CatsooLang documentation
CatsooLang is a little Domain Specific Langage (DSL) that allows you to handle models in a textual programming way into http://www.genmymodel.com. You can handle very low level feature of your model. CatsooLang interpreter is evaluated on the client side only, there is no intensive communication with server (there is some, we will come back on this later). Here is some details about the language:
Each modification you perform on your model is propagated to other users and are stored in the cloud. In that case, your shell will communicate with the server to send the modifications. This mode can be disable, but in that case, each modification will not be stored.
The "CatsooLang Shell" is an interactive mode for "CatsooLang", this is not a UNIX-shell like but more a lua/python/ocaml (you name it) interactive console like that evaluates CatsooLang expressions.
Before jumping into code lines, here is some useful commands:
help
--> displays help with the shortcuts you can use in the interactive shellhistory
--> displays all the commands you typed in the shell from the beginning of your session#sandbox-on
--> goes into safe mode (each model modification is performed locally only and are not saved)#sandbox-off
--> goes back from safe mode to real-time mode (each model modification is saved)Also, there is a special keyword you can call to get information about a function: doc()
.
e.g.,
cl> doc(objects);
1. objects : (obj:Object) -> Collection
Returns all the objects contained in the object, including itself.
@param obj the container object.
@returns A collection of objects.
The result is the function definition the interpreter found. It gives the objects
function signature: this function takes a model element object
(an Object
) and returns a collection.
If there is no documentation (because not yet here) here is the output:
cl> doc(pow);
1. pow : (a:Real x b:Real) -> Real
* Sorry, there is no documentation or doc is empty
If the function does not exist:
cl> doc(foo);
Sorry, there is no 'foo' operation
Finally, if there is many signature for the same function (function overloading), the shell returns all the signature and documentation the interpreter find for this function.
cl> doc(substring);
1. substring : (s:String x begin:Integer x end:Integer) -> String
* Sorry, there is no documentation or doc is empty
2. substring : (s:String x begin:Integer) -> String
* Sorry, there is no documentation or doc is empty
The interactive shell displays the function in the order they are register into the CatsooLang evaluation environement (we will see later what this implies).
The way CatsooLang display messages is using log
keyword (this is not a function per say). In the interaction console, you can use CTRL + i
shortcut to
quickly write log();
and put the cursor between the ()
.
cl> log('foo'); // <-- strings use ' and not "
foo
cl> log(3);
3
cl> log([]);
[]
cl> log([3, 'R']);
[Integer: 3, String: R]
cl> var i := 33.3;
cl> log(i);
33.3
cl> log(__ROOT);
org.eclipse.uml2.uml.internal.impl.ModelImpl@6762 (name: model, visibility: <unset>) (URI: null) (viewpoint: <unset>)
cl> log(foo); // <-- a non existing variable
null
NOTE
You can question variable and expression about their type using the type()
function:
cl> type(3);
'Integer: 3
cl> type('R');
'String: R
cl> type(__ROOT);
'CL_EOBJECT(Model, model, 3798)
cl> type([3, __ROOT]);
'Collection: [Integer: 3, CL_EOBJECT(Model, model, 3886)]
As you can see, there is an already defined variable in CatsooLang evaluation environement called __ROOT
that points to the root of your model.
We can use this variable to display a tree of our model. Here is an example on a simple model owning one class with three attributes:
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 31088) [packageImport]
| +- (PackageImport, 31089) [packageImport]
| +- Class 'UserEntity' [packagedElement]
| | +- Property 'ID' [ownedAttribute]
| | +- Property 'attribute' [ownedAttribute]
| | +- Property 'attribute2' [ownedAttribute]
This function give us a tree representation using a model element as starting point. Each line is built using the following format:
object_type 'object_name' [feature_to_this_element_from_parent]
. So, if we look at the 4th line, we can see
that there is a Class
named UserEntity
contained by the packagedElement
feature of the Model
element named model
.
How can we access the UserEntity
element using CatsooLang? CatsooLang use model element name to navigate through them from the root of your model (a kind of path).
Each name must be separated by ::
, so in our example, we can access to class UserEntity
by typing model::UserEntity
. The beginning of a path can also be a variable
so, the expression __ROOT::UserEntity
gives the same result.
cl> log(model);
org.eclipse.uml2.uml.internal.impl.ModelImpl@6762 (name: model, visibility: <unset>) (URI: null) (viewpoint: <unset>)
cl> log(model::UserEntity);
org.eclipse.uml2.uml.internal.impl.ClassImpl@7a1c (name: UserEntity, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
cl> log(__ROOT::UserEntity);
org.eclipse.uml2.uml.internal.impl.ClassImpl@7a1c (name: UserEntity, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
cl> var r := model;
cl> log(r::UserEntity);
org.eclipse.uml2.uml.internal.impl.ClassImpl@7a1c (name: UserEntity, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
This way of accessing model element is just a quick access. You can 'manually' navigate to the desired object using model features. Here is the way we can access the same element using features:
cl> log(model.packagedElement);
[CL_EOBJECT(Class, UserEntity, 32492)] // <-- the result is a collection of model object with only one element
cl> log(model.packagedElement at(0)); // <-- we access the first element (collection index begin at 0)
org.eclipse.uml2.uml.internal.impl.ClassImpl@7a1c (name: UserEntity, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
The function at
allows you to get an element in a collection at a specific position (for a collection of n element, from 0 to n-1). Let's take a look to the function
signature:
cl> doc(at);
1. at : (collection:Collection x index:Integer) -> Any
* Sorry, there is no documentation or doc is empty // <-- I know, shame on me
The signature tells us that this function takes a collection and an integer as parameter and return something (we will go back on types and typing system later). The important
thing here is the way the function call. There is two ways of calling a function in CatsooLang, either you give all parameters to the function in a foo(bar1, bar2);
way
or you can use the 'sourced' call: bar1 foor(bar2);
. The two calls are equivalent. So, our previous example could be rewritten at(model.packagedElement, 0);
. The way you call
the function is totally up to you.
cl> log(at(model.packagedElement, 0));
org.eclipse.uml2.uml.internal.impl.ClassImpl@7a1c (name: UserEntity, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
cl> log(cos(1.0));
0.5403023058681398
cl> log(1.0 cos());
0.5403023058681398
cl> log(size([1, 2, 3]));
3
cl> log([1, 2, 3] size());
3
CatsooLang is dynamically typed. It means, for example, that each element feature is accessed at evaluation time. If the feature does not exist, an error message is displayed but
the execution continue (no exception system for errors). If an error is displayed, the return result is null
.
cl> log(__ROOT.name);
model
cl> log(__ROOT.foo);
Feature foo doest not exist for org.eclipse.emf.ecore.impl.EClassImpl@18b (name: Model) (instanceClassName: null) (abstract: false, interface: false)
null
cl> log(foo);
null
cl> log(foo := 5);
Variable foo does not exists.
null
You can define a variable using the var
keyword, (e.g., var i := 'R';
). A variable is typed at execution time and its type can change during its 'life'.
Each time, the variable type is deduced from the affectation expression execution (affectation operator is :=
).
cl> var i := 'R';
cl> type('R');
'String: R
cl> i := 0;
cl> type(i);
'Integer: 0
cl> i := i toString();
cl> type(i);
'String: 0
If you want to get a listing of all the defined variable in the current evaluation context, the _getvars()
function returns a collection
of the defined variables (we will see later how collections works):
cl> var i := 'R'; var a := true; var z := [];
cl> log(_getvars());
[__ROOT -> CL_EOBJECT(Model, model, 10702), i -> String: R, a -> Boolean: true, z -> Collection: []]
cl> log(_getvars() at(1) + 'est');
Rest
CatsooLang types are either basic types or types that depend on loaded metamodels. There is 8 basic types, here is their 'hierarchie':
+-----+
| Any |-------------------------------+
+-----+ (contains) |
^ |
| |
+------------+-----------+-----------+---------------+ |
| | | | | |
+----+----+ +----+---+ +----+---+ +----+---+ +--------+---+ |
| Boolean | | String | | Number | | Object | | Collection |#-+
+---------+ +--------+ +--------+ +--------+ +------------+
^
|
+----+-----+
| |
+---+--+ +----+----+
| Real | | Integer |
+------+ +---------+
Any
represents all the CatsooLang types,Boolean
represents the bool type (constantes are true
and false
),String
reprensents string type,Number
is an abstract type that reprents either an int or a real,Integer
represents integer,Real
represents real,Collection
represents collection of elements (can contain Any
type of element),Object
represents the higher level object for model elements (all model elements inherits from this type).Every time a metamodel is loaded into the CatsooLang evaluation environment, each meta-object and meta-object instance will inherit from Object
.
Actually, Object
represents EObject
from Eclipse Modeling Framework
(in the example below, the output format gives the metaclass name, the
instance name (if there is any) and the instance id).
cl> type(uml::Class); // <-- in UML context
'CL_EOBJECT(EClass, Class, 331)
cl> type(UserEntity); // <-- notice that the root is not needed in this path as the object UserEntity is directly contained by the root.
'CL_EOBJECT(Class, UserEntity, 3711)
As we previously seen, you can directly access to your model elements using paths or you can use meta-feature to navigate from an element to another. You can use this way of accessing elements to query your model or get elements from a specific context. Here is some examples:
cl> log(UserEntity.ownedAttribute); // <-- Gets all the attributes of the UserEntity class
[CL_EOBJECT(Property, ID, 3596), CL_EOBJECT(Property, attribute, 3597), CL_EOBJECT(Property, attribute2, 3598)]
cl> log(UserEntity.ownedAttribute.name); // <-- Gets all the attribute names of the UserEntity class
[String: ID, String: attribute, String: attribute2]
cl> log(UserEntity.ownedAttribute.type.name); // <-- Gets all the type names of the UserEntity class attributes
[String: Integer, String: Long, String: Real]
As you can see, when accessing features that return collection (as ownedAttribute
), accessing another feature iterates on each elements and produce a new
collection with the feature value (only if the collection element is an Object). In this case, it is actually syntactic sugar for a slightly more complex
expression (we will discuss them later). If the feature is a single value, you get the result as shown above, but if the feature is a collection, you will
get a collection of collections.
cl> log(__ROOT.packagedElement);
[CL_EOBJECT(Class, UserEntity, 5268)] <-- this is a collection with a single element: UserEntity
cl> log(__ROOT.packagedElement.ownedAttribute);
[Collection: [CL_EOBJECT(Property, ID, 5357), CL_EOBJECT(Property, attribute, 5358), CL_EOBJECT(Property, attribute2, 5359)]] <-- a collection of collection
cl> log(__ROOT.packagedElement.ownedAttribute.ownedElement);
Obj Collection: [CL_EOBJECT(Property, ID, 5445), CL_EOBJECT(Property, attribute, 5446), CL_EOBJECT(Property, attribute2, 5447)] is not an EObject!
[]
cl> log(__ROOT.packagedElement.ownedAttribute flatten().ownedElement);
[Collection: [], Collection: [], Collection: []]
If you are lost in meta-features, CatsooLang defines a function that helps you finding what meta-feature you can use for a specific meta-class:
cl> introspect(uml::Class); // <-- function name is not great...
-= Super types =-
[EncapsulatedClassifier, BehavioredClassifier]
-= Attributes =-
eAnnotations : EAnnotation [0..*]
ownedComment : Comment [0..*]
ownedElement : Element [0..*] - READONLY
owner : Element [0..1] - READONLY
clientDependency : Dependency [0..*]
name : String [0..1]
nameExpression : StringExpression [0..1]
namespace : Namespace [0..1] - READONLY
qualifiedName : String [0..1] - READONLY
visibility : VisibilityKind [0..1]
... <-- there is many other fields
This function displays first the super types of the meta-class (here EncapsulatedClassifier
and BehavioralClassifier
), then all the
meta-attribute (meta-feature) you can use on instance of this meta-class. For each meta-attribute, the feature name is displayed as well
as the feature type and its cardinality. Features annotated with READONLY
are feature you can only read and not set.
TODO
operators =, <>, <=, <, >, >=
cl> log(3 < 3);
false
cl> log(true < false);
false
cl> log(false < true);
true
cl> log('A' < 'B');
true
cl> log(model = model);
true
cl> log(model = model::UserEntity);
false
cl> log(model < model);
null
cl> log(3 < 'R');
Error> cast error between Integer: 3 and String: R
null
In CatsooLang, collections are (almost) immutables, you cannot directly modify them. Operation you perform on them return a new result but does not modify the original collection. Collection can contain any type of elements.
cl> var col := [1, 2, 3, 'A'];
cl> log(col remove('A')); // <-- could be also: log(remove(col, 'A'));
[Integer: 1, Integer: 2, Integer: 3]
cl> log(col);
[Integer: 1, Integer: 2, Integer: 3, String: A]
cl> log(col insertAt('B', 0));
[String: B, Integer: 1, Integer: 2, Integer: 3, String: A]
cl> log(col);
[Integer: 1, Integer: 2, Integer: 3, String: A]
cl> col := col remove('A') insertAt(1, col size() - 1); // <-- function call chaining
cl> log(col);
[Integer: 1, Integer: 2, Integer: 3, Integer: 1]
cl> log(col uniq());
[Integer: 1, Integer: 2, Integer: 3]
cl> log(col);
[Integer: 1, Integer: 2, Integer: 3, Integer: 1]
Collections are almost immutable structure, you can add elements (1 or many) to a collection using the +=
operator. This operator can only be applied on named collections (collection stored in a variable or 'many' feature).
cl> var a := [];
cl> a += 1;
cl> log(a);
[Integer: 1]
cl> a += [2, 1, 3];
cl> log(a);
[Integer: 1, Integer: 2, Integer: 1, Integer: 3]
cl> a := [];
cl> a += [1, [2,3]];
cl> log(a);
[Integer: 1, Collection: [Integer: 2, Integer: 3]]
You can put expression as collection member (between []
), the expressions will be evaluated to get the collection member.
cl> log([3+1, true or false, UserEntity.name, UserEntity]);
[Integer: 4, Boolean: true, String: UserEntity, CL_EOBJECT(Class, UserEntity, 6912)]
You can compare collections using =
or <>
. These operators will test if two collections have same size and if the content
of the first collection is/isn't in the second one. However, these operators does not test content order, only values.
cl> log([] = []);
true
cl> var i := []; log(i = []);
true
cl> log([1, 2, ['A']] = [2, 1, ['A']]);
true
cl> log([1, 2, [4]] = [2, 1, [5]]);
false
cl> log([1, 2, [4]] <> [2, 1, [5]]);
true
CatsooLang allows you to easily iterate on each collection elements using the ->
operator. This operator applies on each collection element
a dedicated function. In the expression ['FOO', 'Bar', 'fooBAr']->toLower();
, the function toLower
will be applied on each element of the
['FOO', 'Bar', 'fooBAr']
collection and will produce a new collection as result: ['foo', 'bar', 'foobar']
. Iteration operator requires that
the function are called in a "sourced" way (foo bar()
).
cl> log(['FOO', 'Bar', 'fooBAr']->toLower());
[String: foo, String: bar, String: foobar]
cl> log([1, 2, 3]->toReal()->cos());
[Real: 0.5403023058681398, Real: -0.4161468365471424, Real: -0.9899924966004454]
In the example below, toLower
signature is toLower : (s:String) -> String
, it takes a String
as parameter and returns a String
, so it can
be applied to the whole collection whithout errors. If an element in the collection cannot match the signature, an error is displayed, but the
execution continue and the result is given whithout the error.
cl> log(4 toLower());
Operation 'toLower' not found in this context or for parameters [Integer: 4]
null
cl> log([4, 'FOO', 'Bar', 'fooBAr']->toLower());
Operation 'toLower' not found in this context or for parameters [Integer: 4]
[String: foo, String: bar, String: foobar]
As you can see in the trace, 4 toLower();
execution displays an error, but also returns null
. In the second expression, even if an error message
had been displayed, the final result does not contains null
. During collection construction, null
elements are simply ignored.
To avoid many function definition to use iteration, CatsooLang defines two special functions _iter
and _apply
that can be used to easily apply
an expression to each element of a collection (named self
). Difference between the two operators is quite subtle, you can try doc()
on these
operators to have details about them.
_iter()
This operator takes any type of element and finally returns them. This operator is used to handle mutable element (collections in some cases and model element), modify them and return them. We will see later how we can modify model elements and how this operator becomes handy.
cl> var a := [1, 2];
cl> log(a->_iter(self + 1));
[Integer: 1, Integer: 2]
cl> a:= [[1], [2]]; a->_iter(self += 1); log(a);
[Collection: [Integer: 1, Integer: 1], Collection: [Integer: 2, Integer: 1]]
_apply()
This operator takes any type of element and finally returns the result of the expression passed as parameter. This operator is used to produce new collections without altering the initial one.
cl> log([1, 2, 4, 5]->_apply(self + 1));
[Integer: 2, Integer: 3, Integer: 5, Integer: 6]
cl> log(UserEntity.ownedAttribute->_apply(self.type.name)); // <-- this expression is equivalent to log(UserEntity.ownedAttribute.type.name); we've seen before
[String: Integer, String: Long, String: Real]
Filtering collection is always a reccurent task. In order to ease this process, CatsooLang defines two operators: select
and reject
(as in OCL/QVT for example).
These operators are used to select/reject elements that satisfy an expression from a collection.
Syntax for select
is: _collection_->select(_id_ | _expr_);
. The same syntax is applied to reject
. The id identifier represents an element from collection on
which the expr is evaluated. If the expr returns true
, then the element is added to the resulting collection. Else, the element is ignored and expr is evaluated
on the next collection element.
cl> log([1, 2, 43, 5, null]->select(e | e > 3));
[Integer: 43, Integer: 5]
cl> log([4, 10, 4, 3, 45]->toReal()->pow(2.)->reject(e | e > 15));
[Real: 9]
In order to filter collection on elements type, use the isKindOf()
and isTypeOf()
functions. _input_ isKindOf(_type_)
tests if the input element has _type_
type or
is a subclass of _type_
whereas _input_ isTypeOf(_type_)
tests if the input type is exactly _type_
. Here is some examples using select
:
cl> log([1, 2., 'R', 'A', UserEntity]->select(e | e isKindOf(String)));
[String: R, String: A]
cl> log([1, 2., 'R', 'A', UserEntity]->select(e | e isKindOf(Number)));
[Integer: 1, Real: 2]
cl> log([1, 2., 'R', 'A', UserEntity]->select(e | e isKindOf(Classifier))); // <-- In UML, Class inherits from Classifier
[CL_EOBJECT(Class, UserEntity, 5052)]
cl> log([1, 2., 'R', 'A', UserEntity]->select(e | e isTypeOf(Number)));
[]
cl> log([1, 2., 'R', 'A', UserEntity]->select(e | e isTypeOf(Real)));
[Real: 2]
cl> log([1, 2., 'R', 'A', UserEntity]->select(e | e isTypeOf(Class)));
[CL_EOBJECT(Class, UserEntity, 5230)]
Note on isKindOf/isTypeOf
functions:
As function parameter, you can use a variable. The function will thus test if the variable type is the same as the input element.
cl> var a := 0; var b := 20; log(a isKindOf(b));
true
cl> var a := 0; var b := 'R'; log(a isKindOf(b));
false
cl> var a := UserEntity::ID; var b := UserEntity::attribute; log(a isKindOf(b));
true
To ease collection filtering on types from a metamodel (metaclass filtering), CatsooLang defines an handy operator: [_metaclass_]
(stolen from QVT) which is syntaxic sugar
for->select(x | x isKindOf(_metaclass_)
. This operator returns a collection containing the element of a specific type or []
if no collection element has the required type.
Associated to this operator, the ![_metaclass_]
selects only one element of the collection which has the required type. If no element has the required type, null
is returned.
Here is two examples:
cl> var a:= [1, 2., 'R', 'A', UserEntity];
cl> log(a[Classifier]);
[CL_EOBJECT(Class, UserEntity, 6031)]
cl> log(a[Number]);
line 2:6 mismatched input 'Number' expecting IDENTIFIER <-- a metaclass identifier is required, this operator does not work with CatsooLang basic types.
log(a[Number]);
^^^^^^
cl> log(a![Class]);
org.eclipse.uml2.uml.internal.impl.ClassImpl@102a (name: UserEntity, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
to be continued
In catsoolang, you can easily create new model elements using the create _metaclass_
keyword.
cl> var a := create Class;
cl> log(a);
org.eclipse.uml2.uml.internal.impl.ClassImpl@1e15 (name: <unset>, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
cl> type(a);
'CL_EOBJECT(Class, 7701)
cl> #load_metamodel ecore('http://www.eclipse.org/emf/2002/Ecore') // <-- we load the ecore metamodel under the 'ecore' name
!! Loading metamodel 'http://www.eclipse.org/emf/2002/Ecore' under 'ecore'
cl> log(ecore::Eclass); // <-- We can access ecore metamodel metaclasses
org.eclipse.emf.ecore.impl.EClassImpl@132 (name: EClass) (instanceClassName: null) (abstract: false, interface: false)
cl> log(create EClass); // <-- Consequently, we can now create instances of ecore metaclasses
org.eclipse.emf.ecore.impl.EClassImpl@2217 (name: null) (instanceClassName: null) (abstract: false, interface: false)
The two operators :=
and +=
are used for single value meta-attributes and multi value meta-attributes assignments:
cl> var a := create Class;
cl> log(a.name);
null
cl> a.name := 'MyFirstClass';
cl> log(a);
org.eclipse.uml2.uml.internal.impl.ClassImpl@227a (name: MyFirstClass, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
cl> log(a.name);
MyFirstClass
cl> log(a.ownedAttribute);
[]
cl> a.ownedAttribute += create Property;
cl> log(a.ownedAttribute);
[CL_EOBJECT(Property, 9277)]
cl> a.ownedAttribute at(0).name := 'myproperty';
cl> log(a.ownedAttribute.name);
[String: myproperty]
cl> tree(a);
+- Class 'MyFirstClass' [-]
| +- Property 'myproperty' [ownedAttribute]
In the above example, we created a model fragment (which is stored into a
). We create a Class
, name it MyFirstClass
then we create a property, we add it to MyClass
and we
name it myproperty
. This require a lot of steps. CatsooLang defines the convenient operator{...}
that allows you to assign many meta-attribute value at the same time. Its syntax
is ``myexprtoaneobj{metaatt1 op expr; metaatt2 op expr; ....}`. Here is some examples that exploit this:
cl> var a := create Class{name := 'C1'};
cl> log(a.name);
C1
cl> a{isAbstract := true; ownedAttribute += create Property{name := 'newprop'}};
cl> tree(a);
+- Class 'C1' [-]
| +- Property 'newprop' [ownedAttribute]
cl> var b := create Class{name := 'C1'; isAbstract := true; ownedAttribute += create Property{name := 'newprop'; type := umllib::String}}; // <-- Almost the same thing, but in one line
cl> tree(b);
+- Class 'C1' [-]
| +- Property 'newprop' [ownedAttribute]
NOTE
The create
keyword is quite permissive and allows you to use variable as _metaclass_
as soon as they reference a metaclass:
cl> var myc := Class;
cl> log(myc);
org.eclipse.emf.ecore.impl.EClassImpl@14b (name: Class) (instanceClassName: null) (abstract: false, interface: false)
cl> log(create myc);
org.eclipse.uml2.uml.internal.impl.ClassImpl@df5 (name: <unset>, visibility: <unset>) (isLeaf: false, isAbstract: false, isFinalSpecialization: false) (isActive: false)
cl> log(create myc{name := 'NewC'}.name);
NewC
Thanks to this feature, you can easily create new model fragments. However, up to now, these fragments are only present in CatsooLang evaluation context and they are not attached to your
model. In order to attach elements to your model, you simply have to alter the meta-attribute of your model element which must own the created fragment. Let's attach our fragment stored
in the b
variable in our model (once the fragment is attached to your model, you can see elements in the 'project' tree view of GenMyModel).
cl> tree(__ROOT); // <-- before we include the fragment
+- Model 'model' [-]
| +- (PackageImport, 10973) [packageImport]
| +- (PackageImport, 10974) [packageImport]
| +- Class 'UserEntity' [packagedElement]
| | +- Property 'ID' [ownedAttribute]
| | +- Property 'attribute' [ownedAttribute]
| | +- Property 'attribute2' [ownedAttribute]
cl> __ROOT.packagedElement += b;
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 11266) [packageImport]
| +- (PackageImport, 11267) [packageImport]
| +- Class 'UserEntity' [packagedElement]
| | +- Property 'ID' [ownedAttribute]
| | +- Property 'attribute' [ownedAttribute]
| | +- Property 'attribute2' [ownedAttribute]
| +- Class 'C1' [packagedElement]
| | +- Property 'newprop' [ownedAttribute]
Now, let's add the a
fragment into a Packaged
named 'mypack':
cl> __ROOT.packagedElement += create Package{name := 'mypack'; packagedElement += a};
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 11753) [packageImport]
| +- (PackageImport, 11754) [packageImport]
| +- Class 'UserEntity' [packagedElement]
| | +- Property 'ID' [ownedAttribute]
| | +- Property 'attribute' [ownedAttribute]
| | +- Property 'attribute2' [ownedAttribute]
| +- Class 'C1' [packagedElement]
| | +- Property 'newprop' [ownedAttribute]
| +- Package 'mypack' [packagedElement]
| | +- Class 'C1' [packagedElement]
| | | +- Property 'newprop' [ownedAttribute]
You can now drag and drop the created elements from your 'project' tree view to your diagram (we will see later how to create graphical view of these elements).
Here is some 'complex' examples that mixe collections, _iter/_apply
and model creations. For each example, we will details how the expression is created.
Example 1: creation of a Package
which contains 10 classes named C1...C10
.
cl> var p := create Package{packagedElement += range(1,10)->_apply('C' + self)->_apply(create Class{name := self})};
cl> tree(p);
+- Package 'null' [-]
| +- Class 'C1' [packagedElement]
| +- Class 'C2' [packagedElement]
| +- Class 'C3' [packagedElement]
| +- Class 'C4' [packagedElement]
| +- Class 'C5' [packagedElement]
| +- Class 'C6' [packagedElement]
| +- Class 'C7' [packagedElement]
| +- Class 'C8' [packagedElement]
| +- Class 'C9' [packagedElement]
| +- Class 'C10' [packagedElement]
There is two main parts of this expression. The first is the 10 classes creation, the second the package creation.
10 classes creations:
range(1,10)
--> creates a collection of 10 integers.
range(1,10)->_apply('C' + self)
--> concatenates each integer to 'C' (in this context, self
is an integer), the result is a collections of strings.
range(1,10)->_apply('C' + self)->_apply(create Class{name := self})
--> for each string (C1
...) create a class that you name after self
(a string in this context).
The result of this expression is a collection of classes.
package creation:
create Package{packagedElement += _previous_expr_}
--> creates a package and attach the previous 10 classes to it. As +=
operator can either add element or add all elements
of a list, the result of packagedElement += ...
is a modification of the packagedElement
collection.
Example 2: creation of 3 classes 'A', 'B', 'C' that inherits from a class named 'SuperClass' and add them to the root of your model (from an empty model).
cl> var sclass := create Class{name := 'SuperClass'};
cl> __ROOT.packagedElement += sclass;
cl> __ROOT.packagedElement += ['A', 'B', 'C']->_apply(create Class{name := self; superClass += sclass});
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 12645) [packageImport]
| +- (PackageImport, 12646) [packageImport]
| +- Class 'SuperClass' [packagedElement]
| +- Class 'A' [packagedElement]
| | +- (Generalization, 12647) [generalization]
| +- Class 'B' [packagedElement]
| | +- (Generalization, 12648) [generalization]
| +- Class 'C' [packagedElement]
| | +- (Generalization, 12649) [generalization]
cl> // here is the same code in a more compact line (assuming an empty model)
cl> var sclass := create Class{name := 'SuperClass'};
cl> __ROOT{packagedElement += sclass; packagedElement += ['A', 'B', 'C']->_apply(create Class{name := self; superClass += sclass})};
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 3764) [packageImport]
| +- (PackageImport, 3765) [packageImport]
| +- Class 'SuperClass' [packagedElement]
| +- Class 'A' [packagedElement]
| | +- (Generalization, 3766) [generalization]
| +- Class 'B' [packagedElement]
| | +- (Generalization, 3767) [generalization]
| +- Class 'C' [packagedElement]
| | +- (Generalization, 3768) [generalization]
Example 3: addition of a property named 'ID' with Integer type in each model class.
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 3764) [packageImport]
| +- (PackageImport, 3765) [packageImport]
| +- Class 'SuperClass' [packagedElement]
| +- Class 'A' [packagedElement]
| | +- (Generalization, 3766) [generalization]
| +- Class 'B' [packagedElement]
| | +- (Generalization, 3767) [generalization]
| +- Class 'C' [packagedElement]
| | +- (Generalization, 3768) [generalization]
cl> __ROOT objects()[Class]->_iter(self.ownedAttribute += create Property{name := 'ID'; type := umllib::Integer});
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 3886) [packageImport]
| +- (PackageImport, 3887) [packageImport]
| +- Class 'SuperClass' [packagedElement]
| | +- Property 'ID' [ownedAttribute]
| +- Class 'A' [packagedElement]
| | +- (Generalization, 3888) [generalization]
| | +- Property 'ID' [ownedAttribute]
| +- Class 'B' [packagedElement]
| | +- (Generalization, 3889) [generalization]
| | +- Property 'ID' [ownedAttribute]
| +- Class 'C' [packagedElement]
| | +- (Generalization, 3890) [generalization]
| | +- Property 'ID' [ownedAttribute]
Example 4: addition of a property named 'ID' with Integer type in each model class that does not inherit from another.
cl> __ROOT objects()[Class]->select(e | e.superClass = [])->_iter(self.ownedAttribute += create Property{name := 'ID'; type := umllib::Integer});
In CatsooLang, functions are defines using the def
keyword. The syntax is the following:
ex:
def foo(p1 : T1, p2 : T2, ..., pn : Tn) : returnType
{
_my_code_here_
}
cl> def helloWorld()
{
log('Hello World');
}
!! helloWorld : () -> Void <-- interpreter answer
cl> helloWorld();
Hello World
When a function is registered in the CatsooLang evaluation environement, the interpreter gives a feedback about the function name and signature. In this example,
we defined a helloWorld
function with no parameters and no return value. Functions can accept 0 or many parameters. The returnType
definition is optional
(only used for documentation purposes). If no returnType
is set, Void
is assumed by the CatsooLang interpreter (be careful, Void
keyword does not exist).
In order to return a value, the keyword return
is used.
cl> def plus1(i : Integer) : Integer { return i+1; }
!! plus1 : (i:Integer) -> Integer
cl> log(plus1(2));
3
cl> log(2 plus1());
3
In this example, you can clearly see the two ways of calling a function in action. In the first call, the parameter is passed as a function parameter whereas in the second
one, the parameter is used as function 'source'. In CatsooLang, function can return nothing (evaluated as null
) or any type of value. The same function can return either
String
, Integer
or any type even if the return type in the signature says otherwise (remember? it is used for documentation only). The example below presents not only
the Any
type in action, but also if(){}else{}
control flow structure, we will see later how if-else
works.
cl> def difftype(a : Any) : Any { if (a isKindOf(Number)) { return a + 1; } else { return 'Other'; }}
!! difftype : (a:Any) -> Any
cl> log(difftype(2));
3
cl> log(difftype(create Class));
Other
You can re-register a function in order to replace the previous one (if you made a mistake). In order to do it, your new function must have the same name and the exact same signature. Be careful, if you want to replace a function, as CatsooLang supports ad-hoc polymorphism, the signature MUST be the EXACT SAME.
cl> def fun(n : Number) : Number { return n + 3; } // <-- oups, should be n + 1
!! fun : (n:Number) -> Number
cl> def fun(n : Number) : Number { return n + 1; }
!! Replacing fun : (n:Number) -> Number
!! fun : (n:Number) -> Number
cl> def fun(i : Integer) : Number { return n + 1; }
!! fun : (i:Integer) -> Number <-- Signature is not the same, no replacement here
cl> doc(fun);
1. fun : (n:Number) -> Number
* Sorry, there is no documentation or doc is empty
2. fun : (i:Integer) -> Number
* Sorry, there is no documentation or doc is empty
As you can see, the fun
function had been registered two times with two different signatures. The doc()
function displays the register order. The function register order
is important. When you call a function, the CatsooLang interpreter evaluate each parameters, then search in its function register the first function with the same name and
with a signature which matches the evaluated parameters. Let's take another look to the previous example.
cl> def fun(n : Number) : Number { return n + 3; }
!! fun : (n:Number) -> Number
cl> def fun(n : Integer) : Number { return n + 1; }
!! fun : (n:Integer) -> Number
* We have registered two functions 'fun' *
cl> type(fun(1));
'Integer: 4
cl> type(fun(1.));
'Real: 4
In both calls, only the first function had been called. As the first fun
function works with Number
and as Integer
and Real
'inherit' from Number
, then the two calls
matche the first fun
signature. In this case, the second registered function is masked and will never be accessible. Here is what happend if we reverse the register order:
cl> def fun(n : Integer) : Number { return n + 1; }
!! fun : (n:Integer) -> Number
cl> def fun(n : Number) : Number { return n + 3; }
!! fun : (n:Number) -> Number
cl> type(fun(1));
'Integer: 2
cl> type(fun(1.));
'Real: 4
In this scenario, the two functions are accessible.
Parameters type in CatsooLang are kind of 'constraints' for signature matching and it support metamodel inheritance. In other words, you can define function that accept a 'kind of parameter type'. Here is a common example over UML:
cl> def foo(c : Classifier) : String { return c.name; }
cl> log(foo(create Class{name := 'MyClass'}));
MyClass
cl> log(foo(create Interface{name := 'MyInterface'}));
MyInterface
* Here is another way of calling the defined function *
cl> log(create Class{name := 'A'} foo());
A
cl> log(create Interface{name := 'B'} foo());
B
Function documentation in CatsooLang is easy to drop on functions. The documentation must be in top of the function signature during its definition and must be written between
""" """
.
cl> def bar() {}
!! bar : () -> Void
cl> doc(bar);
1. bar : () -> Void
* Sorry, there is no documentation or doc is empty
cl> """ *This is documentation for bar* """ def bar(){}
!! Replacing bar : () -> Void
!! bar : () -> Void
cl> doc(bar);
1. bar : () -> Void
This is documentation for bar
Of course, you can enter a new line (using SHIFT + ENTER
) and indent a little bit your function (copy/paste the following function):
cl> """
*My first documented CatsooLang function*
@param s a String
@returns a String
"""
def docfunc(s : String) : String
{
return 'Yeay ' + s;
}
!! docfunc : (s:String) -> String
cl> doc(docfunc);
1. docfunc : (s:String) -> String
My first documented CatsooLang function
@param s a String
@returns a String
The @param
and @returns
documentation keywords are automatically put in italic. CatsooLang documentation also accept ** **/__ __
and * */_ _
from markdown.
This section details a little example about collection iteration, function definitions and type constraints for a "display element" wrapper. The way CatsooLang gives feedback in the
terminal is by using the log()
functionnality. However, this is an element language and it is not a function (why? Arf, why not) so you cannot use it to iterate on elements.
cl> [1,2,3]->log();
line 1:10 no viable alternative at input '->log('
[1,2,3]->log();
^^^^
In order to introduce this behavior, we will simply define a wrapper around the log()
feature:
cl> def mylog(a : Any) {log(a);}
!! mylog : (a:Any) -> Void
cl> [1,2,3]->mylog();
1
2
3
The first parameter type is Any
. This function can be applied to any CatsooLang type (metamodel type or basic type). Nice, but it could be more interesting if you could define exactly
what you want to print. So, we register a second mylog
function:
cl> def mylog(a : Any, disp : Any) {log(disp);}
!! mylog : (a:Any x disp:Any) -> Void
This time, the first parameter is not directly displayed, just the second one. Actually, the two parameters working together in an iteration context are like a lambda abstraction. The
first parameter a
is here to contrain the element type on which the iteration will be performed. It also define a self
variable which will be bound in the disp
lambda term. Here are
some examples:
cl> [1, 2, 3]->mylog(self); // <-- equivalent to [1, 2, 3]->mylog();
1
2
3
cl> [1, 2, 3]->mylog('The number is ' + self);
The number is 1
The number is 2
The number is 3
cl> [1, 2, 3]->mylog(self toReal() pow(self toReal()));
1
4
27
In order to push this concept a little further, take a look to the _iter
and _apply
functions.
Rename all the classes/interfaces contained in a package/model to postfix them with something:
_mypackage_path_ objects()[Classifier]->_iter(self.name := self.name + _mypost_)
.
You can read this expression as:
"from _mypackage_path_
, take all the objects it contains (including itself, so a collection), filter them to keep Classifier
only and on each element assign the name
attribute of the
Classifier
to its own name postfix with _mypost
". Here is an example:
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 5728) [packageImport]
| +- (PackageImport, 5729) [packageImport]
| +- Package 'entity' [packagedElement] <-- One package named 'entity'
| | +- Class 'Entity1' [packagedElement]
| | | +- (Generalization, 5723) [generalization]
| | +- Class 'Entity2' [packagedElement]
| | | +- (Generalization, 5724) [generalization]
| | +- Interface 'Observable' [packagedElement]
| | +- Class 'Abstract' [packagedElement]
| | | +- InterfaceRealization 'null' [interfaceRealization]
cl> log(entity objects()[Classifier]->_iter(self.name := self.name + 'Entity').name); // <-- log gets the modified names (here _apply would give the same result).
[String: Entity1Entity, String: Entity2Entity, String: ObservableEntity, String: AbstractEntity]
cl> tree(__ROOT);
+- Model 'model' [-]
| +- (PackageImport, 6471) [packageImport]
| +- (PackageImport, 6472) [packageImport]
| +- Package 'entity' [packagedElement]
| | +- Class 'Entity1Entity' [packagedElement]
| | | +- (Generalization, 6466) [generalization]
| | +- Class 'Entity2Entity' [packagedElement]
| | | +- (Generalization, 6467) [generalization]
| | +- Interface 'ObservableEntity' [packagedElement]
| | +- Class 'AbstractEntity' [packagedElement]
| | | +- InterfaceRealization 'null' [interfaceRealization]