To Implement KVC validattion method, override -(BOOL)validateIvar:error: or -validateValue:forKey:error:?

The former is described in the KVC Prog Guide, but missing from iOS API installed with Xcode 4.0.2 (with iOS 4.3 SDK) for Snow Leopard which lists the l开发者_如何学运维atter. This is for iOS 2.0 or newer. Thanks.

You can use both. For key named foo, you can either implement a specific method

-(BOOL)validateFoo:(id*)ioValue error:(NSError**)error {
        ...
}

or, a non-specific method

- (BOOL)validateValue:(id *)ioValue forKey:(NSString *)key error:(NSError **)outError {
      if([key isEqualToString:@"foo"]){
           ...
      }else{
          return [super validateValue:ioValue forKey:key error:outError];
      }
  }

It is perfectly sensible that you didn't find validateIvar:error: in the API documentation, because the part Ivar needs to be changed for each key. You don't find every method called setFoo: for keys foo you defined in the system API reference, do you?

You should only implement methods following the pattern validate<Key>:error: (which can be done even in a category). Don't override validateValue:forKey:error:, and do not call validateValue:forKey:error: from within a validation method. You will cause an infinite loop. This is covered in the Key-Value Coding documentation and Core Data documentation, as well as within the Foundation headers. validateValue:forKey:error: is the designated entry point into validation, both for Frameworks like CoreData and your applications.

Code that wishes to perform validation should invoke validateValue:forKey:error: rather than the custom validation method directly.

For example:

if ([[self model] validateValue:&text forKey:@"name" error:&validationError]){
    [[self model] setValue:text forKey:@"name"];
} else {
    [self didFailWithError:validationError];
}

This should not be done from within an accessor method. An action method is the perfect place to do this.

A working example can be seen here