I've been toying around with this in my mind for a while. I hope by typing it out to explain, I'll be able to work out the rough edges. It doesn't provide a change in the gameplay at all, its mostly a way to provide more meaningful data about what's going on in the simulation as far as evolution.
The phylogenic tree is a method in real biology to track ancestry and evolution over time. It's use is limited because when you begin throwing things like sex and viruses into the mix, things get very complicated very fast.
Darwinbots knows exactly what changed a DNA though, so I think it would be possible to construct a meaningful phylogenic tree in the program. It's possible to log the changes in DNA over time into a data structure that can be traversed without too much problem. This implementation assumes that codules are being used, and viruses are using codules. I'll explain trickier details near the end.
When the simulation starts, instead of actual DNA each robot has a link back to a single instance of its specie's DNA. This marks the head of the phylogenic tree for this species, and this phylogenic head carries the instance of the actual DNA.
Everytime an organism is created, a check is made to see if it's mutated at all from its ancestor. If it hasn't, its information (date of birth, net energy change over lifetime, date of death, parent(s), child(ren), any other information we want to provide) is logged into the node of its ancestor in the phylogenic tree.
If its DNA has changed, a new child node is created with the information of what's changed, in addition to its life stats.
When a bot's DNA program executes, it executes the DNA stored in the bot's node, or it searches backwards in the tree until it finds a complete copy of the DNA and then re-traverses the tree, modifying a copy of this root DNA as it goes. Every node doesn't need to store the DNA program, it just needs to provide a way to reconstruct the DNA at that node.
When the tree reaches a certain depth, certain child nodes are given a complete DNA program instead of just the information to reconstruct one. Very old nodes can be destructed back into the instructions to reconstruct the node as system resources becomes scarce.
The above is the base data structure I'm imagining for use in vanilla asexual reproduction. I'm hoping that since you're instancing DNA instead of using local copies for every bot, you can save some RAM. Of course, storing the entire phylogenic tree in memory after millions of cycles will probably use up some RAM as well. Luckily most of the phylogenic tree can reside in virtual memory just fine. It might also save some time with cancerous bots not needing to allocate and deallocate their DNA structures.
It should be fairly trivial to build a visualization function to display the tree, and provide a clear record of various strains, and how narrow the tree really is.
Tricky bits
Point mutations, viruses, and sexual recombination canmake things very tricky.
Viruses aren't that hard. If we assume that viruses are transmitted as whole codules, we just attach some additional information to codules if they're viruses that can be used to track the changes in the viral DNA over time. We record this information in a seperate phylogenic tree for the viruses.
Point mutations could be represented as a different sort of link than the ones between mother/daughter bot. Otherwise they would follow the same rules.
Sex really screws things up. I think trying to record DNA would be a disaster. We'd probably have to settle for the more traditional geneology approach. Bot A and Bot B had kid C during cycle X...
That's pretty much all I have at the moment. I'm looking for some feedback.
