Some time ago I wrote an article on Adaptive Bindings showing how
the pageDef for a an ADF UI does not have to be wedded to a fixed data control or collection / View Object. This article has proved pretty popular, so as a follow up I wanted to cover another "Adaptive" feature of your ADF applications,
the ability to make multiple different connections from an Application Module, at runtime.
Now, I'm sure you'll be aware that if you define your application to use a data-source rather than a hard-coded JDBC connection string, then you have
the ability to change
the target of that data-source after deployment to point to a different database. So that's great, but
the reality of that is that this single connection is effectively fixed within
the application right? Well no, this it turns out is a common misconception.
To be clear, yes a single instance of an ADF Application Module is associated with a single connection but there is nothing to stop you from creating multiple instances of
the same Application Module within
the application, all pointing at different connections. If fact this has been possible for a long time using a custom extension point with code that which extends oracle.jbo.http.HttpSessionCookieFactory. This approach, however, involves writing code and no-one likes to write any more code than they need to, so, is there an easier way? Yes indeed. It is in fact a little publicized feature that's available in all versions of 11g,
the ELEnvInfoProvider.
What Does it Do?
The ELEnvInfoProvider is a pre-existing class (the full path is oracle.jbo.client.ELEnvInfoProvider) which you can plug into your ApplicationModule configuration using
the jbo.envinfoprovider property. Visuallty you can set this in
the editor, or you can also set it directly in
the bc4j.xcfg (see below for an example) .
Once you have plugged in this envinfoprovider, here's
the fun bit, rather than defining
the hard-coded name of a datasource instead you can plug in a EL expression for
the connection to use. So what's
the benefit of that? Well it allows you to defer
the selection of a connection until
the point in time that you instantiate
the AM.
To define
the expression itself you'll need to do a couple of things:
First of all you'll need a managed
bean of some sort – e.g. a sessionScoped
bean defined in your ViewController project. This will need a getter method that returns
the name of
the connection. Now this connection itself needs to be defined in your Application Server, and can be managed through Enterprise Manager, WLST or through MBeans. (You may need to read
the documentation [http://docs.oracle.com/cd/E28280_01/web.1111/b31974/deployment_topics.htm#CHDJGBDD] here on how to configure connections at runtime if you're not familiar with this)
The EL expression (e.g. ${connectionManager.connection} is then defined in
the configuration by editing
the bc4j.xcfg file (there is a hyperlink directly to this file on
the configuration editing screen in
the Application Module editor). You simply replace
the hardcoded JDBCName value with
the expression. So your cfg file would end up looking something like this (notice
the reference to
the ELEnvInfoProvider that I talked about earlier)
<BC4JConfig version="11.1" xmlns="http://xmlns.oracle.com/bc4j/configuration">
<AppModuleConfigBag ApplicationName="oracle.demo.model.TargetAppModule">
<AppModuleConfig DeployPlatform="LOCAL"
JDBCName="${connectionManager.connection}"
jbo.project="oracle.demo.model.Model"
name="TargetAppModuleLocal"
ApplicationName="oracle.demo.model.TargetAppModule">
<AM-Pooling jbo.doconnectionpooling="true"/>
<Database jbo.locking.mode="optimistic">
<Security AppModuleJndiName="oracle.demo.model.TargetAppModule"/>
<Custom jbo.envinfoprovider="oracle.jbo.client.ELEnvInfoProvider"/>
</AppModuleConfig>
</AppModuleConfigBag>
</BC4JConfig>
Still Don't Quite Get It?
So far you might be thinking, well that's fine but what difference does it make if
the connection is resolved "just in time" rather than up front and changed as required through Enterprise Manager?
Well a trivial example would be where you have a single application deployed to your application server, but for different users you want to connect to different databases. Because,
the evaluation of
the connection is deferred until you first reference
the AM you have a decision point that can take
the user identity into account.
However, think about it for a second. Under what circumstances does a new AM get instantiated? Well at
the first reference of
the AM within
the application yes, but also whenever a Task Flow is entered - if
the data control scope for
the Task Flow is ISOLATED. So
the reality is, that on a single screen you can embed multiple Task Flows, all of which are pointing at different database connections concurrently.
Hopefully you'll find this feature useful, let me know...
