This change splits the insertObjectAtArrangedObjectIndex:, addObject:, and removeObject: tests into two versions: one for a regular AC setContent: setup and one for when the AC is bound to an array-like destination implementing KVC.
CPArray insertObject:inArraySortedByDescriptors: now quickly inserts objects at the end of the array if no sort descriptors are specified.
CPArrayController initialises its sort descriptors to an empty array.
Fixed: array controller's addObject added objects to the arranged objects array in a different order than in the content array if no sort descriptor was set.
- setValue:forKeyPath will now get call setValue:forKeyPath again on the value of the first part of the key path
- override valueForKeyPath: and valueForKey: in CPObjectController's CPControllerSelectionProxy to return controller markers when appropriate
- override setValue:forKeyPath: and setValue:forKey: in CPObjectController's CPControllerSelectionProxy to bypass possible controller markers
This commit also removes previous hacks for compound paths and adds test cases for the errors. All this is related to issue #967.
CPArrayController was erroneously using canRemove to determine whether explicit removal could take place, when in fact the documentation for [NSObjectController canRemove] (NSArrayController does not override canRemove) states that it returns "YES if an object can be removed from the receiver using remove:, otherwise NO", implying that it should only be used to check whether there is a selected element that can be removed (for example, to bind the enabled state of a toolbar button to a CPArrayController's canRemove property). It should not be used to decide whether a specific object can be removed or not.
When initializing a CPArrayController with a value that is not an array setContent: would make it an array by replacing value with [value]. This is consistent with how Cocoa does it, but it should first check to make sure that the content is not nil. If it is, the sensible thing to do (and what Cocoa does) is to replace it with an empty array.
This commit also includes a missing import to make CPArrayControllerTest run without error.
It is crucial that we send notification only before and after complete changes. In the middle notifications might cause observers to see or react to inconsistent data (e.g. selection indexes pointing to rows no longer present). Added some unit tests - more might be needed in the future. Fixed: before and after values when observing array controller key paths during content changes or rearranges were wrong. Fixed: the array controller sent out multiple redundant change notifications.
Alexander Ljungberg
Brian Donovan
cacaodev
Klaas Pieter Annema
Ross Boucher