the "facade" is exactly what ofbiz wants to avoid.
the extra overhead of the conversion.
Marc Morin sent the following on 12/10/2010 8:17 AM:
This isn't Object->Relational mapping but Relational->Object mapping. It doesn't change
ANY of the obiz entity semantics, it just provides a facade over the entities that can
optionally be used by Java code (not framework code, but application layer) to recognize that
the business logic IS tightly bound to entity (by name), fields (by name) and relationships
between entities (by name). THis facade just moves these already existing bindings from
"strings" to Java native mechanisms, so that you get compile time errors, if you
rename an entity, field, type, relation, etc...
BUT most importantly, you don't get a bunch of GenericValue's and
List<GenericValue> in the java application code. They become meaningful
business entities that allows code easier to follow and thus less bugs and easier to
maintain.
Marc Morin
Emforium Group Inc.
ALL-IN Software
519-772-6824 ext 201
mmo...@emforium.com
----- Original Message -----
I have thought about this idea, and at first glance it seems cool and
all, but one of the selling points of OFBiz is that it intentionally
stays away from this type of Object->Relational mapping. OFBiz
developers find that direct access to the relational database is
easier to use.
-Adrian
On 12/10/2010 7:20 AM, Marc Morin wrote:
In the spirit of changing the entity/delegator interface more object
friendly, why not take this to then next step and generate POJO
interfaces for each entity. These would extend GenericValue but
provide a simple gettor/settor facade allowing compile time type
checking and removing of the "string" code for much of the business
logic written in java.
We have done such a thing (again in our forked application), and it
makes the Java code much more readable and easier to use. The
general structure is
public class Person extends AbstractGenericValue<Person>
{
public static final String ENTITY = "Person";
// constructor, only called from makeValue, MUST be associated
with a delegator
protected Person(Delegator delegator) {...}
// factory method
public static Person newInstance(Delegator delegator) {...}
// generate finders, by pkey, etc...
public static Person findOne(Delegator delegator, String
partyId){...}
// getter and settors
public String getFirstName() {
return getString("firstName");
} public Person setFirstName(String value) {
set("firstName", value);
return this;
}
// relationships
public Party getParty() throws GenericEntityException {...}
public PartyNameView getPartyNameView() throws
GenericEntityException{...}
}
This allows code that is much easier to debug and less error prone..
example below is for navigating orders.
OrderHeader orderHeader = OrderHeader.findOne(delegator, orderId);
// get the orderItems
List<OrderItem> orderItems = orderHeader.getOrderItemList();
BigDecimal totalQuantity = BigDecimal.ZERO;
for (OrderItem orderItem: orderItems) {
totalQuantity = totalQuantity.add(orderItem.getQuantity());
}
I know we want to encourage business logic in minlang, etc... but if
it is written in java, and there is a LOT of code in java, shopping
cart, etc... this makes that code MUST more readable and
maintainable. The binding between the entity model and the java
implementation can be caught as a compile time error...
significantly lowers the maintenance cost of the code.
This may be pushing a rope, but we use this ALL the time for our
groovy and java code. (would also apply to jsp code obviously).
Minilang code can be type checked by the reader... (want to check
for static errors in code, without the need to "run" the code).
We have implemented the generators, and the refactoring/abstracting
to enable this. We find it works great and doesn't break ANY of the
nice ofbiz extend entity semantics, etc.... Of course if you extend
an entity and then want java business logic to use it... you need to
access those items either with "strings" as stock ofbiz, or redo an
entity-gen. But if there is no java code using the entities, no need
to auto-gen.
As another note, we have done a similar thing with the service
interface.... as you might have guessed, we're a fan of ofbiz
extensibility, but NOT on how it encourages poor Java implementation
practices. ("String" object references, non-type safe, public static
methods everywhere.... etc...)
Marc
