Home > Guides > Core Developers Guide > Configuration Elements > Action Configuration |
The action mappings are the basic "unit-of-work" in the framework. Essentially, the action maps an identifier to a handler class. When a request matches the action's name, the framework uses the mapping to determine how to process the request.
The action mapping can specify a set of result types, a set of exception handlers, and an interceptor stack. Only the name
attribute is required. The other attributes can also be provided at package scope.
In a web application, the name
attribute is matched as part of the location requested by a browser (or other HTTP client). The framework will drop the host and application name and the extension and match what's in the middle: the action name. So, a request for http://www.planetstruts.org/struts2-mailreader/Welcome.do
will map to the Welcome
action.
Within an application a link to an action is usually generated by a Struts Tag. The tag can specify the action by name, and the framework will render the default extension and anything else that is needed. Forms may also submit directly to a Struts Action name (rather than a "raw" URI).
The default entry method to the handler class is defined by the Action interface.
Implementing the Action interface is optional. If Action is not implemented, the framework will use reflection to look for an execute
method.
Sometimes, developers like to create more than one entry point to an Action. For example, in the case of a data-access Action, a developer might want separate entry-points for create
, retrieve
, update
, and delete
. A different entry point can be specified by the method
attribute.
If there is no execute
method and no other method specified in the configuration the framework will throw an exception.
Many times, a set of action mappings will share a common pattern. For example, all your edit
actions might start with the word "edit", and call the edit
method on the Action class. The delete
actions might use the same pattern, but call the delete
method instead.
Rather than code a separate mapping for each action class that uses this pattern, you can write it once as a wildcard mapping.
Here, a reference to "editCrud" will call the edit
method on an instance of the Crud Action class. Likewise, a reference to "deleteCrud" will call the delete
method instead.
Another common approach is to postfix the method name and set it off with an exclamation point (aka "bang"), underscore, or other special character.
To use a postfix wildcard, just move the asterisk and add an underscore.
From the framework's perspective, a wildcard mapping creates a new "virtual" mapping with all the same attributes as a conventional, static mapping. As a result, you can use the expanded wildcard name as the name of validation, type conversion, and message resource files, just as if it were an Action name (which it is!).
Crud_input-validation.xml
Crud_delete-conversion.xml
There's a feature embedded in WebWork 2 that lets the "!" (bang) character invoke a method other than execute
. In WebWork, it doesn't really have a name. During the S2 discussions, we coined the term "dynamic method invocation" to describe how WW/S2 use the bang notation.
Dynamic Method Invocation (DMI) will use the string following a "!" character in an action name as the name of a method to invoke (instead of execute
). A reference to "Category!create.action
", says to use the "Category" action mapping, but call the create
method instead.
For Struts 2, we added a switch to disable DMI for two reasons. First, DMI can cause security issues if POJO actions are used. Second, DMI overlaps with the Wildcard Method feature that we brought over from Struts 1 (and from Cocoon before that). If you have security concerns, or would like to use the "!" character with Wildcard Method actions, then set struts.enable.DynamicMethodInvocation
to FALSE
in the application configuration.
The framework does support DMI, just like WebWork 2, but there are problems with way DMI is implemented. Essentially, the code scans the action name for a "!" character, and finding one, tricks the framework into invoking the other method instead of execute
. The other method is invoked, but it uses the same configuration as the execute
method, including validations. The framework "believes" it is invoking the Category
action with the execute
method.
The Wildcard Method feature is implemented differently. When a Wildcard Method action is invoked, the framework acts as if the matching action had been hardcoded in the configuration. The framework "believes" it's executing the action Category!create
and "knows" it is executing the create
method of the corresponding Action class. Accordingly, we can add for a Wildcard Method action mapping its own validations, message resources, and type converters, just like a conventional action mapping. For this reason, the Wildcard Method is preferred.
In Struts 2.3, an option was added to restrict the methods that DMI can invoke. First, set the attribute strict-method-invocation="true"
on your <package>
element. This tells Struts to reject any method that is not explicitly allowed via either the method
attribute (including wildcards) or the <allowed-methods>
tag. Then specify <allowed-methods>
as a comma-separated list of method names in your <action>
. (If you specify a method
attribute for your action, you do not need to list it in <allowed-methods>
.)
Note that you can specify <allowed-methods>
even without strict-method-invocation
. This restricts access only for the specific actions that have <allowed-methods>
.
If the class attribute in an action mapping is left blank, the com.opensymphony.xwork2.ActionSupport
class is used as a default.
The ActionSupport class has an execute
method that returns "success" and an input
method that returns "input".
To specify a different class as the default Action class, set the default-class-ref
package attribute.
For more about using wildcards, see Wildcard Mappings.
A good practice is to link to actions rather than pages. Linking to actions encapsulates which server page renders, and ensures that an Action class can fire before a page renders.
Another common workflow stategy is to first render a page using an alternate method, like input
and then have it submit back to the default execute
method.
Using these two strategies together creates an opportunity to use a "post-back" form that doesn't specify an action. The form simply submits back to the action that created it.
Usually, if an action is requested, and the framework can't map the request to an action name, the result will be the usual "404 - Page not found" error. But, if you would prefer that an omnibus action handle any unmatched requests, you can specify a default action. If no other action matches, the default action is used instead.
There are no special requirements for the default action. Each package can have its own default action, but there should only be one default action per namespace.
Using wildcards is another approach to default actions. A wildcard action at the end of the configuration can be used to catch unmatched references.
When a new action is needed, just add a stub page.
It's important to put a "catchall" wildcard mapping like this at the end of your configuration so it won't attempt to map every request!