|
|
| |
nx::Class(3) |
NX API |
nx::Class(3) |
nx::Class - API reference of the base-metaclass of the NX objectsystem
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?
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:
- nx::Class create cls ?-superclasses
superClassNames? ?-mixins mixinSpec? ?-filters
filterSpec? ?option value ...? ?initBlock?
- To create a class having the explicit name cls, use
create.
- nx::Class new ?-superclasses superClassNames?
?-mixins mixinSpec? ?-filters filterSpec?
?initBlock?
- 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 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.
- -superclasses ?superClassNames?
- If superClassNames is not specified, returns the superclasses of
the class. If provided, the class becomes the subclass of
superClassNames.
- -filters ?filterSpecs?
- 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.
- -mixins ?mixinSpecs?
- 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.
- alias
- cls ?public | private | protected?
alias ?-debug? ?-deprecated? methodName
?-returns valueChecker? ?-frame object |
method? cmdName
- 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
- cls __class_configureparameter
- Computes and returns the configuration options available for cls
instances, to be consumed as method-parameter specification by
configure.
- create
- cls create instanceName ?option value
option value ...?
- 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.
- delete
- cls delete feature arg
- This method serves as the equivalent to Tcl's rename for removing
structural (properties, variables) and behavioral features (methods) of
the class:
- cls delete property propertyName
- cls delete variable variableName
- cls delete method methodName
- 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).
- filters
- cls filters submethod ?arg ...?
- Accesses and modifies the list of methods which are registered as filters
with cls using a specific setter or getter submethod:
- cls filters add spec ?index?
- 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).
- cls filters clear
- 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.
- cls filters delete ?-nocomplain?
specPattern
- 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.
- cls filters get
- Returns the list of current filter specifications registered for
cls.
- cls filters guard methodName
?expr?
- 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.
- cls filters methods ?pattern?
- 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).
- cls filters set filterSpecList
- 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.
- forward
- cls ?public | protected | private?
forward ?-debug? ?-deprecated? methodName
?-prefix prefixName? ?-frame object?
?-returns valueChecker? ?-verbose? ?target?
?arg ...?
- 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.
- info
- 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.
- cls info heritage ?pattern?
- 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).
- cls info instances ?-closure? ?pattern?
- 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).
- cls info mixinof ?-closure? ?-scope
option? ?pattern?
- 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.
- cls info subclasses ?-closure? ?-dependent?
?pattern?
- 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).
- cls info superclasses ?-closure?
?pattern?
- 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).
- cls info info ?-asList?
- 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.
- cls info filters ?-guards? ?pattern?
- 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).
- cls info method option methodName
- 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.
- cls info methods ?-callprotection level?
?-type methodType? ?-path? ?namePattern?
- 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);
- cls info mixins ?-guards? ?pattern?
- 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).
- cls info slots ?-type className?
?pattern?
- 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).
- cls info variables ?pattern?
- 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.
- method
- cls ?public | protected | private?
method ?-debug? ?-deprecated? name
parameters ?-checkalways? ?-returns
valueChecker? body
- 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).
- mixins
- cls mixins submethod ?arg ...?
- Accesses and modifies the list of mixin classes of cls using a
specific setter or getter submethod:
- cls mixins add spec ?index?
- 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).
- cls mixins classes ?pattern?
- 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).
- cls mixins clear
- 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.
- cls mixins delete ?-nocomplain?
specPattern
- 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.
- cls mixins get
- Returns the list of current mixin specifications.
- cls mixins guard className ?expr?
- 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.
- cls mixins set mixinSpecList
- 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.
- new
- cls new ?-childof parentName? ?option
value option value ...?
- 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.
- property
- cls property ?-accessor public |
protected | private? ?-class className?
?-configurable trueFalse? ?-incremental? ?-trace
set | get | default? spec
?initBlock?
- 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:
- obj propertyName exists
- Returns 1 if the value store of propertyName (e.g., an object
variable) exists and has been given a value, returns 0 otherwise.
- obj propertyName set value
- Sets the property propertyName to value.
- obj propertyName get
- Returns the current value of property propertyName.
- obj propertyName unset
- 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:
- obj propertyName add element
?index?
- Adding element to the managed list value, at the list position
given by index (by default: 0).
- obj propertyName delete ?-nocomplain?
elementPattern
- 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.
- require
- cls require ?public | protected |
private? method methodName
- 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.
- variable
- cls variable ?-accessor public |
protected | private? ?-incremental? ?-class
className? ?-configurable trueFalse? ?-initblock
?-trace set | get | default? script?
spec ?defaultValue?
- 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:
- obj variableName exists
- Returns 1 if the value store of variableName (e.g., an object
variable) exists and has been given a value, returns 0 otherwise.
- obj variableName set varValue
- Sets variableName to varValue.
- obj variableName get
- Returns the current value of variableName.
- obj variableName unset
- 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:
- obj variableName add element
?index?
- Adding element to the managed list value, at the list position
given by index (by default: 0).
- obj variableName delete ?-nocomplain?
elementPattern
- 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.
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:
- cls __alloc instance
- cls __recreate instance ?arg ...?
- cls __dealloc instance
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).
Visit the GSP FreeBSD Man Page Interface. Output converted with ManDoc. |