Scroll to navigation

nx::Class(3nx) NX API nx::Class(3nx)


NAME

nx::Class - API reference of the base-metaclass of the NX objectsystem

SYNOPSIS

nx::Class create cls ?-superclasses superClassNames? ?-mixins mixinSpec? ?-filters filterSpec? ?option value ...? ?initBlock?

nx::Class new ?-superclasses superClassNames? ?-mixins mixinSpec? ?-filters filterSpec? ?initBlock?

cls ?public | private | protected? alias ?-debug? ?-deprecated? methodName ?-returns valueChecker? ?-frame object | method? cmdName

cls create instanceName ?option value option value ...?

cls delete feature arg

cls filters submethod ?arg ...?

cls ?public | protected | private? forward ?-debug? ?-deprecated? methodName ?-prefix prefixName? ?-frame object? ?-returns valueChecker? ?-verbose? ?target? ?arg ...?

cls info heritage ?pattern?

cls info instances ?-closure? ?pattern?

cls info mixinof ?-closure? ?-scope option? ?pattern?

cls info subclasses ?-closure? ?-dependent? ?pattern?

cls info superclasses ?-closure? ?pattern?

cls info info ?-asList?

cls info filters ?-guards? ?pattern?

cls info method option methodName

cls info methods ?-callprotection level? ?-type methodType? ?-path? ?namePattern?

cls info mixins ?-guards? ?pattern?

cls info slots ?-type className? ?pattern?

cls info variables ?pattern?

cls ?public | protected | private? method ?-debug? ?-deprecated? name parameters ?-checkalways? ?-returns valueChecker? body

cls mixins submethod ?arg ...?

cls new ?-childof parentName? ?option value option value ...?

cls property ?-accessor public | protected | private? ?-class className? ?-configurable trueFalse? ?-incremental? ?-trace set | get | default? spec ?initBlock?

cls require ?public | protected | private? method methodName

cls variable ?-accessor public | protected | private? ?-incremental? ?-class className? ?-configurable trueFalse? ?-initblock ?-trace set | get | default? script? spec ?defaultValue?


DESCRIPTION

nx::Class is the base metaclass of the NX object system. All classes (e.g. cls) are (direct or indirect) instances of nx::Class. Therefore, the methods provided by nx::Class are available to all classes. A class cls which does not have nx::Class as its direct or indirect superclass is referred to as an application class. By default, when instantiating a new class from nx::Class, it becomes an application class with nx::Object being set as its superclass. A class cls which is explicitly declared as a (direct or indirect) subclass of nx::Class is referred to as a metaclass, that is, its instances will become classes as well. In other words, a metaclass instantiates and subclasses nx::Class at the same time.

+---------+
| ::nx::* |
+---------+--------------------------------------Y
|                                                |
|  instance of                                   |
|   .-------.                                    |
|  +--------'+     instance of     +----------+  |
|  |         |<....................|          |  |
|  |  Class  |                     |  Object  |  |
|  |         |....................>|          |  |
|  +---------+     subclass of     +-----+----+  |
|   ^   ^                                ^       |
\...|...|................................|......./

| | |
| |subclass.....(xor)......subclass|
| |of +-----------+ of|
| |.........| |..........|
| (metaclass) | /cls/ | (application class)
|.............| |
instance of +-----------+
Classes can be created in the following ways:
To create a class having the explicit name cls, use create.
To create a class having an automatically assigned, implicit name, use new.

The configuration options for direct and indirect instances of nx::Class, which can be passed when calling create and new, are documented in the subsequent section.

CONFIGURATION OPTIONS FOR INSTANCES OF NX::CLASS

Configuration options can be used for configuring objects during their creation by passing the options as non-positional arguments into calls of new and create (see nx::Class). An existing object can be queried for its current configuration using cget and it can be re-configured using configure.

If superClassNames is not specified, returns the superclasses of the class. If provided, the class becomes the subclass of superClassNames.
Retrieves the list of filter methods currently active on instances of the class, if filterSpecs is not set. Otherwise, activates a list of filter methods for the instances of the class. Filters are returned or set in terms of a list of filter specifications.
Returns the list of mixin classes currently active on instances of the class, if mixinSpecs is not specified. Otherwise, the class is extended by the list of mixin classes provided by mixinSpecs. mixin classes are returned or set in terms of a list of mixin specifications.

The configuration options provided by nx::Object are equally available because an application class cls is an indirect instance of nx::Object.

METHODS FOR INSTANCES OF NX::CLASS

Define an alias method for the given class. The resulting method registers a pre-existing Tcl command cmdName under the (alias) name methodName with the class. If cmdName refers to another method, the corresponding argument should be a valid method handle. If a Tcl command (e.g., a proc), the argument should be a fully qualified Tcl command name. If aliasing a subcommand (e.g., array exists) of a Tcl namespace ensemble (e.g., array), cmdName must hold the fully qualified subcommand name (and not the ensemble name of the subcommand).

As for a regular class method, -returns allows for setting a value checker on the values returned by the aliased command cmdName.

When creating an alias method for a C-implemented Tcl command (i.e., command defined using the Tcl/NX C-API), -frame sets the scope for variable references used in the aliased command. If the provided value is object, then variable references will be resolved in the context of the called object, i.e., the object upon which the alias method is invoked, as if they were object variables. There is no need for using the colon-prefix notation for identifying object variables. If the value is method, then the aliased command will be executed as a regular method call. The command is aware of its called-object context; i.e., it can resolve ::nx::self. In addition, the alias method has access to the method-call context (e.g., nx::next). If -frame is omitted, and by default, the variable references will resolve in the context of the caller of the alias method.

To express deprecation of the alias method methodName, set the -deprecated flag. Deprecated methods remain usable from client code, but their usage will be signaled to the developer and/or can be tracked using ::nsf::deprecated. To register methodName with the debugger, set the -debug flag. Entering and exiting a method, which was flagged for debugging, is recorded by calling the redefinable callback procs ::nsf::debug::call and ::nsf::debug::exit, respectively. By default, these callbacks forward to ::nsf::log, which can also be customized at the script level.

__class_configureparameter
Computes and returns the configuration options available for cls instances, to be consumed as method-parameter specification by configure.
This factory method creates an instance instanceName of cls and returns instanceName.
% nx::Class create AClass {

:method init args {
next
}; # initialization method for instances of 'AClass'
}; # defines a class 'AClass' being an instance of 'nx::Class' ::AClass % ::AClass create anInstance; # defines an object 'anInstance' being an instance of 'AClass' ::anInstance % ::anInstance info class ::AClass % ::AClass info class ::nx::Class
create accepts the configuration options option available for this instance, such as those defined by properties of cls (see property).

Note that create is called internally when defining an instance of cls using new.

By calling create on nx::Class itself, the created instance will become a new application class instanceName on which create can also be applied (i.e., it can be instantiated). If the so-created class has ::nx::Class its direct or indirect superclass, instanceName is referred to as a metaclass; that is, a class whose instances are again classes.

This method serves as the equivalent to Tcl's rename for removing structural (properties, variables) and behavioral features (methods) of the class:
Removes a property propertyName, variable variableName, and method methodName, respectively, previously defined for the scope of the class.

delete method can be equally used for removing regular methods (see method), an alias method (see alias), and a forwarder method (see forward).

Accesses and modifies the list of methods which are registered as filters with cls using a specific setter or getter submethod:
Inserts a single filter into the current list of filters of cls. Using index, a position in the existing list of filters for inserting the new filter can be set. If omitted, index defaults to the list head (0).
Removes all filters from cls and returns the list of removed filters. Clearing is equivalent to passing an empty list for filterSpecList to class filter set.
Removes a single filter from the current list of filters of cls whose spec matches specPattern. specPattern can contain special matching chars (see string match). class filters delete will throw an error if there is no matching filter, unless -nocomplain is set.
Returns the list of current filter specifications registered for cls.
If expr is specified, registers a guard expression expr with a filter methodName. This requires that the filter methodName has been previously set using filters set or added using filters add. expr must be a valid Tcl expression (see expr). An empty string for expr will clear the currently registered guard expression for filter methodName.

If expr is omitted, returns the guard expression set on the filter methodName defined for cls. If none is available, an empty string will be returned.

If pattern is omitted, returns all filter names which are defined by cls. By specifying pattern, the returned filters can be limited to those whose names match patterns (see string match).
filterSpecList takes a list of filter specs, with each spec being itself either a one-element or a two-element list: methodName ?-guard guardExpr?. methodName identifies an existing method of cls which becomes registered as a filter. If having three elements, the third element guardExpr will be stored as a guard expression of the filter. This guard expression must be a valid Tcl expression (see expr). expr is evaluated when cls receives a message to determine whether the filter should intercept the message. Guard expressions allow for realizing context-dependent or conditional filter composition.
Every methodName in a spec must resolve to an existing method in the scope of the class. To access and to manipulate the list of filters of cls, cget|configure -filters can also be used.
Define a forward method for the given class. The definition of a forward method registers a predefined, but changeable list of forwarder arguments under the (forwarder) name methodName. Upon calling the forward method, the forwarder arguments are evaluated as a Tcl command call. That is, if present, target is interpreted as a Tcl command (e.g., a Tcl proc or an object) and the remainder of the forwarder arguments arg as arguments passed into this command. The actual method arguments to the invocation of the forward method itself are appended to the list of forwarder arguments. If target is omitted, the value of methodName is implicitly set and used as target. This way, when providing a fully-qualified Tcl command name as methodName without target, the unqualified methodName (namespace tail) is used as the forwarder name; while the fully-qualified one serves as the target.

As for a regular method, -returns allows for setting a value checker on the values returned by the resulting Tcl command call. When passing object to -frame, the resulting Tcl command is evaluated in the context of the object receiving the forward method call. This way, variable names used in the resulting execution of a command become resolved as object variables.

To express deprecation of the forward method methodName, set the -deprecated flag. Deprecated methods remain usable from client code, but their usage will be signaled to the developer and/or can be tracked using ::nsf::deprecated. To register methodName with the debugger, set the -debug flag. Entering and exiting a method, which was flagged for debugging, is recorded by calling the redefinable callback procs ::nsf::debug::call and ::nsf::debug::exit, respectively. By default, these callbacks forward to ::nsf::log, which can also be customized at the script level.

The list of forwarder arguments arg can contain as its elements a mix of literal values and placeholders. Placeholders are prefixed with a percent symbol (%) and substituted for concrete values upon calling the forward method. These placeholders allow for constructing and for manipulating the arguments to be passed into the resulting command call on the fly:

  • %method becomes substituted for the name of the forward method, i.e. methodName.
  • %self becomes substituted for the name of the object receiving the call of the forward method.
  • %1 becomes substituted for the first method argument passed to the call of forward method. This requires, in turn, that at least one argument is passed along with the method call.

    Alternatively, %1 accepts an optional argument defaults: {%1 defaults}. defaults must be a valid Tcl list of two elements. For the first element, %1 is substituted when there is no first method argument which can be consumed by %1. The second element is inserted upon availability of a first method argument with the consumed argument being appended right after the second list element. This placeholder is typically used to define a pair of getter/setter methods.

  • {%@index value} becomes substituted for the specified value at position index in the forwarder-arguments list, with index being either a positive integer, a negative integer, or the literal value end (such as in Tcl's lindex). Positive integers specify a list position relative to the list head, negative integers give a position relative to the list tail. Indexes for positioning placeholders in the definition of a forward method are evaluated from left to right and should be used in ascending order.

    Note that value can be a literal or any of the placeholders (e.g., %method, %self). Position prefixes are exempted, they are evaluated as %cmdName-placeholders in this context.

  • {%argclindex list} becomes substituted for the nth element of the provided list , with n corresponding to the number of method arguments passed to the forward method call.
  • %% is substituted for a single, literal percent symbol (%).
  • %cmdName is substituted for the value returned from executing the Tcl command cmdName. To pass arguments to cmdName, the placeholder should be wrapped into a Tcl list: {%cmdName ?arg ...?}.

    Consider using fully-qualified Tcl command names for cmdName to avoid possible name conflicts with the predefined placeholders, e.g., %self vs. %::nx::self.

To disambiguate the names of subcommands or methods, which potentially become called by a forward method, a prefix prefixName can be set using -prefix. This prefix is prepended automatically to the argument following target (i.e., a second argument), if present. If missing, -prefix has no effect on the forward method call.

To inspect and to debug the conversions performed by the above placeholders, setting the switch -verbose will have the command list to be executed (i.e., after substitution) printed using ::nsf::log (debugging level: notice) upon calling the forward method.

A collection of introspection submethods on the structural features (e.g. configuration options, superclasses) and the behavioral features (e.g. methods, filters) provided by cls to its instances.
If pattern is omitted, returns the list of object names of all the direct and indirect superclasses and per-class mixin classes of cls, in their order of precedence, which are active for instances of cls. If pattern is specified, only superclasses and mixin classes whose names match pattern will be listed (see string match).
If pattern is not specified, returns a list of the object names of all the direct instances of cls. If the switch -closure is set, indirect instances are also returned. A direct instance is created by using create or new on cls, an indirect instance was created from a direct or indirect subclass of cls. If pattern is specified, only instances whose names match pattern will be listed (see string match).
If pattern is not specified, returns a list of the object names of all the objects for which cls is active as a direct mixin class. If the switch -closure is set, objects which have cls as an indirect mixin class are also returned. If pattern is specified, only objects whose names match pattern will be listed (see string match). Valid values of option are all, object, and class. Passing object will have only objects returned which have cls as per-object mixin class. Passing class will have only classes returned which have cls as per-class mixin class. all (the default) will have contained both in the returned list.
If pattern is not specified, returns a list of the object names of the direct subclasses of cls. If the switch -closure is set, indirect subclasses are also returned. If the switch -dependent is on, indirect subclasses introduced by mixin class relations of subclasses of cls are also reported. -closure and -dependent are mutually exclusive. If pattern is specified, only subclasses whose names match pattern will be listed (see string match).
If pattern is not specified, returns a list of the object names of all direct superclasses of cls. If the switch -closure is set, indirect superclasses will also be returned. If pattern is specified, only superclasses whose names match pattern will be listed (see string match).
Returns the available submethods of the info method ensemble for cls, either as a pretty-printed string or as a Tcl list (if the switch -asList is set) for further processing.
If pattern is omitted, returns all filter names which are defined by cls. By turning on the switch -guards, the corresponding guard expressions, if any, are also reported along with each filter as a three-element list: filterName -guard guardExpr. By specifying pattern, the returned filters can be limited to those whose names match patterns (see string match).
This introspection submethod provides access to the details of methodName provided by cls. If methodName is not the name of an existing method, an empty string is returned. To disambiguate between a non-existing method and an empty string as valid return value (e.g., for info method args|parameters|args|...), use info method exists.

Permitted values for option are:

  • args returns a list containing the parameter names of methodName, in order of the method-parameter specification.
  • body returns the body script of methodName.
  • callprotection returns the call-protection level set for methodName; possible values: public, protected, private.
  • debug returns 1 if methodName is in debug mode, 0 otherwise.
  • definition returns a canonical command list which allows for (re-)define methodName.
  • definitionhandle returns the method handle for a submethod in a method ensemble from the perspective of cls as method provider. methodName must contain a complete method path.
  • deprecated returns 1 if methodName is deprecated, 0 otherwise.
  • exists returns 1 if there is a methodName provided by cls, returns 0 otherwise.
  • handle returns the method handle for methodName.
  • origin returns the aliased command if methodName is an alias method, or an empty string otherwise.
  • parameters returns the parameter specification of methodName as a list of parameter names and type specifications.
  • registrationhandle returns the method handle for a submethod in a method ensemble from the perspective of the method caller. methodName must contain a complete method path.
  • returns gives the type specification defined for the return value of methodName.
  • submethods returns the names of all submethods of methodName, if methodName is a method ensemble. Otherwise, an empty string is returned.
  • syntax returns the method parameters of methodName as a concrete-syntax description to be used in human-understandable messages (e.g., errors or warnings, documentation strings).
  • type returns whether methodName is a scripted method, an alias method, a forwarder method, or a setter method.
Returns the names of all methods defined by cls. Methods covered include those defined using alias and forward. The returned methods can be limited to those whose names match namePattern (see string match).

By setting -callprotection, only methods of a certain call protection level (public, protected, or private) will be returned. Methods of a specific type can be requested using -type. The recognized values for methodType are:

  • scripted denotes methods defined using class method;
  • alias denotes alias methods defined using class alias;
  • forwarder denotes forwarder methods defined using class forward;
  • setter denotes methods defined using ::nsf::setter;
  • all returns methods of any type, without restrictions (also the default value);
If pattern is omitted, returns the object names of the mixin classes which extend cls directly. By turning on the switch -guards, the corresponding guard expressions, if any, are also reported along with each mixin as a three-element list: className -guard guardExpr. The returned mixin classes can be limited to those whose names match patterns (see string match).
If pattern is not specified, returns the object names of all slot objects defined by cls. The returned slot objects can be limited according to any or a combination of the following criteria: First, slot objects can be filtered based on their command names matching pattern (see string match). Second, -type allows one to select slot objects which are instantiated from a subclass className of nx::Slot (default: nx::Slot).
If pattern is omitted, returns the object names of all slot objects provided by cls which are responsible for managing properties and variables of cls. Otherwise, only slot objects whose names match pattern are returned.

This is equivalent to calling: cls info slots -type ::nx::VariableSlot pattern.

To extract details of each slot object, use the info submethods available for each slot object.

Defines a scripted method methodName for the scope of the class. The method becomes part of the class's signature interface. Besides a methodName, the method definition specifies the method parameters and a method body.

parameters accepts a Tcl list containing an arbitrary number of non-positional and positional parameter definitions. Each parameter definition comprises a parameter name, a parameter-specific value checker, and parameter options.

The body contains the method implementation as a script block. In this body script, the colon-prefix notation is available to denote an object variable and a self call. In addition, the context of the object receiving the method call (i.e., the message) can be accessed (e.g., using nx::self) and the call stack can be introspected (e.g., using nx::current).

Optionally, -returns allows for setting a value checker on values returned by the method implementation. By setting the switch -checkalways, value checking on arguments and return value is guaranteed to be performed, even if value checking is temporarily disabled; see nx::configure).

To express deprecation of the method name, set the -deprecated flag. Deprecated methods remain usable from client code, but their usage will be signaled to the developer and/or can be tracked using ::nsf::deprecated. To register name with the debugger, set the -debug flag. Entering and exiting a method, which was flagged for debugging, is recorded by calling the redefinable callback procs ::nsf::debug::call and ::nsf::debug::exit, respectively. By default, these callbacks forward to ::nsf::log, which can also be customized at the script level.

A method closely resembles a Tcl proc, but it differs in some important aspects: First, a method can define non-positional parameters and value checkers on arguments. Second, the script implementing the method body can contain object-specific notation and commands (see above). Third, method calls cannot be intercepted using Tcl trace. Note that an existing Tcl proc can be registered as an alias method with the class (see alias).

Accesses and modifies the list of mixin classes of cls using a specific setter or getter submethod:
Inserts a single mixin class into the current list of mixin classes of cls. Using index, a position in the existing list of mixin classes for inserting the new mixin class can be set. If omitted, index defaults to the list head (0).
If pattern is omitted, returns the object names of the mixin classes which extend cls directly. By specifying pattern, the returned mixin classes can be limited to those whose names match pattern (see string match).
Removes all mixin classes from cls and returns the list of removed mixin classes. Clearing is equivalent to passing an empty list for mixinSpecList to mixins set.
Removes a mixin class from a current list of mixin classes of cls whose spec matches specPattern. specPattern can contain special matching chars (see string match). class mixins delete will throw an error if there is no matching mixin class, unless -nocomplain is set.
Returns the list of current mixin specifications.
If expr is specified, a guard expression expr is registered with the mixin class className. This requires that the corresponding mixin class className has been previously set using class mixins set or added using mixins add. expr must be a valid Tcl expression (see expr). An empty string for expr will clear the currently registered guard expression for the mixin class className.

If expr is not specified, returns the active guard expression. If none is available, an empty string will be returned.

mixinSpecList represents a list of mixin class specs, with each spec being itself either a one-element or a three-element list: className ?-guard guardExpr?. If having one element, the element will be considered the className of the mixin class. If having three elements, the third element guardExpr will be stored as a guard expression of the mixin class. This guard expression will be evaluated using expr when cls receives a message to determine if the mixin is to be considered during method dispatch or not. Guard expressions allow for realizing context-dependent or conditional mixin composition.
At the time of setting the mixin relation, that is, calling mixins, every className as part of a spec must be an existing instance of nx::Class. To access and to manipulate the list of mixin classes of cls, cget|configure -mixins can also be used.
A factory method to create autonamed instances of cls. It returns the name of the newly created instance. For example:
% nx::Class create AClass; # defines a class 'AClass' being an instance of 'nx::Class'
::AClass
% set inst [::AClass new]; # defines an autonamed object being an instance of 'AClass'
::nsf::__#0
% $inst info class
::AClass
The factory method will provide computed object names of the form, e.g. ::nsf::__#0. The uniqueness of generated object names is guaranteed for the scope of the current Tcl interpreter only.

It is a frontend to create which will be called by new once the name of the instance has been computed, passing along the arguments option to new as the configuration options (see create).

If -childof is provided, the new object will be created as a nested object of parentName. parentName can be the name of either an existing NX object or an existing Tcl namespace. If non-existing, a Tcl namespace parentName will be created on the fly.

Defines a property for the scope of the class. The spec provides the property specification as a list holding at least one element or, maximum, two elements: propertyName?:typeSpec? ?defaultValue?. The propertyName is also used as to form the names of the getter/setter methods, if requested (see -accessor). It is, optionally, equipped with a typeSpec following a colon delimiter which specifies a value checker for the values which become assigned to the property. The second, optional element sets a defaultValue for this property.

If -accessor is set, a property will provide for different getter and setter methods:

Returns 1 if the value store of propertyName (e.g., an object variable) exists and has been given a value, returns 0 otherwise.
Sets the property propertyName to value.
Returns the current value of property propertyName.
Removes the value store of propertyName (e.g., an object variable), if existing.
The option value passed along -accessor sets the level of call protection for the generated getter and setter methods: public, protected, or private. By default, no getter and setter methods are created.

Turning on the switch -incremental provides a refined setter interface to the value managed by the property. First, setting -incremental implies requesting -accessor (set to public by default, if not specified explicitly). Second, the managed value will be considered a valid Tcl list. A multiplicity of 1..* is set by default, if not specified explicitly as part of spec. Third, to manage this list value element-wise (incrementally), two additional setter methods become available:

Adding element to the managed list value, at the list position given by index (by default: 0).
Removing the first occurrence of an element from the managed list value which matches elementPattern. elementPattern can contain matching characters (see string match). An error will be thrown if there is no match, unless -nocomplain is set.

By setting -configurable to true (the default), the property can be accessed and modified through cget and configure, respectively. If false, no configuration option will become available via cget and configure.

If neither -accessor nor -configurable are requested, the value managed by the property will have to be accessed and modified directly. If the property manages an object variable, its value will be readable and writable using set and eval.

The -trace option causes certain slot methods to be executed whenever get, set, or default operations are invoked on the property:

  • set: slot value=set obj propertyName value
  • get: slot value=get obj propertyName
  • default: slot value=default obj propertyName

A property becomes implemented by a slot object under any of the following conditions:

  • -configurable equals true (by default).
  • -accessor is one of public, protected, or private.
  • -incremental is turned on.
  • initBlock is a non-empty string.
Assuming default settings, every property is realized by a slot object.

Provided a slot object managing the property is to be created, a custom class className from which this slot object is to be instantiated can be set using -class. The default value is ::nx::VariableSlot.

The last argument initBlock accepts an optional Tcl script which is passed into the initialization procedure (see configure) of the property's slot object. See also initBlock for create and new.

Attempts to register a method definition made available using ::nsf::method::provide under the name methodName with cls . The registered method is subjected to default call protection (protected), if not set explicitly.
Defines a variable for the scope of the class. The spec provides the variable specification: variableName?:typeSpec?. The variableName will be used to name the underlying Tcl variable and the getter/setter methods, if requested (see -accessor). spec is optionally equipped with a typeSpec following a colon delimiter which specifies a value checker for the values managed by the variable. Optionally, a defaultValue can be defined.

If -accessor is set explicitly, a variable will provide for getter and setter methods:

Returns 1 if the value store of variableName (e.g., an object variable) exists and has been given a value, returns 0 otherwise.
Sets variableName to varValue.
Returns the current value of variableName.
Removes variableName, if existing, underlying the property.
The option value passed along -accessor sets the level of call protection for the getter and setter methods: public, protected, or private. By default, no getter and setter methods are created.

Turning on the switch -incremental provides a refined setter interface to the value managed by the variable. First, setting -incremental implies requesting -accessor (public by default, if not specified explicitly). Second, the managed value will be considered a valid Tcl list. A multiplicity of 1..* is set by default, if not specified explicitly as part of spec (see above). Third, to manage this list value element-wise (incrementally), two additional setter operations become available:

Adding element to the managed list value, at the list position given by index (by default: 0).
Removing the first occurrence of an element from the managed list value which matches elementPattern. elementPattern can contain matching characters (see string match). An error will be thrown if there is no match, unless -nocomplain is set.

By setting -configurable to true, the variable can be accessed and modified via cget and configure, respectively. If false (the default), the interface based on cget and configure will not become available. In this case, and provided that -accessor is set, the variable can be accessed and modified via the getter/setter methods. Alternatively, the underlying Tcl variable, which is represented by the variable, can always be accessed and modified directly, e.g., using eval. By default, -configurable is false.

The -trace option causes certain slot methods to be executed whenever get, set, or default operations are invoked on the variable:

  • set: slot value=set obj variableName value
  • get: slot value=get obj variableName
  • default: slot value=default obj variableName

A variable becomes implemented by a slot object under any of the following conditions:

  • -configurable equals true.
  • -accessor is one of public, protected, or private.
  • -incremental is turned on.
  • -initblock is a non-empty string.
Provided a slot object managing the variable is to be created, a custom class className from which this slot object is to be instantiated can be set using -class. The default value is ::nx::VariableSlot.

Using -initblock, an optional Tcl script can be defined which becomes passed into the initialization procedure (see configure) of the variable's slot object. See also initBlock for create and new.

OBJECT LIFE CYCLE

nx::Class provides means to control important stages through which an NX object passes between and including its creation and its destruction: allocation, recreation, deallocation.


/cls/->create(/instance/)
.---------------. exists? [false] .----------------. .-------------------.
---->|Class::create()|----><>---------------->|Class::__alloc()|-----------><>---->|Object::configure()|
`---------------' | (1) `----------------' ^ (3) `---------+---------'
[true] | | | (4)
| .-------------------. | .------------------.
`->|Class::__recreate()|-------------------------' |/instance/->init()|
(2) `-------------------' `------------------'
/instance/->destroy()
.-----------------. .------------------.
---->|Object::destroy()|---->|Class::__dealloc()|
`-----------------' (5) `------------------'
Object creation is controlled by the factory method create, provided by nx::Class to its instance cls. create produces a new object instance as an instance of cls in a number of steps.
[1]
If instance does not represent an existing object, an internal call to __alloc, provided by nx::Class, runs the allocation procedure for a fresh instance of cls.
[2]
If instance corresponds to an existing object, the recreation procedure is triggered by calling __recreate defined by nx::Class.
[3]
The newly allocated or recreated object instance is then configured by dispatching configure, provided by nx::Object, which consumes the configuration options passed into create. This will establish the instance's initial state, e.g. by setting object variables and object relations according to the configuration options and corresponding default values.
[4]
Finally, the initialization method init is dispatched, if available for instance. init can be defined by cls on behalf of its instance instance, e.g. to lay out a class-specific initialization behavior.
% nx::Class create Foo {:property x}
% Foo method init {} {set :y [expr {${:x} + 1}]}
% Foo public method bar {} {return ${:y}}
% Foo create f1 -x 101
% f1 cget -x
101
% f1 bar
102
Alternatively, the object instance may define a per-object init on its own. A per-object init can be chained to a class-level init using nx::next, just like a regular method.

Note that the definition of an init method must contain an empty parameter specification, since init is always called with an empty argument list.

Object destruction, such as triggered by an application-level destroy call (5), is finalized by __dealloc offerd by nx::Class.

In the following, the three built-in procedures --- allocation, recreation, and deallocation --- are explained:

  • Allocation: __alloc creates a blank object instance as an instance of cls and returns the fully-qualified instance. __alloc is primarily used internally by create to allocate a Tcl memory storage for instance and to register instance with the Tcl interpreter as a new command.
  • Recreation: Recreation is the NX scheme for resolving naming conflicts between objects: An object is requested to be created using create or new while an object of an identical object name, e.g. instance, already exists:
% Object create Bar
::Bar
% Object create Bar; # calls Object->__recreate(::Bar, ...)
::Bar
In such a situation, the built-in __recreate first unsets the object state (i.e., Tcl variables held by the object) and removes relations of the object under recreation with other objects. Then, second, standard object initialization is performed by calling configure and init, if any.

Alternatively, recreation will be performed as a sequence of destroy and create calls in the following recreation scenarios:

  • An existing class is requested to be recreated as an object.
  • An existing object is requested to be recreated as a class.

% Object create Bar
::Bar
% Class create Bar; # calls Bar->destroy() & Class::create(::Bar, ...)
An object of an object system other than NX (e.g. XOTcl2) is asked to be recreated.
Deallocation: __dealloc marks an instance instance of cls for deletion by returning its Tcl memory representation to the Tcl memory pool and by unregistering the corresponding Tcl command with the Tcl interpreter.

Beware that __dealloc does not necessarily cause the object to be deleted immediately. Depending on the lifecycle of the object's environment (e.g. the Tcl interp interpreter, the containing namespace) and on call references down the callstack, the actual memory freeing/returning operation may occur at a later point.

The three methods __alloc, __recreate, and __dealloc are internally provided and internally called. By default, they are not part of the method interface of cls and cannot be called directly by clients of cls. In addition, __alloc, __recreate, and __dealloc are protected from redefinition by a script.

To extend or to replace the built-in allocation, recreation, and deallocation procedure, the methods __alloc, __recreate, and __dealloc can be refined by providing a custom method implementation:

  • as a per-object method of cls;
  • as a method of a per-object mixin class extending cls;
  • as a method of a per-class mixin class extending nx::Class;
  • as a method of a subclass specializing nx::Class, from which cls is to be instantiated.

This custom implementation can redirect to the built-in __alloc, __recreate, and __dealloc, respectively, by using nx::next. By providing such a custom implementation, __alloc, __recreate, and __dealloc, respectively, become available as callable methods of cls:

COPYRIGHT

Copyright (c) 2014-2019 Stefan Sobernig <stefan.sobernig@wu.ac.at>, Gustaf Neumann <gustaf.neumann@wu.ac.at>; available under the Creative Commons Attribution 3.0 Austria license (CC BY 3.0 AT).
2.3.0 Class