foo }
When thing.test()
is called, thing is automatically inserted at the beginning of the parameter list. However, for backward-compatibility, this does not occur when a function is stored by name (rather than by reference) directly in the object (rather than being inherited from a base object). By convention, the function is named by combining the "type" of object and the method name.
An object is a prototype or base if another object derives from it:
other := {} other.base := thing other.test()
In this case, other inherits foo and test from thing. This inheritance is dynamic, so if thing.foo
is modified, the change will be reflected by other.foo
. If the script assigns to other.foo
, the value is stored in other and any further changes to thing.foo
will have no effect on other.foo
. When other.test()
is called, its this parameter contains a reference to other instead of thing.
For convenience and familiarity, the "class" keyword can be used to construct a base object. A basic class definition might look like this:
class ClassName extends BaseClassName { var ClassVar := Expression class NestedClass { ... } Method() { ... } }
When the script is loaded, this constructs an object and stores it in the global variable ClassName. Therefore, to reference this class inside a function, a declaration such as global ClassName
is required unless the function is assume-global. If extends BaseClassName
is present, BaseClassName must be the full name of a previously defined class. The full name of each class is stored in object.__Class
.
Class definitions can contain class variable declarations, method definitions and nested class definitions.
var ClassVar := Expression
Class variable declarations such as the one above do little more than storing a value in the class object. As with any value or method contained by the class object, this value can be inherited by objects derived from ClassName. Expression is evaluated only once, before the script's auto-execute section. Class and static variable declarations are evaluated in the order they are defined in the script.
class NestedClass { ... }
Nested class definitions allow a class object to be stored inside another class object rather than a separate global variable. In the example above, class NestedClass
constructs an object and stores it in ClassName.NestedClass
. Therefore, NestedClass may be inherited by any class or object which derives from ClassName.
Method() { ... }
Method definitions look identical to function definitions. Each method has a hidden parameter named this
, which typically contains a reference to an object derived from the class. However, it could contain a reference to the class itself or a derived class, depending on how the method was called. Methods are stored by reference in the class object.
In addition to the hidden parameter this
, method definitions can use the pseudo-keyword base
to invoke the base class of the class which contains the method definition. For example, base.Method()
in the method above would be equivalent to BaseClassName.Method.(this)
, except that the global variable BaseClassName does not need to be declared. Note that this differs from this.base.base.Method()
in two other ways:
this
is derived from a sub-class of the current class.this
automatically, not this.base.base
.base
only has special meaning if followed by a dot .
or brackets []
, so code like obj := base, obj.Method()
will not work. Scripts can disable the special behaviour of base by assigning it a non-empty value; however, this is not recommended. Since the variable base must be empty, performance may be reduced if the script omits #NoEnv.
Whenever a derived object is created with the new
keyword [requires v1.1.00+], the __New
method defined by its base object is called. This method can accept parameters, initialize the object and override the result of the new
operator by returning a value. When an object is destroyed, __Delete
is called. For example:
m1 := new GMem(0, 20) m2 := {base: GMem}.__New(0, 30) class GMem { __New(aFlags, aSize) { this.ptr := DllCall("GlobalAlloc", "uint", aFlags, "ptr", aSize, "ptr") if !this.ptr return "" MsgBox % "New GMem of " aSize " bytes at address " this.ptr "." } __Delete() { MsgBox % "Delete GMem at address " this.ptr "." DllCall("GlobalFree", "ptr", this.ptr) } }
Meta-functions are methods defined by an object's base which can define exactly how the object should act when an unknown key is requested. For example, if obj.key
has not been assigned a value, it invokes the __Get meta-function. Similarly, obj.key := value
invokes __Set and obj.key()
invokes __Call.
In these cases, the base object is invoked as follows:
return
s, the return value is used as the result of the operation and no further processing occurs.
Set: If the operation was successful, __Set should return the new value of the field, which may differ from the original r-value. This allows assignments to be chained, as in a.x := b.y := z
. An empty string should be returned if the assignment failed, allowing the script to detect the error.
Get: If found, the value of that field is returned.
Call: If found, the function stored in the field (by name or reference) is called, passing the original target object as the this parameter.
If (and only if) no base object handles the operation, processing will continue as normal:
Otherwise a new key-value pair is created and stored in the object.
Known limitations:
return
without a value is equivalent to return ""
. This may be changed in a future version so that return
can be used to "escape" from a meta-function without overriding the default behaviour.__Get and __Set can be used to implement properties whose values are calculated or restricted in some way. For example, they could be used to implement a "Color" object with R, G, B and RGB properties, where only the RGB value is actually stored:
red := new Color(0xff0000), red.R -= 5 cyan := new Color(0x00ffff) MsgBox % "red: " red.R "," red.G "," red.B " = " red.RGB MsgBox % "cyan: " cyan.R "," cyan.G "," cyan.B " = " cyan.RGB class Color { __New(aRGB) { this.RGB := aRGB } __Get(aName) { if (aName = "R") return (this.RGB >> 16) & 255 if (aName = "G") return (this.RGB >> 8) & 255 if (aName = "B") return this.RGB & 255 } __Set(aName, aValue) { if aName in R,G,B { aValue &= 255 if (aName = "R") this.RGB := (aValue << 16) | (this.RGB & ~0xff0000) else if (aName = "G") this.RGB := (aValue << 8) | (this.RGB & ~0x00ff00) else ; (aName = "B") this.RGB := aValue | (this.RGB & ~0x0000ff) ; 'Return' must be used to indicate a new key-value pair should not be created. ; This also defines what will be stored in the 'x' in 'x := clr[name] := val': return aValue } } }
When a call such as obj.func(param)
is made, obj.func may contain a function name or an object. If obj.func contains an object, it is invoked using obj as the key. In most cases obj.func[obj]
does not exist and obj.func's __Call meta-function is invoked instead. This may be used to change the behaviour of function calls in an abstract way, as shown in the example below:
; Create a prototype for an array of functions. FuncArrayType := {__Call: "FuncType_Call"} ; Create an array of functions. funcArray := {1: "One", 2: "Two", base: FuncArrayType} ; Create an object which uses the array as a method. obj := {func: funcArray} ; Call the method. obj.func("foo", "bar") FuncType_Call(func, obj, params*) { ; Call a list of functions. Loop % ObjMaxIndex(func) func[A_Index](params*) } One(param1, param2) { ListVars Pause } Two(param1, param2) { ListVars Pause }
Combining this technique with class definitions provides a convenient way to define dynamic properties similar to those in the previous section:
blue := new Color(0x0000ff) MsgBox % blue.R "," blue.G "," blue.B class Properties { __Call(aTarget, aName, aParams*) { ; If this Properties object contains a definition for this half-property, call it. ; Take care not to use this.HasKey(aName), since that would recurse into __Call. if IsObject(aTarget) && ObjHasKey(this, aName) return this[aName].(aTarget, aParams*) } } class Color { __New(aRGB) { this.RGB := aRGB } class __Get extends Properties { R() { return (this.RGB >> 16) & 255 } G() { return (this.RGB >> 8) & 255 } B() { return this.RGB & 255 } } ;... }
When a multi-parameter assignment such as table[x, y] := content
implicitly causes a new object to be created, the new object ordinarily has no base and therefore no custom methods or special behaviour. __Set
may be used to initialize these objects, as demonstrated below.
x := {base: {addr: Func("x_Addr"), __Set: Func("x_Setter")}} ; Assign value, implicitly calling x_Setter to create sub-objects. x[1,2,3] := "..." ; Retrieve value and call example method. MsgBox % x[1,2,3] "`n" x.addr() "`n" x[1].addr() "`n" x[1,2].addr() x_Setter(x, p1, p2, p3) { x[p1] := new x.base } x_Addr(x) { return &x }
Since x_Setter
has four mandatory parameters, it will only be called when there are two or more key parameters. When the assignment above occurs, the following takes place:
x[1]
does not exist, so x_Setter(x,1,2,3)
is called ("..."
is not passed as there are too few parameters).x[1]
is assigned a new object with the same base as x
.x[1][2]
does not exist, so x_Setter(x[1],2,3,"...")
is called.x[1][2]
is assigned a new object with the same base as x[1]
.x[1][2][3]
does not exist, but since x_Setter
requires four parameters and there are only three (x[1][2], 3, "..."
), it is not called and the assignment completes as normal.When a non-object value is used with object syntax, the default base object is invoked. This can be used for debugging or to globally define object-like behaviour for strings, numbers and/or variables. The default base may be accessed by using .base
with any non-object value; for instance, "".base
. Although the default base cannot be set as in "".base := Object()
, the default base may itself have a base as in "".base.base := Object()
.
When an empty variable is used as the target of a set operation, it is passed directly to the __Set meta-function, giving it opportunity to insert a new object into the variable. For brevity, this example does not support multiple parameters; it could, by using a variadic function.
"".base.__Set := Func("Default_Set_AutomaticVarInit") empty_var.foo := "bar" MsgBox % empty_var.foo Default_Set_AutomaticVarInit(ByRef var, key, value) { if (var = "") var := Object(key, value) }
Object "syntax sugar" can be applied to strings and numbers.
"".base.__Get := Func("Default_Get_PseudoProperty") "".base.is := Func("Default_is") MsgBox % A_AhkPath.length " == " StrLen(A_AhkPath) MsgBox % A_AhkPath.length.is("integer") Default_Get_PseudoProperty(nonobj, key) { if (key = "length") return StrLen(nonobj) } Default_is(nonobj, type) { if nonobj is %type% return true return false }
Note that built-in functions may also be used, but in this case the parentheses cannot be omitted:
"".base.length := Func("StrLen") MsgBox % A_AhkPath.length() " == " StrLen(A_AhkPath)
If allowing a value to be treated as an object is undesirable, a warning may be shown whenever a non-object value is invoked:
"".base.__Get := "".base.__Set := "".base.__Call := Func("Default__Warn") empty_var.foo := "bar" x := (1 + 1).is("integer") Default__Warn(nonobj, p1="", p2="", p3="", p4="") { ListLines MsgBox A non-object value was improperly invoked.`n`nSpecifically: %nonobj% }
AutoHotkey uses a basic reference-counting mechanism to automatically free the resources used by an object when it is no longer referenced by the script. Script authors should not invoke this mechanism explicitly, except when dealing directly with unmanaged pointers to objects. For more information, see ObjAddRef.
; Increment the object's reference count to "keep it alive": ObjAddRef(address) ... ; Decrement the object's reference count to allow it to be freed: ObjRelease(address)
However, ObjAddRef does not need to be used when initially obtaining an address via Object(obj)
.
Generally each new copy of an object's address should be treated as an object reference, except that the script is responsible for calling ObjAddRef and/or ObjRelease as appropriate. For example, whenever an address is copied via something like x := address
, ObjAddRef should be called to increment the reference count. Similarly, whenever the script is finished with a particular copy of the object's address, it should call ObjRelease. This ensures the object is freed when the last reference to it is lost - and not before then.
To run code when the last reference to an object is being released, implement the __Delete meta-function.
Known Limitations:
; Create two objects. x := {}, y := {} ; Create a circular reference. x.child := y, y.parent := x ; BAD: x := "", y := "" ; Good: y.parent := "" ; Break the cycle. x := "", y := ""
Although memory used by the object is reclaimed by the operating system when the program exits, __Delete will not be called unless all references to the object are freed. This can be important if it frees other resources which are not automatically reclaimed by the operating system, such as temporary files.
In some rare cases it may be necessary to pass an object to external code via DllCall or store it in a binary data structure for later retrieval. An object's address can be retrieved by x := &obj
; however, if the variable obj is cleared, the object may be freed prematurely. To ensure this does not occur, use ObjAddRef as shown above or the Object()
function as shown below:
address := Object(object)
Additionally, this function can also be used to convert the address back into a reference:
object := Object(address)
In either case the object's reference-count is automatically incremented so that the object is not freed prematurely.
Note that this function applies equally to objects not created by Object(), such as COM object wrappers or File objects.