While evaluating an expression, the ELResolver associated with the ELContext is consulted to do the initial resolution of the first variable of an expression. It is also consulted when a . or [] operator is encountered, except for the last such operator in a method expression, in which case the resultion rules are hard coded.

For example, in the EL expression ${employee.lastName}, the ELResolver determines what object employee refers to, and what it means to get the lastName property on that object.

Most methods in this class accept a base and property parameter. In the case of variable resolution (e.g. determining what employee refers to in ${employee.lastName}), the base parameter will be null and the property parameter will always be of type String. In this case, if the property is not a String, the behavior of the ELResolver is undefined.

In the case of property resolution, the base parameter identifies the base object and the property object identifies the property on that base. For example, in the expression ${employee.lastName}, base is the result of the variable resolution for employee and property is the string "lastName". In the expression ${y[x]}, base is the result of the variable resolution for y and property is the result of the variable resolution for x.

Though only a single ELResolver is associated with an ELContext, there are usually multiple resolvers considered for any given variable or property resolution. ELResolvers are combined together using CompositeELResolver s, to define rich semantics for evaluating an expression.

For the getValue(ELContext, Object, Object) , getType(ELContext, Object, Object) , setValue(ELContext, Object, Object, Object) and isReadOnly(ELContext, Object, Object) methods, an ELResolver is not responsible for resolving all possible (base, property) pairs. In fact, most resolvers will only handle a base of a single type. To indicate that a resolver has successfully resolved a particular (base, property) pair, it must set the propertyResolved property of the ELContext to true. If it could not handle the given pair, it must leave this property alone. The caller must ignore the return value of the method if propertyResolved is false.

The getFeatureDescriptors(ELContext, Object) and getCommonPropertyType(ELContext, Object) methods are primarily designed for design-time tool support, but must handle invocation at runtime as well. The isDesignTime() method can be used to determine if the resolver is being consulted at design-time or runtime.