UML has been getting a lot of criticism from all sides, even from the modeling community. Sure, it has its warts:

• it is a huge language, that wants to be all things to all kinds of people (business analysts, designers, developers, users)
• it has a specification that is lengthy, hard to navigate and often vague, incomplete or inconsistent
• it is modular, but its composition mechanism (package merging) is esoteric and not well understood by most
• it is extensible, but language extensions (profiles and stereotypes) are 2nd-class citizens
• it lacks a reference implementation
• its model interchange specification is so vague that often two valid implementations won’t work with each other
• its committees work behind closed doors, there is no opportunity for non-members to provide feedback on specifications while they are in progress (membership is paid)
• <add your own grudges here>

However, even though I see a lot of room for improvement, I still don’t think there is anything better out there. The more I become familiar with the UML specification, the more impressed I am about its completeness, and how issues I had never thought about before were dealt with by its designers. And it seems that the OMG recognizes some of the issues I raised above as shortcomings and is working towards addressing them. Unfortunately, some fundamental problems are likely to remain.

In my opinion (hey, this is my blog!), for a modeling language to beat UML:

• it must be general purpose, not tailored to a specific architecture or style of software
• it must not be tailored to an implementation language
• it must be based on or compatible with the object paradigm
• it must not be limited to one of the dominant aspects of software (state, structure, behavior)
• it must be focused on executability/code generation (and thus suitable for MDD) as opposed to documentation/communication
• it must be modular, and user extensions should be 1st class citizens
• its specification should follow an open process
• it must not be owned/controlled by a single company
• it must not require royalties for adoption/implementation

My suspicion is that the next modeling language that will beat the UML as we know today is the future major release of UML. Honestly, I would rather see a new modeling language built from scratch, focused on building systems, that didn’t carry all that requirement/communication/documentation-oriented crap^H^H^H^Hbaggage that UML has (yes, I am talking about you, use case, sequence, instance and collaboration diagrams!), and developed in a more open and agile process than the OMG can possibly do. But I am not hopeful. The current divide between general purpose and domain specific modeling communities is not helping either.

So, what is your opinion? Do you think there are any better alternatives that address the shortcomings of UML without imposing any significant caveats of their own? Have your say.

The TextUML Toolkit has since release 1.2 had a metamodel extension package (inaptly named ‘meta’). This metamodel extension package defined new metaclasses not available in UML such as:

• closure - an activity that has another activity as context
• conversion action - an action that flows an input directly as output just changing the type
• literal double - a literal value for double precision numeric values
• signature - a classifier that contained parameters, the type of a closure
• meta value specification - a value specification for meta references
• collection literals - a value specification that aggregates other value specifications

Turns out extending the UML metamodel by definining new packages and metaclasses is a bad idea. Some reasons (yes, I am feeling ‘bullety’):

• it is non-standard
• other UML tools cannot read instances of your metaclasses, some won’t read your models at all if they have *any* unknown metaclasses
• there is little documentation on how to maintain these kinds of metamodel extensions
• since it is not the mainstream approach, we are bound to encounter more issues

Because of that, release 1.6 will rely exclusively on profiles and stereotypes for extending the UML metamodel.

What to expect

For conventional users of the Toolkit, this change might possibly go unnoticed, barring any potential bugs introduced in the process.

People using the built-in base package and the base_profile will be directly affected. The elements in these models are still provided, by they are now in the new mdd_extensions profile, or one of the new mdd_types, mdd_collections and mdd_console packages.

We apologize for any inconvenience these changes might bring, but we strongly believe they are required to make the TextUML Toolkit a better product. If you have any trouble moving to 1.6 (to be released later this month), make sure to hit the user forums or report issues.

I have been trying to figure out whether the TextUML notation for action semantics can be dealt with properly by tools such as Xtext and EMFText (class models and state machines should be fine). For example, given this structural model fragment:

class Advertisement
    attribute summary : String;
    attribute description : String;
    attribute keywords : Keyword[*];
    attribute category : Category;
    operation addKeyword(keywordName : String);
    static operation findByCategoryName(catName : String) : Advertisement[*];
end;

association AdvertisementKeyword
    role Advertisement.keywords;
    role advertisement : Advertisement;
end;

class Keyword specializes Object
    attribute name : String;
end;

class Category specializes Object
    attribute name : String;
    attribute description : String;
end;

association AdvertisementCategory
    role Advertisement.category;
    role ad : Advertisement[*];
end;

Notice the Advertisement class declares two operations. Their behaviour in TextUML could be written as:

    operation Advertisement.addKeyword;
    begin
        var newKeyword : Keyword;
        newKeyword := new Keyword;
        newKeyword.name := keywordName;
        link AdvertisementKeyword(keywords := newKeyword, advertisement := self);
    end;

    operation Advertisement.findByCategoryName;
    begin
        return Advertisement extent.select(
            (a : Advertisement) : Boolean {
                return a->AdvertisementCategory->category.name = catName;
            }
        );
    end;

Note that TextUML allows the behavior to be specified inline when declaring an operation in a class, or in separate, as above (that explains the lack of parameters, modifiers etc).

In the resulting UML model, the behaviour of Advertisement.addKeyword would roughly map to this (using a textual pseudo-notation for UML activities that is hopefully more readable than raw XMI):

activity(name: "addKeyword") {
    structuredActivityNode {
        variable(name: "newKeyword", type: #String)
        writeVariable(variable: #newKeyword, value: createObject(class: #Keyword))
        writeAttribute(
            attribute: #Keyword.name,
            target: readVariable(variable: #newKeyword),
            value: readVariable(variable: #keywordName)
        )
        createLink(
            association: #AdvertisementKeyword,
            end1: #AdvertisementKeyword.keyword,
            end1Value: readVariable(variable: #newKeyword),
            end2: #AdvertisementKeyword.advertisement,
            end2Value: readSelf()
        )
    }
}

and the behaviour Advertisement.findByCategoryName would map to this:

activity(name: "findByCategoryName") {
    structuredActivityNode {
        // implicit variable for return value
        variable(name: "@result", type: #Advertisement, upperBound: *)
        // implicit variable for parameter value
        variable(name: "catName", type: #String)
        writeVariable(
            variable: #@result,
            value: callOperation(
                operation: #Advertisement.select,
                target: readExtent(class: #Advertisement),
                filter: metaValue(#@findByCategoryName_closure1)
            )
        )
    }
}

// a closure is an activity that has a reference to a context activity
closure(name: "@findByCategoryName_closure1") {
        // implicit variable for return value
        variable(name: "@result", type: #Boolean)
        // implicit variable for parameter value
        variable(name: "a", type: #Advertisement)
        writeVariable(
            variable: #@result,
            value: callOperation(
                operation: #Object.equals,
                // variables from the context activity are available here
                target: readVariable(variable: #catName)
                args: readAttribute(attribute: #Category.name, target: readLink(
                    association: #AdvertisementCategory,
                    fedEnd: #Advertisement.ad,
                    fedEndValue: readVariable(variable: #a),
                    readEnd: #Advertisement.category
                ))
            )
        )
}

Note that UML does not have closures, this is an extension to the UML metamodel which I wrote about here before.

Some background on the metaclasses involved: ReadVariableAction, CreateObjectAction, CreateLinkAction, ReadExtentAction etc are all action metaclasses. Actions are the building blocks for modeling activity behaviour in UML.

The million dollar question is: can Xtext and EMFText handle more complex textual notations like this? Is this out of the happy path? Has anyone done something similar? I am under the impression I could use Xtext or EMFText better if I used them based on a intermediate metamodel for behavior that would be closer to the concrete syntax (to get all the IDE bells and whistles for free) and then transformed that to UML in a separate step.

If you have the answers for any of these questions (or even if you have comments and questions of your own), please chime in.

Even though it has always been possible to open models generated by the TextUML Toolkit in UML2-based diagramming tools, until 1.2 the Toolkit assigned new ids to each element every time a model was regenerated. That meant that any diagrams based on the model being regenerated would become invalid as any cross-references from the diagram to the model would be broken.

Starting with 1.3 M1 (test build available from the milestones update site), the TextUML Toolkit is now better compatible with diagramming tools based on UML2. You can edit the source in TextUML and save it to cause model generation to occur, and any existing diagrams should still remain valid.

Here you can see the TextUML Toolkit being used side-by-side with the UML2 Tools graphical editor:

Actually, I found out that UML2 Tools can be a great companion for the TextUML Toolkit because it seems to just do the right thing when it notices the underlying model has been changed externally: new elements show up, removed elements disappear, preserved elements preserve their layout. I tried doing the same with Papyrus 1.11 and unfortunately it does not seem to handle external updates properly. I haven’t tried with other UML2-based diagram editors, so at this point I am not sure which one represents the trend here.

I strongly believe queries are an essential part of a domain model. As such, in our quest to have (UML) models that can fully (yet abstractly) describe object models for the common enterprise applications, we cannot leave out first class support for queries.

But how do you do queries in UML? The obvious answer seems to be OCL, but that is not the approach I am taking as OCL and UML have serious interoperability/duplication issues. Instead, I took the  middleweight extension approach.

First, we model a protocol for manipulating collections of objects (showing only a subset here):

class Collection specializes Basic
  operation includes(object : T) : Boolean;
  operation isEmpty() : Boolean;
  operation size() : Integer;
  operation exists(predicate : {(:T) : Boolean}) : Boolean;
  operation \any(predicate : {(:T) : Boolean}) : T;
  operation select(filter : {(:T) : Boolean}) : T[*];
  operation collect(mapping : {(:T) : any}) : any[*];
  operation forEach(predicate : {(:T)});
  operation union(another : T[*]) : T[*];
  (…)
end;

That protocol is available against any collection of objects, which in UML can be obtained by navigating an association, reading an attribute, invoking an operation, obtaining the extent of a class (remember Smalltalk’s allInstances), anything where the resulting value has multiplicity greater than one.

Note most of the operations in the Collection protocol take blocks/closures as arguments. Closures are used in this context to define the filtering criterion for a select, or the mapping function for a collect.

For instance, for obtaining all accounts that currently do not have sufficient funds, this method would do it:

static operation findNSFAccounts() : Account[*];
begin
    return Account extent.select(
        (a : Account) : Boolean {return a.balance < 0}
    );
end;

Note the starting collection is the extent of the Account class. That is very similar to what is done in the context of query languages for object-oriented databases, such as OQL or JDOQL. We then filter the class extent by selecting only those accounts that have a negative balance, by passing a block to the select operation.

When mapping that behavior to SQL, we could end up with a query like this:

select _account_.* from Account _account_ where _account_.balance < 0

Another example: we want to obtain all customers with a balance above a given amount, let’s say, to send them a letter to thank them for their business. The following method specifies that logic:

static operation findBestCustomers(minBalance : Real) : Customer[*];
begin
    return (Account extent.select(
          (a : Account) : Boolean { return a.balance >= minBalance }
    ).collect(
          (a : Account) : Customer { return a->AccountOwner->owner }
    ) as Customer);
end;

Note that we start off with the extent of Account class, filter it down to the accounts with good balance using select, and then map from that collection to a collection with the respective account owners by traversing an association using collect.

If that was going to be mapped to SQL, one possible mapping would be:

select _customer_.* from Account _account_
    inner join Customer _customer_
        on  _account_._accountID_ = _customer_._customerID_
    where _account_.balance >= ?

Much of this can be already modeled if you try it out with the TextUML Toolkit 1.2. But, you might ask, once you model that, what can you do with UML models containing queries like the ones shown here?

Since the models are complete (include structure and behavior), you can:

1. Execute them. Imagine writing automated tests against your models, or letting your customer play with them before you actually start working on the implementation.
2. Generate complete code. The generated code will include even your custom queries, not only those basic ones (findAll, findByPK) code generators can usually produce for you.

If you would like to see tools that support that vision, keep watching this blog.

So, what is your opinion?
Do you see value in being able to specify queries in your models? Is this the right direction? What would you do differently?

UML is known to be a huge language, and that has two problems: it is too complex, having way more features than most applications will ever need, and can still be insufficient, as no single language will ever cover everybody’s needs.

In the article “Customizing UML: Which Technique is Right for You?”, James Bruck and Kenn Hussey (both from the UML2 team) do a great job at covering the several options for extending (or restricting) UML (James also made a related presentation at last year’s EclipseCon, together with Christian Damus, of Eclipse OCL fame). Cutting to the chase, these are the options they identify:

• using keywords (featherweight extensions)
• using profiles, stereotypes and properties/tagged values (lightweight extensions)
• extending the metamodel by specializing the existing metaclasses (middleweight extensions)
• using package merges to select the parts of UML you need (heavyweight extensions)

Each option has its own strengths and weaknesses, as you can see in the referred article/presentation. At this time, the TextUML notation supports two of those approaches: profiles and metamodel extensions.

Adding closures to UML

Even though profiles are the most popular (and recommended) mechanism for extending UML, it is not enough in some cases/applications. That has been the case in the TextUML Toolkit, for instance, when implementing closures in the TextUML action language (yes, the Toolkit eats its own dog food).

According to the wikipedia entry, “a closure is a function that is evaluated in an environment containing one or more bound variables. When called, the function can access these variables. ”

It really makes sense to (meta) model a closure as some kind of UML activity, which is basically a piece of behavior that can be fully specified in UML. Methods, for instance, are better modeled in UML as activities. Activities are composed of activity nodes and actions, which are similar to blocks of code and instructions, respectively.

The only thing that is missing in the standard Activity metaclass is the ability for a closure to have an activity node from another activity as context, so it can access context’s local variables. So here is a possible (meta) modeling of closures in UML using the TextUML syntax:

[Standard::Metamodel]
model meta;

apply Standard;

(*
  A closure is a special kind of activity that has another
  activity’s activity node as context. A closure might
  reference variables declared in the context activity node.
*)
[Standard::Metaclass]
class Closure specializes uml::Activity
  (* The activity node that provides context to this closure. *)
  reference contextNode : uml::StructuredActivityNode;
end;

end.

Or, for those of you who prefer a class diagram (courtesy of the EclipseGraphviz integration):

Note a model contributing language extensions must be applied the Standard::Metamodel stereotype, and each metaclass must be assigned the stereotype Standard::Metaclass.

Of course, there is no point in being able to metamodel closures, if there is no way to refer to them. We need a kind of type that we can use to declare variables and parameters that can hold references to closures. That also means we need to be able to invoke/dereference a closure reference, and there is no support for referring to metamodel elements in the UML action language. That means we need more language extensions. But I will leave that to another post. My goal here was to show how simple it is to create a simple UML metamodel extension with the TextUML Toolkit.

What about you, have you ever needed to extend UML using a metamodel extension? What for?

No tool is an island. That is even more important when we are talking about highly focused single-purpose tools such as the TextUML Toolkit. As you probably know, the TextUML Toolkit is a tool for UML modeling using a textual notation, but that is about it. The TextUML Toolkit itself won’t help if you need features such as code generation, reverse engineering or graphical modeling/visualization. That might look as a negative thing to some, but that is by design. I am a strong believer of separation of concerns, component-based software and using the best tool for the job, and that is reflected in the design of the Toolkit.

This post will describe how to use models created by other UML2-compatible modeling tools from models you create using the TextUML Toolkit. I plan to cover other areas of integration in future posts.

Reading models created by other UML tools using the TextUML notation

If you have the TextUML Toolkit installed, one of the editors available for UML files is the TextUML Viewer. As the name implies, it is not really an editor, i.e., you cannot edit models with it, just visualize them using the TextUML notation. But it is an useful tool for reading UML models you don’t have the source for, i.e. models created by other UML2 modeling tools.

Using models created by other tools…

There are a few different ways of using models created by other UML modeling tools from your models created in TextUML.

…by co-location

In this method, you just drop the UML model you want to use to at the root of your TextUML (i.e. MDD) project. All models at the root of the same MDD project are considered to be in the same repository and thus you can make cross-model references by either using qualified names or importing the package and using simple names.

This method is dirty simple, but it forces you to keep the foreign model at a specific location, potentially forcing you to keep multiple copies of the model.

…with project references

In this method (available starting in version 1.2), the UML model you want to use is in a different project than the one you want to refer it from. You just open the properties dialog for your TextUML (MDD) project and add the project providing the UML model you want to use to the list of referenced projects. Now models at the root of that project will also be visible from your TextUML (MDD) project, in other words, they are also seen as part of the same repository.

This method is more powerful than the previous one, as now models can be shared by reference instead of by copying. There is a limitation still: the models you use ought to be part of some project in the workspace, so you cannot refer to models stored at arbitrary locations in the file system or models that are contributed by Eclipse bundles.

…by explicitly loading them

For this method, you use the load directive to load an external model located by a URI.

This method allows you to load any model, provided its location can be represented as a resolvable URI. That is the case of any file system path, or Eclipse platform URLs. The caveat is that some URIs such as file system URIs are not portable, so they are are bound to be invalid on other machines.

The entire truth…

…is that even if you are using only the TextUML Toolkit for building UML models, these are the same mechanisms for creating models that make cross-package references (see also the TextUML Guide). For the TextUML Toolkit, all UML2-compatible models are equal, no matter how they were created.

Yesterday I checked in support for UML primitive types in the TextUML Toolkit. As an example of the syntax, here is the UMLPrimitiveTypes model, shipped in Eclipse UML2, rendered by the TextUML Source viewer:

[Standard::ModelLibrary]
model UMLPrimitiveTypes;

apply Standard;

primitive Boolean;

primitive Integer;

primitive String;

primitive UnlimitedNatural;

end.

As you can see, primitive types are declared using the keyword ‘primitive’, and cannot specialize other types or have structural features (thus there is no ‘end’ terminating the type declaration). To be honest, I am not sure this is the intended semantics, as the UML specification does not seem to explicitly disallow that.

Here is the breakdown of the change sets (r188-191):

• automated tests (r190)
• parser and model generation (r191)
• textual renderer (r189)
• editor (r188)

To be frank, the only test case added was quite trivial:

public void testPrimitiveType() throws CoreException {
  String source = "";
  source += "model someModel;\n";
  source += "primitive Primitive1;\n";
  source += "end.";
  parseAndCheck(source);
  PrimitiveType found = (PrimitiveType) getRepository().findNamedElement(
    "someModel::Primitive1", IRepository.PACKAGE.getPrimitiveType(), null
  );
  assertNotNull(found);
}

But it is a good start. It ensures parsing works, and the model generated has the intended element created under the right parent package.

Any suggestions of what UML feature to cover next? And what is your take? Can UML primitive types have structural features or specialize other types/implement interfaces?

(By the way, if you want to understand how the TextUML Toolkit is implemented, feature posts like this are a good starting point)

Just checked in a new feature in the TextUML Toolkit: support for data types (a.k.a. structs). This is an example of a data type declaration:

datatype UserName
  attribute firstName : String;
  attribute lastName : String;
end;

You can declare operations and specialize other classifiers as usual.

Change set

Here are the individual changes per plug-in (r177-r186):

• automated tests (180 and 185)
• parser and model generation (181, 184 and 186)
• textual renderer
• editor

There, nice and easy. Well, actually, I ended up doing a second commit because I forgot to ensure a data type specializes only other data types. But now that has been taken care of.

As usual, suggestions of UML features to support in the TextUML Toolkit are most welcome.

Just checked in a new feature in the TextUML Toolkit: required extensions for stereotypes (honestly, I didn’t know about that feature in UML until I read this post on the Eclipse UML2 newsgroup).

The following is a stereotype extending two metaclasses (uml::Class and uml::Operation):

profile my_profile;

import uml;

stereotype foo extends Class, Operation required
end;

end.

In the example above, the extension of Operation is required, the extension of Class is not.

Change set

The change set to implement required extensions in the TextUML Toolkit (issue #2266268) was quite small (btw, kudos to the UML2 project for providing the best API for UML out there).

Here are the individual changes per plug-in (r153-r156):

• automated tests
• parser and model generation
• textual renderer
• editor

Any other UML feature you would like to see exposed in the TextUML notation? Suggestions are welcome.

The first milestone build of the next TextUML Toolkit release is now available from the milestone update site. This preview build is the first to include support for modeling behavior using action semantics. I hinted at this capability here before, and I plan to cover action semantics in the TextUML notation in following posts. But most people will get the gist of the notation from the examples below:

Here is an example of a UML model of an Account class represented in TextUML:

model bank;

import base;

class Account
    attribute accountNumber : String;
    attribute balance : Integer;

    operation withdraw(amount : Integer);
    begin
        self.balance := self.balance - amount;
    end;    

    operation deposit(amount : Integer);
    begin
        self.balance := self.balance + amount;
    end;

    static operation newAccount(number : String,owner : Client): Account;
    begin
        var newAccount : Account;
        newAccount := new Account;
        newAccount.accountNumber := number;
        newAccount.balance := 0;
        link ClientAccount(owner := owner, accounts := newAccount);
        return newAccount;
    end;
end;

class Client specializes Object
    attribute name : String;
    static operation newClient(name : String): Client;
    begin
        var newClient : Client;
        newClient := new Client;
        newClient.name := name;
        return newClient;
    end;
end;

association ClientAccount
    navigable role owner : Client[1];
    navigable role accounts : Account[0, *];
end;

end.

And here an example of a test driver ‘program’ (note the use of a closure for looping through a collection of objects):

package test;

apply base_profile;

import bank;
import base;

class TestDriver
  [entryPoint]
  static operation run();
  begin
      var john : Client, mary : Client,
             account1 : Account, account2 : Account, account3 : Account;
      john := Client#newClient(”John Doe”);
      mary := Client#newClient(”Mary Doe”);

      Console#writeln(”Created: “.concat(john.name));

      account1 := Account#newAccount(”1234-1″, john);
      account2 := Account#newAccount(”1238-2″, mary);
      account3 := Account#newAccount(”2231-7″, john);

      Console#writeln(”account owner: “.concat(account1->ClientAccount->owner.name));
      Console#writeln(”account number: “.concat(account1.accountNumber));
      Console#writeln(”initial balance is: “.concat(account1.balance.toString()));

      account1.deposit(2000);
      Console#writeln(”after deposit, balance is: “.concat(account1.balance.toString()));

      account1.withdraw(500);
      Console#writeln(”after withdrawal, balance is: “.concat(account1.balance.toString()));

       /* Now show information for all accounts. */
  	Account extent.forEach(
          (a : Account) {
              Console#writeln(”*******”);
              Console#writeln(”Owner: “.concat(a->ClientAccount->owner.name));
              Console#writeln(”Number: “.concat(a.accountNumber));
              Console#writeln(”Balance: “.concat(a.balance.toString()));
          }
      );
  end;
end;

end.

And here is the output of the test driver program, produced by running it on the Libra UML runtime (not part of the TextUML Toolkit):

Created: John Doe
account owner: John Doe
account number: 1234-1
initial balance is: 0
after deposit, balance is: 2000
after withdrawal, balance is: 1500
*******
Owner: Mary Doe
Number: 1238-2
Balance: 0
*******
Owner: John Doe
Number: 1234-1
Balance: 1500
*******
Owner: John Doe
Number: 2231-7
Balance: 0

Alternatively, one could have generated 100% of the code for a target platform of choice, given that all the information necessary for that is in the model. Note that code generation is not part of the TextUML Toolkit.

Do you want to give it a try? Install M1 from the milestone update site. You can also fetch the example projects from here.

It is certainly a rough stone. I am counting on the community feedback to figure out what areas need to be smoothed first. Please provide your feedback or ask questions on the TextUML Toolkit forums.

Do you know what UML can do for you? I mean, did you know that UML models can actually do things?

One of the least known features of UML is that you can model detailed imperative behavior. The UML “instruction set” can do things like:

• create and destroy objects
• create and destroy links (associations) between objects
• read and write attributes and local variables
• invoke operations and functions
• throw and catch exceptions
• conditional statements
• loops

That is quite amazing, isn’t? And all that while still preserving a high level of abstraction. Such capability is generally referred to as ‘action semantics’. Action semantics provides the basic framework for executability in UML and has been there for quite a while now. It was originally added to the spec, first as patch, in UML 1.5 (2003), and then more seamlessly integrated into UML 2.0 and following spec releases.

Action Semantics and TextUML

An even more well-kept secret is that the TextUML notation supports UML action semantics and thus the creation of fully executable UML models. This support is not yet shipped as part of the TextUML Toolkit, but will be in the next release. Meanwhile, if you want to give it a try or take a closer look, you will have to grab the source from the SVN repository.

I plan to go into more details in the near future, but just to wet your appetite, here is one example of an executable UML model described in the TextUML notation:

package hello;

apply base_profile;
import base;

class HelloWorld
    [entryPoint]
    static operation hello();
    begin
        Console#writeln(”Hello, World”);
    end;
end;

end.

Cool, isn’t? If you had a UML runtime, this model could be executed even before you made a decision about what platform to target. Also, if your code generator were action semantics aware, you could trigger code generation for the target platform(s) of choice, with the key difference that you could achieve (or get very close) to full code generation, as the model now also describes behavior. No more of that monkey business of having to edit the generated code and manually fill in all those /* IMPLEMENT ME! */ methods.

Do you think this has value? Would you want to work with a tool that supported that? I am really keen on knowing your opinion.

This is actually old news for many people, but recently I learned (by pure chance) that the OMG issued a RFP for a “Concrete Syntax for a UML Action Language”. Letters of intent are due on December 8th. Submissions, one year after. OMG members only need apply (Aww…). I wonder if anyone in the Eclipse Modeling project is involved in submitting a proposal. Anyone?

Soapbox: since version 1.5, UML has had support for algorithmic behavior specification, commonly called action semantics. It is still hardly used, and most people that consider they know a lot of UML have never noticed it. Some people believe that the lack of an official concrete syntax for action semantics is a barrier to adoption. You see, the OMG defined the semantics and an abstract syntax, but left concrete syntax as an exercise for tool vendors.

As I wrote here before, I don’t really see the value in the OMG godfathering one concrete syntax over all others. Of course, we are talking here about a syntax for human beings, not for tools. Tools certainly don’t need a human-readable concrete syntax, a standard binary or XML format will do. The problem is: we all have our own preferences for what makes a good syntax, and there is no single syntax that will make everybody happy, so we are bound to have multiple concrete syntaxes anyway. We all like interoperability between tools, but when it comes to sugar, we like choice.

But maybe I am wrong. Maybe an OMG-blessed C-like concrete syntax for UML is all that is missing for Executable UML to become mainstream in the software development community. Go figure, we are an amusing bunch. Personally, I don’t care that much. I have been a Java developer for around 12 years now, so I can certainly stand another C-like syntax. We are not talking about a language anyway, it is just a syntax for an existing language, and syntax, a bit like UI, is inherently disposable, if you take it away, the real stuff is still there.

One clear positive outcome of the RFP is that submitters must provide, along with the proposal for a concrete syntax, any changes to fUML* that would be required to support such action language. That will probably help closing some gaps in the UML specification that make it hard to execute if you are stuck with the standard.

There are many other interesting bits in the proposal, but I will leave a more detailed analysis to a future post.

* the Executable UML Foundation Submission says:“Foundational UML Subset (fUML) is that subset of UML required to write ‘programs’ in UML”

I posted this earlier this week on the UML2 newsgroup, thought I would share it here too…

I published a TextUML rendition of the UML 2.1 metamodel that is shipped by Eclipse UML2 here (warning: 5.9k lines, 370Kb of text). It starts like this:

(...)
model uml;

apply Ecore;
apply Standard;

import ecore;

(* A comment is a textual annotation that can be attached to a set of elements. *)
[Standard::Metaclass]
class Comment specializes Element
    (* Specifies a string that is the comment. *)
    [Ecore::EAttribute(isID=false,isTransient=false,isVolatile=false,isUnsettable=true,annotations=[])]
    public attribute body : String;
    (* References the Element(s) being commented. *)
    public attribute annotatedElement : Element[0, *];
end;
(…)

Thought it could be useful to anyone wanting to have a quick look at the UML metamodel. The TextUML renderer does not support all UML element types, so some elements might be missing/incomplete.

I think that is a nice demo of the textual browsing capabilities in the TextUML Toolkit. Of course, much more could be done in that area. If you feel like lending a hand, any help (bug reports included) will be much appreciated.

Shorthand notation for both composite and aggregate associations has just been checked in (r99-r100). It follows the same spirit as the existing shorthand for plain associations (a.k.a. references).

composition <end-name> : <referenced-type-name>;

or

aggregation <end-name> : <referenced-type-name>;

That will result in a new unnamed association with two member ends, one named, owned by the declaring classifier, with composite or shared aggregation type, and another unnamed, owned by the association itself.

Change set explained

When adding a new feature to the notation, the typical change set contains:

• updates to the affected grammar production rules
• implementation of the corresponding model generation handling code
• a few test cases.

In the case of the features presented in this post, the grammar change was quite simple (nicer view):

--- trunk/plugins/com.abstratt.mdd.frontend.textuml.core/textuml.scc	2008/09/16 05:47:13	99
+++ trunk/plugins/com.abstratt.mdd.frontend.textuml.core/textuml.scc	2008/09/16 05:47:20	100
@@ -341,7 +341,9 @@

   attribute_decl = attribute identifier colon type_identifier optional_subsetting semicolon ;

-  reference_decl = reference identifier colon type_identifier optional_subsetting  semicolon ;
+  reference_decl = reference_type identifier colon type_identifier optional_subsetting  semicolon ;
+
+  reference_type = {association} reference | {composition} composition | {aggregation} aggregation ; 

   optional_subsetting = subsets qualified_identifier | {empty} ;

In other words, where one could declare a reference, one can now declare also a composition or an aggregation (gotta love the readability of SableCC grammars…).

The corresponding model generator (a.k.a. compiler) changes were quite trivial as well, even more so if you consider I inadvertently checked in a fix to an unrelated bug around comments (just ignore the changes in lines not in the 300’s).

In terms of lines of code, the bulk of the changes were in the test class for associations, where I added three test cases. I had postponed writing tests for the reference shorthand notation, so I took this opportunity to write a test for it too.

As most test cases in the TextUML Toolkit test suite, the new test cases in AssociationTests have the following layout:

• one or more compilation units with TextUML source code are declared
• source code is compiled and the resulting errors verified (most test cases won’t expect errors)
• the resulting model is checked - were the expected elements created? Do they have the expected features?

Talking about tests

The TextUML Toolkit has a modestly sized test suite (~130 test cases) at this time. It could use dozens, probably hundreds more, but most features are covered to so some extent.

However, as important as coverage, is how much one can learn about the piece of software being tested by reading its test suite. And here is where I believe the TextUML Toolkit’s test suite shines. For example, by reading the test classes, one could (maybe more effectively than by reading the notation guide), learn how the following UML features are exposed by the TextUML notation:

• classes and structural features
  
• packages, imports, profiles
• associations
• multiplicity
• template classes

Trivia: can you find out from the test suite the difference between a private and a public package import? Give it a try and let me know if I am lying… by the way, the notation guide does not cover that yet, although the UML specification, of course, does.

I often see the TextUML Toolkit being compared to tools that produce UML diagrams from textual descriptions. Examples of tools like that are ModSL, MetaUML and UMLGraph. But that doesn’t really make sense, and here is why: these tools produce diagrams from text. The TextUML Toolkit produces models from text. But what is the difference between models and diagrams?

According to Wikipedia:

• A diagram is a 2D geometric symbolic representation of information according to some visualization technique.
• A model is a pattern, plan, representation (especially in miniature), or description designed to show the main object or workings of an object, system, or concept.

Note that even though both terms are defined around the word “representation”, the term “diagram” implies graphical visualization, whereas the term “model” admits any kind of media, basically because models have no concrete form per se.

Now, please, if you are not convinced yet, read aloud 5 times: MODELS ARE NOT DIAGRAMS!

If that didn’t work, well, maybe the facts below will help:

• models, not diagrams, are the subject matter of model-driven development.
• models, not diagrams, can be validated.
• models, not diagrams, can serve as input to code generation.
• models, not diagrams, can be automatically generated from reverse engineering source code.
• models, not diagrams, can be executed.

Even though diagrams are commonly used for representing models, they are not the only way, and non rare not the most appropriate one (yes, I am talking again about the virtues of textual notations, but I won’t repeat myself).

P.S. Maybe it helps adding to the confusion the fact that the TextUML Toolkit has an optional feature that will generate class diagrams automatically from the UML models created with it. However, that is just an optional, loosely integrated feature, and it is definitely not the Toolkit’s thing. Weirdly enough, from my observation, that feature is the main reason most people become interested in the TextUML Toolkit. Well, go figure.

I just completed the code for supporting parameter direction modifiers. When this feature makes into the next build, modelers will be able to choose for named parameters any of the standard parameter direction kinds: out, inout, or (the default) in. By the way, the syntax for specifying return types remains the same. Here is an example:

  operation op1(in p1 : String, out p2 : String);

Java developers might consider this feature unnecessary as there is only one way of passing parameters in Java, but for distributed applications (such as those using CORBA or web services) these modifiers are quite relevant.

The corresponding support for rendering operation parameter direction kind modifiers has been added to the TextUML model viewer (decompiler) and EclipseGraphviz graphical rendering component.

This is the first new feature since version 1.0 was released.  The impact of implementing it was quite minimal to the code base, so I am looking forward to implementing the next notation feature. The question is: what feature? Any suggestions?

One potential downside people often point out about using TextUML as a notation for UML is that it is yet another language to learn. But that is not really a well founded argument. TextUML is not a full-blown language, it is just an alternative notation for UML, the de facto standard language for modeling. The language semantics of UML are defined in terms of an abstract syntax, and even though the specification includes sections about the graphical notation, it clearly supports the idea of alternate concrete syntaxes.

Moreover, we really shouldn’t care much about the actual notation being used, be it the standard graphical notation, TextUML or any other textual notation. I surely don’t. Regardless what notation you use for creating UML models, in the end there is only one language. All you know about the semantics of UML model elements is still true no matter what syntax you choose (for instance, if you know your UML, you should easily become productive with the TextUML Toolkit by just glancing over the notation guide now and then). In fact, I predict a time where people will want to move between different notations across tasks and time. Also, in the same team, different people will be collaboratively creating UML models using different notations.

This will only be possible because supporting a new concrete syntax is much simpler than supporting a whole new language (by the way, that is the same reason why I believe UML-based DSLs to be a better option than homegrown proprietary DSLs, but I digress). Any reasonably good programmer armed with a parser generator and knowledge of the metamodel should be able to write a compiler for a textual notation for UML, and Eclipse makes it really easy to provide basic IDE features such as those you see in the TextUML Toolkit. Also, there are tools in the horizon such as IMP and TMF that will make these tasks really a breeze.

Meanwhile, you can stick with what is here today.

Contrary to my original intention of working this week on code generation and the sample application, I have been doing some improvements in the graphical visualization of UML models in EclipseGraphviz. EclipseGraphviz (a spin-off of the TextUML Toolkit) is an open source component (EPL) that integrates Graphviz into Eclipse, and among other things, can generate UML class diagrams from Eclipse UML2 models on the fly.

After fixing a couple of bugs, I decided to add a preference page to give some level of control over the way the diagrams are laid out. Here is what it looks like:

You can now turn on or off several adornments such as association end names, multiplicities and membership. For example, this diagram was rendered using the options shown above:

Whereas the following one was rendered for the same model while having only the “Create constraints for association navigability” option checked:

As you can see, enabling constraints based on association navigability can significantly alter the diagram layout. One good reason for enabling that option is to avoid diagrams that grow too much horizontally (as Graphviz will put all nodes of the same rank on the same row), but I personally prefer the former layout.

Side note: I considered dropping the graphical visualization feature altogether for 1.0, as I see it as just a nice to have, and am not totally happy with the quality of the diagrams generated by Graphviz. But every bit of feedback I got for the TextUML Toolkit was that the graphical visualization was really nice, so I changed my mind. But unless serious bugs are found in this area, I have no plans of touching the EclipseGraphviz code again until after I complete the 30-day challenge.

Side note #2: I expanded the FAQ to cover the issue of notation choice and the role of EclipseGraphviz in the TextUML Toolkit.

There was a recent thread about UML on Slashdot, as a reaction to this blog post . The headline: “Is UML Really Dead, Or Only Cataleptic?“. Many posters are clearly bitter towards UML. There seems to be a strong preference for using UML as a communication tool as opposed to as a basis for partial or full code generation (see post on UML modes). Many also complain that the graphical notation is cumbersome and that it hurts productivity (+1!). A few actually like UML; however, it is said that the UML specification is vast and complex and that you should pick the parts of UML that make sense for your case/goals (+1 here too).

Lots of interesting points, but the most negative posts are just misinformed. But that is Slashdot, what should I expect? Java and Eclipse have, in general, a poor receptivity on Slashdot, so I guess UML is in good company.

Of course, while I disagree with those that ditch UML because it was not properly employed in some project they worked, I strongly agree with the complaints about UML (graphical) diagrams being cumbersome and hard to deal with, as I have written here before. But that is not a problem with UML per se, but with the fact most still see it as a graphical language.

UML certainly has its share of problems (design by committee, no reference implementation), but I strongly believe it is very useful and that there isn’t anything out there (I am talking to you, home grown DSLs) that can replace it as the lingua franca for model-driven development.

TextUML is a textual notation for UML. The TextUML Toolkit is an Eclipse-based IDE-like tool for creating UML models using the TextUML notation.

Other tools follow the same approach. Emfatic (now an EMFT subproject) has been doing the same for EMF Ecore for a long time; the TMF project aims to be for textual modeling what GMF is for graphical modeling, and will be based on GMT’s TCS and xText components.

Still, people are often puzzled when I explain what the TextUML Toolkit is. A common question is: “if I am going to write code (sic), why do I need UML anyway?“.

Dean Wampler from Object Mentor wrote on his blog a while ago a post entitled “Why we write code and don’t just draw diagrams“. It is a short post, but he presents very good points on why a graphical notation is usually not suficient and is bound to be less productive than a textual one when it comes to modeling details. For instance, on the saying “a picture is worth a thousand words“, Dean wrote:

“What that phrase really means is that we get the ‘gist’ or the ‘gestalt’ of a situation when we look at a picture, but nothing expresses the intricate details like text, the 1000 words. Since computers are literal-minded and don’t ‘do gist’, they require those details spelled out explicitly.”

Couldn’t have said it better.

I strongly advise you to read the original post in its entirety, but I will leave you with another pearl from Dean’s post (emphasis is mine):

“I came to this realization a few years ago when I worked for a Well Known Company developing UML-based tools for Java developers. The tool’s UI could have been more efficient, but there was no way to beat the speed of typing text.”

Enough said.

One common misconception is that UML is a graphical language and that any tools adopting alternative notations (such as a textual one) are inherently non-compliant. That can’t be farther from the truth. Read on to understand.

The fact is that the UML specification clearly separates abstract syntax (the kinds of elements and how they relate) and semantics (what they mean) from concrete syntax (what they look like), and states that there are two types of compliance:

• abstract syntax compliance
• concrete syntax compliance

Concrete syntax compliance means that users can continue to use a notation they are familiar with across different tools. This is important when UML is used as a communication tool in a team environment. Architects, designers, programmers and even many business analysts speak the same language.

Abstract syntax compliance means that users can move models across different tools, even if they use different notations. This is essential when UML is used as a basis for model-driven development. You might want to use tool A for creating the model, tool B for validating the model, tool C for somehow transforming/enhancing the model and tool D for generating code from it (a common form of MDD tool chain).

The TextUML Toolkit uses a textual notation that strictly exposes the UML semantics, but it is not compliant with the language concrete syntax by any means. On the other hand, the toolkit uses Eclipse UML2’s XMI flavor for persisting models and thus is fully compliant regarding the abstract syntax. That is consistent with the vision for the product: a tool from developers for developers that want to build software in a more sane way: the model-driven way. Developers can create models using a notation they are more productive with. Models can then be used as input for code generation using many of the tools available in the market. If non-developers might frown upon a textual modeling notation, they will always have the option of using their favorite graphical-notation based tools for reading the models. I mean, if their tool of choice is abstract syntax compliant as well.

I just added a brand new FAQ section to the web site. It was long overdue.

Here is the first batch of questions:

1. Can a tool that uses a textual notation be considered UML compliant?
2. What are the differences between TextUML and the real UML?
3. I thought UML was all about raising the level of abstraction. Why should I go back to coding?
4. How much does the TextUML Toolkit cost?
5. Is the TextUML Toolkit open source?
6. What diagrams can I create with the TextUML Toolkit?

Do you have any questions that are not answered in the FAQ or do not agree with or understand an answer? Fire your comments, I am always happy to hear from you.

Andreas Awenius, from Empowertec, was kind enough to devote not only one, but two blog postings on the TextUML Toolkit. While his first post was just a paragraph or two acknowledging the use of a textual notation for UML modeling in the TextUML Toolkit, his second post was a very detailed and balanced analysis on the whether using textual notations is a good idea. Even though the conclusion of his posting is that a graphical notation is superior to a textual one, I still agree with many points Andreas has made in his analysis:

1. the presentation notation and the persistence format are completely unrelated concerns (exactly, see my previous comment on this).
2. diagrams are views for models, focused on a specific aspect, and as such must show only what is needed and omit anything that is not essential. (perfect)
3. there are “multiple ways to manipulate the repository, e.g. the conventional - graphical - way or a textual notation”. (yes!)

These are the very same facts that justify the existence of a tool such as the TextUML Toolkit.

1. since user-oriented notation and persistence format are completely different things, you can please users with the most appropriate syntax for their needs (and that is text for many developers), without affecting interoperability with other tools, as tools interoperate at the level of the persistence format. The TextUML Toolkit achieves that by producing UML2 models as the native object file format.
2. the graphical notation is great for creating views, but it is not that suitable for exposing all the details in a model. A textual notation has no problem doing that. The graphical notation is still useful, and this is why TextUML Toolkit will include a read-only graphical view for browsing the repository. And here is the first point I disagree with Andreas. He says: “a diagram must carefully be crafted manually, leaving out all irrelevant detail”. I say: automatically generated diagrams can still obey a set of preferences defined by the user (level of detail, color, font, etc). UMLGraph has been doing this for a long time (albeit by breaking some basic rules of separation of concerns, but I digress). The TextUML Toolkit will embed an upcoming version of UMLGraph in its next milestone.
3. this is basically a conundrum of points #1 and #2. The model created will be the same (from the point of view of tools) regardless the user-facing notation. If users say they would rather use a textual notation, why forcing them to use a graphical one when there are no technical reasons to do so? Here is where Andreas makes several points to back his opinion based on technical and strategical reasons, most of which I (strongly to moderately) disagree. Instead of trying to refute each of those arguments (that is not how you encourage people to provide elaborate feedback), I will use them as an opportunity for driving awareness efforts around the Toolkit, the notation and our future offerings in upcoming posts to this blog.

The ‘raison d’être‘ for this blog is to establish connections with the modeling community at large, users and tool providers. I really appreciate the time and energy that Andreas committed to sharing his views on the TextUML Toolkit and the textual notation approach and am both grateful and honored for that. Thanks a lot, Andreas. Stay in touch. I hope at some point I can convince you that you are wrong!

Last Thursday happened the 1st Eclipse DemoCamp in Vancouver. It was a very dynamic event, with nine short talks packed into little more than one and a half hours. Attendance was high, I was really amazed to see such a great turnaround. The audience saw a diverse range of interesting Eclipse-based projects being developed in Vancouver/Victoria.

I demoed the TextUML Toolkit and got good feedback and some very interesting questions. The idea of a using a textual notation for creating UML models is still not a very easy sell, and that continues to surprise me. Being a developer, I find working with a textual notation so much more natural than a graphical one that it makes it hard for me to see things from other people’s shoes. I plan to go over the issue of textual versus graphical notations on a future post.

All in all, it was a great event. Many thanks to Tasktop’s Robert Elves and Mik Kersten for organizing it, to the Eclipse Foundation for sponsoring the event, and congrats to all presenters.

In a recent thread on the EMF newsgroup, I came to realize that there are (at least) two EMFs, each belonging to a different class of product, and attending completely distinct requirements.

EMF’s native metamodel, Ecore, is a generic, lean object-oriented metamodel based on a subset of UML.

EMF’s runtime framework provides Java applications with runtime support for object models “including change notification, persistence support with default XMI serialization, and a very efficient reflective API for manipulating EMF objects generically” (source).

I am a happy user of EMF’s Ecore metamodel as a poor man’s UML (EMF team, please take that as a compliment). From Ecore-based models, using a compatible template engine, I can generate all code/artifacts that are prone to automation, be they Java code or not.

The runtime aspect of EMF is certainly useful to many applications, but certainly not to all or maybe even most. That is not to say that the EMF runtime does not provide a lot of value, as it clearly does given its popularity. Nor does it imply that the EMF runtime API has design flaws that prevent its use to be more widespread. The fact is that frameworks, while designed for extension, always impose a certain set of architectural decisions in order to provide value out-of-the-box. Those decisions are bound to make its use more suitable to some scenarios and applications than others. The goal is to make as many people happy as possible. It is clear that the EMF team has pulled off that trick. But that does not mean that using EMF-generated model code is appropriate for every Java application out there.

Model driven development maximizes reuse via a complete separation between problem domain and technological concerns. This separation is critical to allow us to build software in an obsolescence-proof way. EMF’s Ecore provides a good foundation for MDD in this sense, regardless the technology choices for the software being developed. The EMF runtime framework, albeit a valuable tool for a significant range of applications, brings with it a specific set of choices in terms of design and implementation decisions, and as such has a less universal applicability.

However, time and again I read comments in the EMF newsgroup that lead me to believe that this duality might have been accidental, and that the sole reason Ecore was created was to support the generation of Java applications based on the EMF runtime. If that is really the case, this is something that both amuses and worries me. My concern is that if the EMF team does not acknowledge the importance of Ecore as an independently useful product, technical decisions in the evolution of EMF might break the use of Ecore in contexts other than EMF-based Java applications.

One weird thing about UML is that there aren’t collection types or array types. Basically, multiplicity and typing are totally independent concerns, represented by the metaclasses TypedElement and MultiplicityElement.

A typed element is a named element that has a type, and that is all about it. Examples of typed elements are value specifications, properties, parameters, pins and variables.

A multiplicity element, on the other hand, is an element that when instantiated potentially admits a collection of values. An optionally defined lower bound value (defaults to 1) can determine the minimum number of instances expected. Whether multiple values are in fact admitted will depend on the upper bound of the multiplicity element, which defaults to 1 (no multiple values allowed), but can be set to any positive integer, or infinity. A multiplicity element that can actually be multivalued can also be characterized regarding ordering (whether values can be accessed by position) and uniqueness (whether values can be repeated).

Some kinds of elements are both typed and support multiplicity (such as properties, parameters, pins and variables), however a few are one or the other (let’s ignore those in this discussion).

Mapping UML to Java

To try to illustrate all that was said above, let’s see a few examples of Java variable declarations and the equivalent declaration in UML (using TextUML syntax):

There are few interesting differences though:

1. in UML, it is the typed element itself that defines multiplicity, and not the type
2. c:Client, c:Client[1] and c:Client[1,1] are all equivalent
3. c:Client[*], c:Client[1,*] are equivalent
4. if a value is optional, the lower bound must be specified to be 0 (example: c:Client[0,1]). There is no Java equivalent for that.
5. unique and unordered are the default in UML (you can use the modifiers ‘nonunique’ and ‘ordered’ to override the defaults)

There are some implications related to assignment when the source and destination are multiplicity elements:

1. the upper bound of the destination must be the same or greater than the upper bound of the source, or a type mismatch will ensue
2. what happens if source and destination differ on ordering or unicity? The UML spec does not sem to cover that (let me know if you think otherwise). In TextUML, any required transformations are performed automatically behind the scenes. For example: if the source is non-unique and the destination is unique, duplicates will be silently suppressed, or if the source is unordered and the destination is ordered, an arbitrary order will be defined.

Well, that was it for today. If you have questions, suggestions or think I got something wrong here, feel free to add a comment.

One of the least known and understood concepts of UML is templates. Section 17.5 on version 2.1.1 of the UML specification covers the Templates package in 31 pages. What follows is an attempt at providing a summary of the mechanism in a way that is easy to understand without actually omitting any important details.

A simple example

The following example in TextUML should be easy to understand for any developer familiar with C++ parametrized types or Java generics:

class Foo
end;

class Bar<T>
    attribute prop1 : T;
    operation op11(par1 : T);
end;

class Fred
    attribute attr1 : Bar<Foo>;
end;

Class ‘Bar’ is a template class, whose template signature contains a single parameter: ‘T’. The type of the property ‘prop1′ is defined as the template parameter ‘T’. Class ‘Fred’ declares ’someOp1′, an operation that takes a parameter whose type is a binding of the template class ‘Bar’. ‘Bar’’s template parameter ‘T’ is bound to the class ‘Foo’. Implicitly, the type of ‘Fred.attr1′ when expanded against Foo should look something like:

class BarOfFoo
    attribute prop1 : Foo;
    operation op11(par1 : Foo);
end;

Note that the expanded class has actually no name, but I am calling it ‘BarOfZoo’ for pedagogical reasons.

Looking closer at the abstractions

• TemplateableElements - abstract super-class for elements that can be declared as templates, or that can bind other templates to a set of parameters. Four kinds of elements can be declared as templates in UML 2.*: Classifier, Operation, Package and StringExpression, and thus only those metaclasses specialize TemplateableElement*.
• ParameterableElements - abstract class that is specialized by any type of element that can be used as parameters to templates.
• TemplateSignature - a template signature is owned by a template element and contains the set of parameters declared by a template element.
• TemplateBinding - a template binding represents the “instantiation” of a template in the form of a directed relationship between a template signature and a a bound element, another templateable element. In addition to tying the template to the bound element through the template’s signature, the template binding contains a set of template parameter substitutions. Which takes us to the next abstraction…
• TemplateParameterSubstitution - a template parameter substitution is created for every template parameter declared by a template signature. It binds an open parameter to an actual parameter, which is a ParameterableElement.

Would you like to play with templates in UML? For now, you will have to look elsewhere. There is some support for templates in the TextUML Toolkit 1.0 M2, but it is, to put it mildly, half-baked. Full template support is planned for M3 (whenever it happens), and that is exactly what I am working right now. I know I am close to getting it right, but assignment compatibility involving template/bound classifiers get be really tricky to implement. Well, whenever I am done, you will learn it first here.
*in the Eclipse UML2 API, Property also specializes TemplateableElement. That seems to be a deviation from the spec, and a bug report has been submited.

A user has just reported that the TextUML Toolkit M2 build did not work at all for him. Ouch. Thankfully, the user included in the bug report all the information I needed to allow a quick diagnostic: turns out he was using a Java 5 VM, and he was getting exceptions like this:

java.lang.UnsupportedClassVersionError: Bad version number in .class file

 at java.lang.ClassLoader.defineClass1(Native Method)

 ...

I then realized that for the M2 build some of the plug-ins were accidentally compiled having Java 6 as the target platform, and thus wouldn’t work on a Java 5 VM. This is being fixed as I write this, and a new M2a build will be available shortly.

So, for those of you running with a Java 5 VM, please make sure you download the TextUML Toolkit M2a build. Otherwise, if you are using a Java 6 VM, you should be fine with M2, no need to download the build again.

Thanks to Jon for reporting the issue. Keep those bug reports coming!

Profiles and stereotypes form a lightweight mechanism for extending the UML metamodel.

A stereotype allows you to tag elements in your model so they can be interpreted differently from ordinary model elements, much like annotations work in languages such as C# and Java. These tags can then be used to drive code generation, for example, so for every class marked as “persistent”, appropriate persistence code should be generated.

A profile is a special kind of package intended to contain stereotype declarations that extend UML to cover some specific domain or platform.

Let’s see how to declare and use profiles and stereotypes with the help of the TextUML Toolkit.

Declaring a stereotype

To declare a stereotype in TextUML, one uses the following syntax:

profile <profile-name>;

stereotype <stereotype-name> extends <metaclass-1 [,...metaclass-n]…]
[<property-1>; [... property-n;]…]
end;

end.

In other words, a stereotype can be declared as applicable to one or more metaclasses (i.e. types of elements in a UML model), and a stereotype can optionally declare properties (more on properties later) . For instance, a classes could be tagged with the <<persistent>> stereotype:

profile business_apps;

import uml;

stereotype persistent extends Class
end;

end.

Or operations could be marked as <<transactional>>, meaning that a transaction will be started whenever the operation starts executing, and finished when its execution ends:

profile business_apps;

import uml;

stereotype transactional extends Operation
end;

end.

Any UML element can be affected by stereotypes, but stereotypes are declared as targetting (potentially multiple) specific element types. For instance, the UML specification has an example of a profile for Enterprise JavaBeans that defines a <<Session>> stereotype for session beans. The<<Session>> stereotype declares a property that allows modelers to define whether the session bean component is stateful or stateless.

profile EJB;

enumeration StateKind
  STATELESS, STATEFUL
end;

stereotype Bean extends uml::Component
end;

stereotype Session specializes Bean
  property kind : StateKind;
end;

stereotype Entity specializes Bean
end;

end.

Applying a stereotype

Now that we know how to declare stereotypes, lets see how to use them. First of all, you must apply the profile defining the stereotypes to the model declaring elements you want to apply stereotypes to:

model bank;

apply business_apps;

/* other model elements here */

end.

You can then attach stereotypes defined in the applied profile to the suitable model elements in your model:

model bank;

apply business_apps;

[persistent]
class Account
    attribute accountNumber : base::String;
    attribute balance : base::Real;
    attribute changes : AccountChange[0,*];
    [transactional] operation withdraw(amount : Real);
    [transactional] operation deposit(amount : Real);
    operation balance() : Real;
    [transactional] operation transfer(other : Account, amount : Real);
end;

end.

In the example above, we applied the <<persistent>> stereotype to the Bank class, and the <<transactional>> stereotype to the withdraw, deposit and transfer operations. In order to have access to these stereotypes, we had to apply the “business_apps” profile to our model.

Conclusion

In this post/article on UML basics using TextUML, we saw how to declare stereotypes and apply them to elements in UML models. We learned that a stereotype must explicitly declare the metaclasses they are applicable to, and that optionally stereotypes might declare properties. Finally, we saw that before a stereotype can be used in a model, the profile declaring the stereotype must be applied to the model.

The primary goal of the EclipseGraphviz project is to support Eclipse-based applications that want to use Graphviz as an easy way of producing non-interactive structured diagrams without requiring the complexity of GEF or GMF.

The TextUML Toolkit is the only application currently known to be based on the EclipseGraphviz project. To support rendering UML models generated using the TextUML textual notation, EclipseGraphviz has now the ability of rendering any UML model generated using the Eclipse UML2 API.

In this screenshot, you can see the UML model used in the article “Getting started with UML2” rendered with EclipseGraphviz:

The main benefits are that the EclipseGraphviz graphical rendering of UML2 models is quite lightweight, and Graphviz produces great layouts automatically. The main caveat is that the diagrams are not interactive. Not to mention that EclipseGraphviz is still quite in its early stages, so it lacks maturity and features. And it does not run yet on non-Windows platforms.

There is where you can make a difference. Check out the project from the Subversion repository, give it a try, and contribute with bug reports, feature requests and patches.

After a few minor bug fixes and improvements to the test build, the M2 milestone of TextUML Toolkit is available. With this milestone you can:

• create UML models using the TextUML textual representation
• visualize the model created using the conventional graphical notation for class diagrams

This milestone does not support non-Windows platforms. But progress is being made in that direction. See below a screenshot of TextUML Toolkit running on Ubuntu Feisty Fawn:

Just go to the download page and fetch it, it is free with no restrictions (or guarantees). Make sure to check out the tutorial. Your feedback is most appreciated.

Now off to a long weekend on the beach…

In the context of providing class diagram visualization for TextUML Toolkit, I have developed a simple graphical viewing framework for Eclipse. It is content type based, and allows you to view anything a content provider has been registered for. For instance, any image file supported by SWT:

But you can also view the graphical representation of a Graphviz DOT file, and it will even update as you edit the file:

And finally, and also the reason why I had to develop support for DOT visualization in the first place, you can also visualize a UML model (here showing a model created using the TextUML Toolkit):

All these features (except for the TextUML Toolkit itself) are part of the EclipseGraphviz project, which is open source (EPL). No releases yet as this is still very new, but you can grab the source from the Subversion repository. If you would like to contribute to the EclipseGraphviz project, or to the graphical viewing framework, your help will be most welcome.

RC

UML has become the lingua franca for modeling applications using the object-oriented paradigm. People use UML in many different ways (see the post on UML modes), ranging from as a communication tool to as a full fledged programming language that supports full code generation. This last way of using UML should puzzle most readers - how can UML models lead to full code generation?

UML has two diagrams that are used for behavior specification: the activity diagram and the state diagram. These two diagrams (more one than the other, depending on the nature of the subsystem being modeled), plus the class diagram (for modeling the structural aspects of the object model) provide the framework that supports the design of complex applications in a way that is fully complete (and thus allowing 100% code generation) while still implementation independent (see earlier post on platform independence). All the other diagrams (use case, sequence, collaboration) are interesting for gathering requirements, but are useless in modeling a solution that can be automatically transformed into a running application, and thus we will ignore them here.

Specifying structure with the Class Diagram

The class diagram is the most well understood of all diagrams in UML. You can model all structural aspects of your object model in the form of classes, attributes, operations and relationships between classes. This specification of structural aspects can then be used for generating (boilerplate) code, database schema, configuration files and so on so forth. This is great already, as that is most of the work. But without including behavioral aspects, it is impossible to do full code generation solely from the class diagram, you are forced to fill the empty methods with handwritten code (unfortunately this is how most vendors expect you to do model-driven development). Still, the class diagram is a fundamental one in which it provides a base framework the other diagrams can build upon.

Specifying dynamics with the State Diagram

The UML state diagram (derived from David Harel’s state charts) allows for a full design of the dynamic aspects of a system. One can model complex state machines using the state diagram, always in the context of a class described in the class diagram. Many mainstream applications do not have any interesting dynamics though, so in those cases the state diagram has limited value. However, in applications for certain industries (such as robotics, telecom and automotive) it is the most important diagram.

Specifying business logic with the Activity Diagram

The most underrated of the UML diagrams, the activity diagram has a key role: it is the only one to allow the modeler to specify behavior in a precise way. The Activity Diagram provides elements (such as actions, pins, data and control flows, signals) that allow specifying the meaning of a behavioral element (such as the body of an operation from the class diagram, or the effect of a state transition from the state diagram).

But no matter how important the UML activity diagram is, it has one strong limitation: it demands too much detailed information in order to be fully defined (and thus actually useful in the context of code generation). Any simple logic that could be written in a few lines in, say, Java, requires a graph with so many nodes that it is virtually impractical to use it with the graphical notation, severely hampering its more widespread adoption.

Have I just suggested that activity diagrams are useless for any serious usage? No! It is just the case that the graphical notation is too cumbersome, and it is not just a problem with the specific choice of symbols - there will never be any graphical representation that can be as expressive and concise for specifying behavior as your programming language of choice (even if your favorite language is as verbose as COBOL). So it is a matter of representation: a textual notation is much more appropriate than a graphical one. The activity diagram itself is fine, thanks.

So what is the textual notation for the activity diagram? There is none. I mean, not one defined by the OMG. Many companies have defined their own action languages (such as Pathfinder AL, Bridgepoint OAL, Kennedy Carter’s ASL) with compilers that provide a textual front-end for the UML activity diagram. TextUML itself has a bigger cousin (currently an ongoing work) that allows specifying UML activities in a way that is familiar to any programmer. Want to see what an action language looks like? Expect a new post on the subject (including our very own action language) here soon.

One notable benefit of model-driven development that is often underrated is improved reuse. This is a direct consequence of appropriate separation of concerns promoted by this development approach. The more intertwined concerns are when addressed in a piece of code, the harder it is to reuse that piece of code. The reason is simple: whenever you tie a solution for one concern to a solution for another concern, you are in trouble: you cannot reuse that piece of code where only one of the concerns is relevant, or the solution for one of the concerns is not appropriate (even if the solution for the other concern is).

Model-driven development promotes an approach where problem-domain concerns are addressed separately from implementation concerns. That means artifacts dealing with problem-domain concerns are free from any specifics on target platforms, and also that artifacts addressing implementation-related concerns are totally unaware of any problem domain knowledge.

That is fantastic, and the reason is twofold:

1. it makes it viable to build a repository of platform-independent problem-domain specific components, likely created by people that are experts in their domain, that can be reused on different target platforms;
2. it allows implementation specialists to code their technology-specific implementation strategies as standalone artifacts (i.e. templates), which can then be shared and reused in applications for the most varied problem domains.

The software industry has been looking for a long time for a way of encapsulating knowledge about specific problem domains in the form of platform-independent software components. Model-driven development with true executable models enables that dream.

For many decades, valuable business logic has been imprisoned into obsolescence-prone implementation-specific artifacts. We are working hard on a product that will help stop this insanity and finally make the dream of truly reusable component repositories a reality.

One key aspect of MDA is platform independence. However, even some of the brightest people in our industry misunderstand what platform independence means in MDA.

Platform independence has a different meaning in MDA than it has, for instance, in Java. Java promotes platform-independence by providing a common environment that insulates the application from platform details such as the instruction set and system API (for instance, for memory allocation, file system manipulation, networking, GUI, threading, etc). The application still has to address all these concerns, but it does that through API and mechanisms that work the same way regardless the actual underlying platform, and thus can run on any platform the Java environment is available. In other words, the Java environment is the platform.

MDA promotes platform-independence by adopting a design-centric approach. Models are removed from implementation related concerns and thus are inherently platform independent: a single design can be reused for building the same system for multiple target platforms. The implementation details are taken care of by target platform specific templates. The templates are applied to the user models then generating concrete platform-specific artifacts (running code, documentation, database schema, configuration files). Differently from Java (even if Martin Fowler says so), MDA does not promote another platform. What it does is to promote a clear separation between problem domain concerns and implementation concerns (as covered before here in the inaugural series entitled “Where we are coming from“).

The benefits of this separation are many: from unprecedented levels of reuse to better opportunities for work specialization. I plan to cover these benefits in detail on future posts.

RC

On his bliki, Martin Fowler eloquently describes the three different modes in which UML can be used: UML as sketch, UML as blueprint and UML as programming language. Let’s revisit the different modes from the perspective of tools for the job.

UML as sketch

Description: In this mode, UML is essentially a tool for conceiving and communicating ideas between team members. As such, there is no need for completeness or validity of the object models, and actually any information not essential for communicating the idea at hand is intentionally omitted for conciseness and clarity. UML as sketch is increasingly popular, even at shops where model-driven development is considered an abomination.

Tools for the job: developers don’t need special tools for using UML as a sketching tool, paper and pen or a whiteboard are great. The only flaws are when you would like to archive a drawing or send it to others by email. In that case, Visio (or any other generic diagram editor) is a common choice. Another less known option is Whiteboard Photo, that allows you to take a snapshot of your hand drawings (on a whiteboard or on paper) and have them automatically translated into great looking clean 2-D charts.

UML as blueprint

Description: in this mode, the focus is on having UML models used as an input to the implementation phase, and thus need to be valid and complete from the point of view of structure. Behavioral aspects are described in textual form (such as in use case descriptions) or by means of diagrams depicting scenarios (such as sequence or collaboration diagrams) and are inherently informally and incompletely described. The models can be submitted as input to code generation but the generated code has to be enhanced by hand to cover the behavioral aspects.

Tools for the job: generic tools won’t cut it here - you will need diagram editing tools that support all (or most) of UML diagrams, and (if you are taking the code generation route) persist your models in a format that your code generation tools can understand. Most UML tools out there support (or intend to support) this.

The TextUML Toolkit we provide supports this mode too. Your class diagrams are fully verified, and are persisted using a standard format.

UML as programming language

Description: In this mode, UML models must be complete both structurally and behaviorally as they must be simulatable and serve as input to code generation (by the way, Fowler really misses the point when he argues that a graphical language such as UML is not appropriate for fine grained behavior specification - who said UML is graphical anyway?).

Tools for the job: there are only a few tools currently in the market that support this mode, they tend to target embedded software development, and I bet you will not find how much they cost on their websites unless you call them.

Corcova Libra will support this, and when made available will visibly sport a reasonably fair price on our web site. Also, Libra will aim at mainstream business application developers, instead of focusing on specific vertical markets.

Conclusion

UML as sketch is cool and useful, but from the point of view of software engineering (our focus here) is meaningless.

UML as blueprint is increasingly practiced in shops where (partial) code generation has been adopted. It has benefits over writing the entire code by hand, but it still requires all the interesting code to be written manually, and there are a lot of issues with that.

Finally, UML as programming language (in other words, full blown model-driven development) is the most interesting of the three modes, even if there is a lot of skepticism and misinformation towards it. I will talk about that in an upcoming post.

RC

Even though a textual notation such as TextUML is much more productive when creating UML models, the graphical notation is still better for a high-level overview. So, for the next milestone of the TextUML Toolkit, the main feature planned is a model visualizer using the conventional graphical notation of UML.

The first option considered in implementing that feature was the UML2 Tools project, part of the Eclipse.org Model Development Tools project. UML2 Tools supports viewing and editing most of UML diagram types, and works seamlessly with models created using UML2 and EMF, which the TextUML Toolkit already uses, and that is great. However, it also requires GEF and GMF, what is bad because it significantly increases the download size for the TextUML Tookit (currently at 31MB), already high for the value it currently provides. Also, from some initial experimentation, it seems to be very computation and/or memory intensive. Because of that, it was decided a more lightweight alternative should be found, even more so if you consider interactivity and support for diagrams other than the class diagram are very low on our list.

Enter Graphviz and UMLGraph

Graphviz is an awesome tool for graph visualization. Graphviz takes simple text files (in the dot notation) describing the graph structure and produces great looking diagrams.

However, Graphviz is very generic, and does not know anything about the UML notation. That is when UMLGraph comes into play. UMLGraph produces dot files that generate OMG compliant UML diagrams (like the ones showing here).

So the direction is to adopt Graphviz and UMLGraph into a lightweight, non-interactive graphical viewer for UML class diagrams. We don’t dismiss having a more powerful graphical viewer (and editor) in the future, probably based on the UML2 Tools project, but the focus now is on simplicity.

There are two difficulties into adopting Graphviz and UMLGraph though:

1. Graphviz is a native standalone tool, not a Java API
2. UMLGraph, as a Java doclet, supports only one input representation: Java source with Javadoc comments

Turns out the first obstacle can be easily overcome by wrapping the Graphviz binaries into platform-specific fragments and providing a simple Java API on top of it.

The second obstacle is a bit trickier though. UMLGraph ties the usage of the doclet API into the generation of output in dot form. Breaking that coupling requires a non-trivial refactoring effort. I have been talking to UMLGraph author Diomidis Spinellis and he is keen on the decoupling, and that is the route we are going to take.

Would you like to see what it looks like? Make sure to check out the M2 milestone coming out in mid-June.

RC

There hasn’t been a post here for a while, but there is a good reason for that: we are glad to announce that we now have an actual web site (go see for yourself) and have simultaneously made the first public release! TextUML Toolkit 1.0 M1 is the first milestone of the tool we have developed for creating UML models in a way that is more natural to hard-core developers. If you want to know more, go read the overview, download the tool and try out the tutorial.

After a long period of silence, I almost pity that I have somehow crammed too many great news on a single post…

RC