|The use of the term "persistence" here refers to page-level persistence, NOT database persistence.|
Most instance variables in Tapestry are automatically cleared at the end of each request. This is important, as it pertains to how Tapestry pages are shared, over time, by many users.
However, you often want to store some data on a single page, and have access to it in later requests to that same page, without having to store it in a database between requests. (To store values across multiple pages, see Session Storage.)
Making page data persist across requests to a single page is accomplished with the @Persist annotation. This annotation is applied to private instance fields of components:
Such annotated fields will retain their state between requests. Generally, speaking, this means that the value is stored into the session (but other approaches are possible).
Whenever you make a change to a persistent field, its value is saved. On later requests to the same page, the value for the field is restored.
The value for each field is the strategy used to store the field between requests.
The session strategy stores field changes into the session; the session is created as necessary.
A suitably long session attribute name is used; it incorporates the name of the page, the nested component id, and the name of the field.
Session strategy is the default strategy used unless otherwise overridden.
The flash strategy stores information in the session as well, just for not very long. Values are stored into the session, but then deleted from the session as they are first used to restore a page's state.
The flash is typically used to store temporary messages that should only be displayed to the user once.
The field is persisted onto the client; you will see an additional query parameter in each URL (or an extra hidden field in each form).
Client persistence is somewhat expensive. It can bloat the size of the rendered pages by adding hundreds of characters to each link. There is extra processing on each request to de-serialize the values encoded into the query parameter.
Client persistence does not scale very well; as more information is stored into the query parameter, its length can become problematic. In many cases, web browsers, firewalls or other servers may silently truncate the URL which will break the application.
Use client persistence with care, and store a minimal amount of data. Try to store the identity (that is, primary key) of an object, rather than the object itself.
By default the value for the Persist annotation is the empty string. When this is true, then the actual strategy to be used is determined by a search up the component hierarchy.
For each component, the meta-data property tapestry.persistence-strategy is checked. This can be specified using the Meta annotation.
If the value is non-blank, then that strategy is used. This allows a component to control the persistence strategy used inside any sub-components (that don't explicitly use a different strategy).
In any case, if no component provides the meta data, then the ultimate default, "session", is used.
Fields marked with @Persist may not have default values (whether set inline, or inside a constructor).
If you reach a point where you know that all data for a page can be discarded, you can do exactly that.
The method discardPersistentFieldChanges() of ComponentResources will discard all persistent fields for the page, regardless of which strategy is used to store the property. This will not affect the page in memory, but takes effect for subsequent requests.
The Servlet API was designed with the intention that there would be only a modest amount of server-side state, and that the stored values would be individual numbers and strings, and thus, immutable.
However, many web applications do not use the HttpSession this way, instead storing large, mutable objects in the session. This is not a problem for single servers, but in a cluster, anything stored in the session must be serialized to a bytestream and distributed to other servers within the cluster, and restored there.
Most application servers perform that serialization and distribution whenever HttpSession.setAttribute() is called. This creates a data consistency problem for mutable objects, because if you read a mutable session object, change its state, but don't invoke setAttribute(), the changes will be isolated to just a single server in the cluster.
Tapestry attempts to solve this: any session-persisted object that is read during a request will be re-stored back into the HttpSession at the end of the request. This ensures that changed internal state of those mutable objects is properly replicated around the cluster.
But while this solution solves the data consistency problem, it does so at the expense of performance, since all of those calls to setAttribute() result in extra session data being replicated needlessly if the internal state of the mutable object hasn't changed.
Tapestry has solutions to this, too:
Tapestry knows that Java's String, Number and Boolean classes are immutable. Immutable objects do not require a re-store into the session.
You can mark your own session objects as immutable (and thus not requiring session replication) using the ImmutableSessionPersistedObject annotation.
The OptimizedSessionPersistedObject interface allows an object to control this behavior. An object with this interface can track when its mutable state changes. Typically, you should extend from the BaseOptimizedSessionPersistedObject base class.
The SessionPersistedObjectAnalyzer service is ultimately responsible for determining whether a session persisted object is dirty or not (dirty meaning in need of a restore into the session). This is an extensible service where new strategies, for new classes, can be introduced.