Hello Mavirick and JemaheChi,
I've been away for a long time due to my new job; but i keep an eye on Qhimm. If you have any hints about your new hashing algorithm i would be happy to create a "test version" of Berrymapper to test the algorithm.
Anyway, as we have a large number of hashed textures now ( characters, worldmap,GF,weapons, battlefields...) we can define already a close-to-perfect set of key pixels for your algorithm. We could even have different algorithm fit to each type of texture (one for characters, one for objects, one for worldmap).
Thanks Shunsq, I appreciate the support
I had a preliminary version back in March and I used a thrown-together hashing tool to confirm that it would work. The issue was that the textures I had to work with (before I lost them all
), even as many as there were, didn't come close to the full set.
Yeah, we have a large number of textures that have been used in different mods, but even if we had them all in one place it wouldn't be enough. The problem is that the textures we
don't replace in any mods are as important--perhaps even more so--than the ones we do; without them we could unwittingly end up with tons of unforeseen collisions and then you'd start seeing the replacement textures all over the place, or they would end up behind the scenes in palette textures, masks, etc.
The reason the new hashing algorithm was never released was because I didn't want people to start using it before the pixel set was solid. Each time the hash algo changes in any way, all of the hashmap CSV's for every mod have to be changed. This means a lot of work for modders who have to keep up with each iteration and a lot of confusion for players who wonder why the hashmap for one mod works great while another causes problems. Maybe this decision was a mistake. Maybe, as I mentioned in my last comment, it would be possible to define the pixels used in the hash externally, such that modders and players could have more control over the set.
Fortunately, FatedCourage has done an incredible job collecting textures from a playthrough of the entire game, and as of today he's gotten all of them to me: a grand total of 293,848 textures collected through Tonberry's debug feature. Obviously a ton of these will be duplicates, but through the same analysis I did earlier this year we should be able to start making headway on a good pixel set. Stay tuned.