In my previous post I’ve explained that the new Codable protocol of Swift 4 has some problems with references to objects (i.e. class instances). Or, better said, the two implementations of coders available
have this problem: they expand all references to new objects and create doublets if references in the object tree occur more than once for a single object. In the case of loops in the object tree (references of objects to each other) the encoding leads to an endless loop.
The old NSKeyedArchiver/NSKeyedUnarchiver of the Objective C-world do not have this problem: they respect references and restore them correctly at unarchiving. So, I want to implement this old way, i.e. the NSCoding protocol in my LinkedList as well. Remark: LinkedList uses references to its payload as well even in loops but that is taking care of even in the use of the Coding protocol.
Maybe you have read my series of posts about LinkedLists. It was great fun for me to put these things together, but recently, I stumbled over a problem with serialisation, esp. the Codable protocol and references to objects. In Swift 5 there are two Encoder/Decoder pairs available:
It turns out that these encoders are not really suitable for archiving object graphs because they expand all references to objects, i.e. put the objects instead of the references in the archive. If you have a simple tree of objects and you do not care about doublets that may be fine but as soon as you have loops in your object graph you are in trouble. Encoding such a graph with these two encoders lead to an endless loop. This may happen quite frequently; just think about delegation where you typically have references of the delegate and its master against each other.
This is the first post of my upcoming series devoted to linked lists. I like this data structure very much because of its elegance and efficiency in situations where you have to insert and remove objects quite often and randomly. Unfortunately there is no standard data type „Linked List“ in Swift, so we have to build one by ourself. That’s not too difficult and there are quite a few example implementations in the web (e.g. from Chris Pilcher or Hugo Tunius). I’ll take these as a starting point and add some of my own thinkings and optimisations.