How JSF State Management works

How JSF State Management Works

Overview

StateManager manages state, but you, as an end-user, have to give it hints about what state needs to be kept. That's either via implementing Serializable (and correct use of transient is part of implementing Serializable) or by implementing StateHolder.

Data is stored in two different ways in JSF: in scoped beans or in the component tree. The scope of scoped beans are hopefully self-explanatory. The state of the components themselves are stored in the response, and then restored when a request arrives. This effectively gives data stored by the components a "page" scope since they exist so long as the page doesn't change. Note that components store value bindings and method bindings (#{} – el expressions) as string literals, so the backing beans they point to are not stored in the component tree at page scope.

Tomahawk has a SaveState component named <t:saveState> that allows you to store data (including entire backing beans) as part of the component tree, effectively making the page-scoped (or more since you can preserve such values across pages. In JSF 2.0, a new scope called @ViewScope was added to provide the same functionality as t:saveState. You still need to properly implement Serializable or StateHolder on your backing beans.

Backtracking

Note that StateManager is pluggable, so it's possible to implement your own alternatives to the standard state management. Thus you could implement a strategy that allowed backtracking.

Another alternative to handling backtracking is to use client-side state saving and avoid the use of session-scoped data. This also requires resyncing any other persisting backing stores (i.e. databases) when the page is reloaded.