<!-- Download Links -->
Note: after downloading be sure to consult ReadMe.txt first.
<!-- Add the rest of your HTML here -->
This sample is an introduction to Rockford Lhotka's Component-based Scalable
Logical Architecture (CSLA) as described in his book, Visual Basic 6 Business
Objects (Wrox Press, 1998). The CSLA is
an implementation of n-tier design, where an application is logically
partitioned into four layers - presentation, UI-centric business objects,
data-centric business objects, and data services. The purpose of the
architecture is to achieve two overarching objectives.
Flexibility in the physical placement of each layer.
The latter means that we can decide whether to place all the layers on a single
machine or to distribute them between client and server machines with little or
no impact on the business logic code. It also means that we can provide a
variety of presentation layers, ranging from a traditional Windows interface to
a Web interface, with little or no impact on the business logic code.
A typical physical architecture, utilizing a rich Windows-style user interface,
is to place the UI- centric objects on a client workstation where they are
close to the user interface. The UI- centric objects communicate with the
data-centric objects, which are placed on an application server, and the latter
communicate with the data services, which are on a database server.
The code presented here in C# is not a full implementation of the architecture.
Rather, it serves to illustrate the basic ideas. It presents a single
UI-centric business object that captures basic information about a person -
social security number, name and date of birth. It shows how a Microsoft .Net
Windows Forms user interface can interact with the business object while
allowing the latter to validate all the business rules as the user enters data.
The user interface is completely insulated from detailed business rule
validation. It just needs to act on a property of the business object - whether
it is currently in a valid or invalid state.
UML Class Diagram
The Person Class
The Person class has three supplied properties:
Social Security Number
and one computed property:
It has three business rules.
Social Security Number contains exactly 11 alphanumeric characters (though the
sample asks the user for digits).
Name must be non-empty.
Birth Date must be in a valid format.
A Person object is not considered to be in a valid state until these rules are
In order to provide the richest and most responsive user experience the idea is
to validate these rules while the user is entering their data, rather than
after it has all been entered. Other user interfaces, such as an HTML 3.2
interface, can make do with a "batch"-style interface but we want to be able to
use a variety of user interfaces and a Windows-style GUI provides the richest.
The basic idea is that the controls for saving the user's data should only
become available when the user has supplied all the data necessary to make a
Person valid. It is the user interface developer's job to write the code to do
this. But they don't need to write detailed validation code they need only
query the Person's "Valid" property.
However, in practice, the UI developer has to know more about a Person than
this. But what they have to know is not to do with business rules but with the
operational state of a Person at any time.
Let's consider the operations that can be performed on a Person:
Add a new Person.
Save it to a database.
Load an existing Person from the database.
Delete a Person.
A rich user interface will have features that enable a user to:
Edit their data, save their changes and close the application.
Edit their data, save their changes, do further editing, save and close.
Edit their data, cancel their changes, and close.
And so on.
In a data entry form the above is typically implemented using OK, Cancel and
Apply buttons. This might not seem particularly problematic. However, a feature
of this architecture is that the Person object's state is being updated as the
user types. If the user cancels their edits it is important to restore the
object to the last valid state it was in prior to the edits. This means we have
to keep a copy of the current state after each Apply operation.
In this particular example, this doesn't matter but in a more complex
application, say, with collections of Persons, it would. And we are trying to
illustrate the general principles.
The UI developer also needs to know a few other things about a Person, such as
whether a Person is new, modified (dirty), or is currently being edited.
For example, it should not be possible to load a new Person from the database
while one is being edited.
To facilitate editing, three methods are provided:
This enables editing.
Saves or deletes object if appropriate and terminates editing. Clients should
immediately call BeginEdit if they wish to continue editing.
Cancels all changes since the last ApplyEdit operation or since the object was
marked for deletion. Terminates editing.
In database terminology these three methods function like "BeginTrans",
"Commit" and "Rollback" respectively, but applied to an object rather than a
Managing Business Rules
In Rockford Lhotka's original Visual Basic implementation he created a
BrokenRules class that was made available to every business object in the
application. In this sample, I have moved this functionality into an abstract
base class, BusinessObject. I could also have placed an abstract interface to
the three editing methods here while deferring the implementation to Person,
but I chose not to for this sample.
The validity or invalidity of a Person object is determined by maintaining a
collection of broken business rules. The object is invalid while the collection
is non-empty. When the count goes to zero the object becomes valid.
The method that maintains the collection is:
void RuleBroken(string rule, bool isBroken)
rule is a description of the business rule. This can be as simple
as just the name of the property, e.g., "Birth Date"
isBroken indicates whether the rule is broken or not.
When a rule becomes broken it's added to the collection. If it's repeatedly
broken the algorithm just skips adding it. When it becomes unbroken it is
removed from the collection.
When a Person object is first constructed all its fields will be empty, so all
its rules will be broken. Remember, the rules are:
Social Security Number contains exactly 11 alphanumeric characters.
Name must be non-empty.
Birth Date must be in a valid format.
In the Person constructor we write:
RuleBroken("Social Security Number", true);
RuleBroken("Birth Date", true);
Of course, this is not normally how we should construct objects. Normally
objects should always be constructed in a valid state. However, the object
being discussed here is a special kind. And when anything important is being
done to it, such as persisting it or restoring it, the method ensures that the
object is valid before it's persisted or after it's restored.
Making a Person Valid
The idea is that a Person should move from an invalid to a valid state as the
user types in their data. Initially each item that is the subject of a business
rule will be incorrect. Then one by one, as the user completes a data entry
field, each should become valid until Person becomes valid.
This is achieved by appropriately constructed Person set_xxx properties and
Form TextChanged events.
So, for social security number we have:
Person.set_SocialSecurityNumber property. (simplified for
socialSecurityNumber = value;
RuleBroken("Social Security Number", socialSecurityNumber.Length != 11);
person.SocialSecurityNumber = txtSocialSecurityNumber.Text;
The TextChanged event occurs on each keystroke and for each keystroke, the
Person set_SocialSecurityNumber property is called which calls RuleBroken.
Note the expression:
socialSecurityNumber.Length != 11.
This describes as true the condition that makes the rule broken.
On the first keystroke social security number is not 11 characters long so it
is added to the broken rules collection. On subsequent keystrokes it remains
broken and so
RuleBroken will skip adding it to the collection.
But when the 11th character is typed
RuleBroken receives a false
value for its argument, indicating that the rule is no longer broken, so it is
removed from the collection.
This procedure is repeated for the name and birth date properties until the
Person object becomes valid.
Informing a Client that a Person is Valid or Invalid
The Person class exposes an IsValid property but also exposes Valid and Invalid
custom events. These two events are handled by the client, in this case a
Windows Forms application, to enable or disable saving of a Person object. This
means enabling or disabling the OK and Apply buttons.
To fire the events the Person class (actually the base class) declares an event
handler delegate and two events.
public delegate void EventHandler(object sender, EventArgs e);
public event EventHandler OnInvalid;
public event EventHandler OnValid;
To fire the
if (OnValid != null)
OnValid (this, EventArgs.Empty);
RuleBroken executes this code when the count of broken business
rules falls to zero.
The client handles the event as follows:
person.OnValid += new Person.EventHandler(Person_OnValid);
private void Person_OnValid(object sender, System.EventArgs e)
btnOK.Enabled = true;
btnApply.Enabled = true;
This event handler code is simplified compared to the actual implementation.
But it illustrates the general idea.
OnInvalid event is implemented similarly.
The Age Property
The age property is interesting. It is exposed as read-only so the client
cannot set it. Instead, it is computed from the birth date. However, it can be
exposed to the user interface as a dynamic property that is updated each time
the user enters a different birth date.
(Note: The sample application expects a date in UK format - day/month/year
Suppose the user enters 11/1/1969 for date of birth. This gives an age of 33.
The idea is to have 33 displayed to them in a read-only field the instant they
type the 9 for 1969. If they make a mistake and decide to change, say, the year
to 1959 the age field should immediately update itself to show 43. This
behaviour can be achieved by again defining an event in the Person class:
public event EventHandler OnNewAge;
In the Person object, once the birth date is valid this event is raised and it
is raised again on each subsequent occasion that it becomes valid, if the new
age differs from the previous valid age.
The client handles the event as follows:
private void Person_OnNewAge(object sender, System.EventArgs e)
lblAge.Text = person.Age.ToString();
Persistence is implemented using a "manager" object, PersonManager. This has
Load, Save and Delete methods and communicates with the Person object. The data
access is via ADO.NET and a Microsoft Access database.
The interface between the two is very simplistic. In a production environment
we would need to use .Net remoting and serialization, as the PersonManager
class would typically reside on a server.
Also, later in his book, Lhotka streamlines his concepts and refers to Person
and PersonPersist objects to refer to the UI-centric and data-centric halves of
the business object.
As indicated, the sample employs a Windows Forms user interface. It is not
intended to win any prizes for user interface design. Its purpose is to
illustrate the principal concepts discussed above.
Editing a new Person
The screenshot below shows an incompletely edited Person. The birth year is in
two-digit format but it needs to be four digits, so the OK and Apply buttons
Once the mistake is corrected the Person object becomes valid and the OK and
Apply buttons are enabled. Note also that the age field is now filled in.
Loading a new Person
A Person is loaded using their social security number. The screenshot below
shows an incomplete social security number. It needs to be 11 digits long, so
the Load button is disabled. The name and birth date fields are also disabled,
as we're not in edit mode.
Once the 11th digit is added the Load button is enabled.
Once the Person is loaded the social security number field itself becomes
But if we now uncheck the check box to enable editing, all fields are enabled
and we can also delete the Person at this point because it is no longer new.
An ASP.NET Client
I have not included the code but, for illustration, the screenshots below show
an ASP.NET Web Forms implementation. This demonstrates a "batch"-style user
interface. Validation does not take place until all the fields have been
entered and the user clicks one of the Save buttons. Instead of validation
being done in the TextChanged events, the ASP.NET required field and custom
validators are used. These are invoked when a Submit button, i.e., one of the
Save buttons, is clicked. The first screenshot shows an invalid date with an
error message. This is corrected in the second screenshot.
The UI was a bit trickier to implement due to the statelessness of a web
application. For example, after a submit command the form is reloaded and this
invalidates the current state; the Loading checkbox must have AutoPostback=true
and so on. However, the essential point is that no changes are required to the