It's pretty simple really. Well the idea is easy, coding it is a bit harder. But definately doable.
Mojang could just release an application that will convert your already existing maps into a map for the next version. How will it do this? I'll explain:
It should first generate a certain number (let's say 8) amount of chunks extra around the whole map, with the OLD chunk generator. These 8 chunks will be the chunks which will make the transition seem seamless. Cause what it can do after that is generate the SAME 8 chunks with the new chunk generator. Then interpolate these chunks. With this I mean that the first chunk, the closest, should be mostly the old generator, but a tiny bit of the new generator. The last chunk, the 8th, will be mostly the new generator. This way, the application will make a seamless conversion.
Let's say it's indeed 8. Then you have the chunk on the border of the already existing map (this chunk already existed) being 100% old generator. then the far edge of the 8th new chunk will be 100% new generator.
There could be a few problems with this of course. For example, what if the map has a hole of only 3 by 4 chunks that aren't generated yet, but all the chunks around it are? Stuff like this could be a problem, but that can be resolved. You could use an algorithm to find these chunks, and then fill them with the old generator.
There will probably be some other problems when doing this, but I'm pretty sure they can be resolved, being a coder myself.
I hope there is anyone out there who understood my post. But basically, to me this seems like a quite easy solution. Mojang of course doesn't want this in Minecraft itself, as Minecraft would then have to remember every old map generator. Which is why I propose a conversion application for every version that changes the biomes. I'm sure that with some work this can be done. Hell I'd even be ok with making this if I wouldn't have to go through the hassle of decoding Minecraft's sourcecode every new version...
That's actually a painfully long and complex solution when you get down to it.
The issue is that to transition from one generator to the next within a 16x16x64 chunk (over 16k blocks) is quite difficult. I suspect you could take the same chunk and generate it twice with both generators then interpolate the two together. But you will likely still end up with odd sections of terrain. But still better than cliffs.
You would have to find a way to determine the average density of each block by looking around all its neighbors. Perhaps interpolating based on the density of each block between both chunks.
It will look abrupt, but better than nothing. Rather than one long cliff, you'll have one long smoothed out cliff. It will still be odd.
I think potentially that idea could work if it generated a few chunks out, so it was a smoother transition.
But don't forget. If they create a separate program to do this, it directly takes away from their time adding new content that matters and fixing bugs.
That's actually a painfully long and complex solution when you get down to it.
The issue is that to transition from one generator to the next within a 16x16x64 chunk (over 16k blocks) is quite difficult. I suspect you could take the same chunk and generate it twice with both generators then interpolate the two together. But you will likely still end up with odd sections of terrain. But still better than cliffs.
You would have to find a way to determine the average density of each block by looking around all its neighbors. Perhaps interpolating based on the density of each block between both chunks.
It will look abrupt, but better than nothing. Rather than one long cliff, you'll have one long smoothed out cliff. It will still be odd.
I think potentially that idea could work if it generated a few chunks out, so it was a smoother transition.
But don't forget. If they create a separate program to do this, it directly takes away from their time adding new content that matters and fixing bugs.
Probably th converting will take a while yeah. My point was mostly to make a seperate conversion application for every new version with biome changes though, but then got caught up in what the application could look like :tongue.gif:
But anyway, what you say could work as well, but I think that would only make it more complex. Finding the density of a chunk only will not work. You'll need the density of the closest part of the chunk. What if there's only one cave in the whole chunk? The density will be high, so if the chunk next to it is generated, the interpolation where the cave is would look strange. Or am I missing your point?
Also, Interpolation usually gives way better results than you think. I'm pretty sure that biome transitions are actually done this way. At the edge between 2 biomes it interpolates between the 2. I'm not sure but I think that's what happens.
But yeh, my point is mostly that there could be a seperate application to make the map ready for the new map generation code. The actual algorithm I thought of was not thought through yet and I only thought of it while making the post itself, so that might be wrong. My point is more that there could be some algorithm that can make it look natural, it just requires a bit more thought to think of it.
And yeh, it'll take more time away from adding new features, but once the main concept behind it is done, then for every new version just implement the old generation code and the new code, and you're done. Making the actual application the first time will take the longest. After that it is basically copy/pasting... Not really but you know what I mean :tongue.gif:
This is one of the reasons I'm holding off on any major creations. I've been poking around and testing out all the new features, but with each release/pre-release I just end up making a new world. Even the SMP server I play on I've sorta stopped doing any building since it isn't even sure if we will keep the old maps or decide to start all over from scratch. But, as stated before, this is the danger of playing in a beta :smile.gif:
If they stopped changing the level generation code when people first started complaining about it, we wouldn't have biomes, caves or strongholds at all.
Thats just the price to pay for playing a beta game using a dynamic design process.
Are you aware that in 2 weeks this "it's a beta" excuse is gonna go boom? It's quite unusual to officially release a game with its core elements half-finished i.e. unusable. It's something I hope we won't get used to.
Sooner or later someone should realize this is not a hobby side project anymore, it's a product with millions of customers who deserve to get a finished game when the beta tag will be removed; maps needing to be erased/heavily edited after each update is fine in beta stage, after that is something highly disappointing for a sandbox game based on playing with the maps.
Probably th converting will take a while yeah. My point was mostly to make a seperate conversion application for every new version with biome changes though, but then got caught up in what the application could look like :tongue.gif:
But anyway, what you say could work as well, but I think that would only make it more complex. Finding the density of a chunk only will not work. You'll need the density of the closest part of the chunk. What if there's only one cave in the whole chunk? The density will be high, so if the chunk next to it is generated, the interpolation where the cave is would look strange. Or am I missing your point?
Also, Interpolation usually gives way better results than you think. I'm pretty sure that biome transitions are actually done this way. At the edge between 2 biomes it interpolates between the 2. I'm not sure but I think that's what happens.
But yeh, my point is mostly that there could be a seperate application to make the map ready for the new map generation code. The actual algorithm I thought of was not thought through yet and I only thought of it while making the post itself, so that might be wrong. My point is more that there could be some algorithm that can make it look natural, it just requires a bit more thought to think of it.
And yeh, it'll take more time away from adding new features, but once the main concept behind it is done, then for every new version just implement the old generation code and the new code, and you're done. Making the actual application the first time will take the longest. After that it is basically copy/pasting... Not really but you know what I mean :tongue.gif:
I don't think a separate application is a solution here. This process should be seamless just like terrain generation already is. It might be fine for modders and what not but I don't think a casual user will be impressed in a good way about having to individually convert all of their worlds before being able to play.
Well, I just registered to post about it, but then found this topic. I'm in exactly the same situation as OP was. Yes, I'm attached to my beta world, shame on me. I stick to already explored territory most of the time. That's why I experienced this issue so late. But...
From what I understand, as many posted here, we're not in beta anymore. Now I'm in "rage quit" mode, but some time later I will start new world probably. So here's my question: is biome generation stable now? Was there any official statement about this issue? Or should I expect another "rage mode" in my not so far future?
Biome generation was tweaker in one of the 1.1.0 pre-releases, and 1.1.0 might be released tomorrow, so you should wait a little longer.
Is mojang going to start accounting for this or not? I can only take so many mountains cut perfectly in half with a giant perfectly vertical cliff directly next to an ocean biome before I start feeling like I'm wasting my time building a world.
Really.. You are just going to have to say "no" eh? Based on massive expertise no doubt.
What this thread asks for is actually near impossible. In many ways, it's lucky at all we can even have a world that can be continued when the world gen changes. MANY games wouldn't even go that far, because it's problematic, as is demonstrated in this very game.
It could easily be the case that a world could never have any new features, and you always had to start over. But instead, we do at least get this compromise. But expecting perfect chunk alignment is completely overboard. When there are major changes to the map gen, it radically changes maps. Lining them up would be insanely difficult and require customized code for transitional areas and so forth.
You are basically asking them to waste lots of time on something when there are so many more pressing issues.. something that is easily overcome by MAKING A NEW MAP. You have been LUCKY to continue old maps at all. To whine that they don't line up perfectly demonstrates ignorance of the system, and unreasonable expectations.
Really.. You are just going to have to say "no" eh? Based on massive expertise no doubt.
What this thread asks for is actually near impossible. In many ways, it's lucky at all we can even have a world that can be continued when the world gen changes. MANY games wouldn't even go that far, because it's problematic, as is demonstrated in this very game.
It could easily be the case that a world could never have any new features, and you always had to start over. But instead, we do at least get this compromise. But expecting perfect chunk alignment is completely overboard. When there are major changes to the map gen, it radically changes maps. Lining them up would be insanely difficult and require customized code for transitional areas and so forth.
You are basically asking them to waste lots of time on something when there are so many more pressing issues.. something that is easily overcome by MAKING A NEW MAP. You have been LUCKY to continue old maps at all. To whine that they don't line up perfectly demonstrates ignorance of the system, and unreasonable expectations.
No, it is not impossible. They need to spend some time developing a solid terrain generation code, one that will develop an acceptable ratio of mountains to plains to hills to oceans. Once a basic template is made, the specific details can be modified without altering the base form of the landscape.
Tools exist to smooth transitions for those of you who genuinely care about eliminating cliffs. It's true that automatic detection and transitions between different terrain generators in game would be preferable to external tools.
The mcmerge tool and the decliff extension for MCEdit may be used to smooth transitions between chunk borders.
Well, I just registered to post about it, but then found this topic. I'm in exactly the same situation as OP was. Yes, I'm attached to my beta world, shame on me. I stick to already explored territory most of the time. That's why I experienced this issue so late. But...
From what I understand, as many posted here, we're not in beta anymore. Now I'm in "rage quit" mode, but some time later I will start new world probably. So here's my question: is biome generation stable now? Was there any official statement about this issue? Or should I expect another "rage mode" in my not so far future?
For the time being, yes, biome generation is stable right now. Even with the newest snapshot, and the new beaches it brought with it, terrain formations remain mostly the same as 1.0.0, with just little changes here and there. A couple extra hills, some new sand in place of old grassy coasts, tiaga biomes get snow & ice, and decorations (lakes, trees, etc) moved a bit. That's all.
For the time being, yes, biome generation is stable right now. Even with the newest snapshot, and the new beaches it brought with it, terrain formations remain mostly the same as 1.0.0, with just little changes here and there. A couple extra hills, some new sand in place of old grassy coasts, tiaga biomes get snow & ice, and decorations (lakes, trees, etc) moved a bit. That's all.
There are some small visible cliffs created by the new hills colliding into old terrain.
But lets be honest here. If anyone is petty enough to complain about this^ then they are petty enough to complain about anything.
Edit: I am severely misunderstanding the Minecraft community though. I expect to see a few "1.1.1 TERRAIN MADE UGLY CLIFFS GAME IS RUINED FOREVER" threads here in the next few days.
Rollback Post to RevisionRollBack
I still use "I" as my inventory key. Everyone who doesn't press "I" is wrong. How can people who press "e" live with themselves?
With this update, Pine forests will get snow. I personally don't mind, but not everyone would like their base to suddenly snow in. All we are asking for is a little bit more predictability and continuity. The changes we see now are small, but one day they'll add a new biome again and it will suddenly be snowing in your desert base. They need to find a way to avoid that.
I created a new world with 1.0, but I've been considering going back to my beta 1.4 world because I think we'll never get in a situation where world generation becomes stable anyway. I would need to delete a few chunks to make room for a stronghold, but everything else can be found with some exploring.
In my 1.0 world I built my base in the middle of the ocean and I try to focus all my builds in relative close proximity. I did that because it's a million times easier to transplant an ocean area into a new map than it is to make smooth transitions between new and old chunks on land. Maybe I'll take my 1.4 world buildings and move them in under the ocean on my new map.
"Oh look, the mighty patch cliffs! We're in 1.8.1 country, baby!" :laugh.gif:
That's actually a painfully long and complex solution when you get down to it.
The issue is that to transition from one generator to the next within a 16x16x64 chunk (over 16k blocks) is quite difficult. I suspect you could take the same chunk and generate it twice with both generators then interpolate the two together. But you will likely still end up with odd sections of terrain. But still better than cliffs.
You would have to find a way to determine the average density of each block by looking around all its neighbors. Perhaps interpolating based on the density of each block between both chunks.
It will look abrupt, but better than nothing. Rather than one long cliff, you'll have one long smoothed out cliff. It will still be odd.
I think potentially that idea could work if it generated a few chunks out, so it was a smoother transition.
But don't forget. If they create a separate program to do this, it directly takes away from their time adding new content that matters and fixing bugs.
Probably th converting will take a while yeah. My point was mostly to make a seperate conversion application for every new version with biome changes though, but then got caught up in what the application could look like :tongue.gif:
But anyway, what you say could work as well, but I think that would only make it more complex. Finding the density of a chunk only will not work. You'll need the density of the closest part of the chunk. What if there's only one cave in the whole chunk? The density will be high, so if the chunk next to it is generated, the interpolation where the cave is would look strange. Or am I missing your point?
Also, Interpolation usually gives way better results than you think. I'm pretty sure that biome transitions are actually done this way. At the edge between 2 biomes it interpolates between the 2. I'm not sure but I think that's what happens.
But yeh, my point is mostly that there could be a seperate application to make the map ready for the new map generation code. The actual algorithm I thought of was not thought through yet and I only thought of it while making the post itself, so that might be wrong. My point is more that there could be some algorithm that can make it look natural, it just requires a bit more thought to think of it.
And yeh, it'll take more time away from adding new features, but once the main concept behind it is done, then for every new version just implement the old generation code and the new code, and you're done. Making the actual application the first time will take the longest. After that it is basically copy/pasting... Not really but you know what I mean :tongue.gif:
Sooner or later someone should realize this is not a hobby side project anymore, it's a product with millions of customers who deserve to get a finished game when the beta tag will be removed; maps needing to be erased/heavily edited after each update is fine in beta stage, after that is something highly disappointing for a sandbox game based on playing with the maps.
I don't think a separate application is a solution here. This process should be seamless just like terrain generation already is. It might be fine for modders and what not but I don't think a casual user will be impressed in a good way about having to individually convert all of their worlds before being able to play.
Biome generation was tweaker in one of the 1.1.0 pre-releases, and 1.1.0 might be released tomorrow, so you should wait a little longer.
QUOTED. FOR. TRUTH.
Really.. You are just going to have to say "no" eh? Based on massive expertise no doubt.
What this thread asks for is actually near impossible. In many ways, it's lucky at all we can even have a world that can be continued when the world gen changes. MANY games wouldn't even go that far, because it's problematic, as is demonstrated in this very game.
It could easily be the case that a world could never have any new features, and you always had to start over. But instead, we do at least get this compromise. But expecting perfect chunk alignment is completely overboard. When there are major changes to the map gen, it radically changes maps. Lining them up would be insanely difficult and require customized code for transitional areas and so forth.
You are basically asking them to waste lots of time on something when there are so many more pressing issues.. something that is easily overcome by MAKING A NEW MAP. You have been LUCKY to continue old maps at all. To whine that they don't line up perfectly demonstrates ignorance of the system, and unreasonable expectations.
No, it is not impossible. They need to spend some time developing a solid terrain generation code, one that will develop an acceptable ratio of mountains to plains to hills to oceans. Once a basic template is made, the specific details can be modified without altering the base form of the landscape.
The mcmerge tool and the decliff extension for MCEdit may be used to smooth transitions between chunk borders.
For the time being, yes, biome generation is stable right now. Even with the newest snapshot, and the new beaches it brought with it, terrain formations remain mostly the same as 1.0.0, with just little changes here and there. A couple extra hills, some new sand in place of old grassy coasts, tiaga biomes get snow & ice, and decorations (lakes, trees, etc) moved a bit. That's all.
There are some small visible cliffs created by the new hills colliding into old terrain.
But lets be honest here. If anyone is petty enough to complain about this^ then they are petty enough to complain about anything.
Edit: I am severely misunderstanding the Minecraft community though. I expect to see a few "1.1.1 TERRAIN MADE UGLY CLIFFS GAME IS RUINED FOREVER" threads here in the next few days.
I still use "I" as my inventory key. Everyone who doesn't press "I" is wrong. How can people who press "e" live with themselves?
Yep, considering that would take like 15 dirt blocks to fix, I don't consider the newest update to be "save-breaking" like 1.8 and 1.9.
I created a new world with 1.0, but I've been considering going back to my beta 1.4 world because I think we'll never get in a situation where world generation becomes stable anyway. I would need to delete a few chunks to make room for a stronghold, but everything else can be found with some exploring.
In my 1.0 world I built my base in the middle of the ocean and I try to focus all my builds in relative close proximity. I did that because it's a million times easier to transplant an ocean area into a new map than it is to make smooth transitions between new and old chunks on land. Maybe I'll take my 1.4 world buildings and move them in under the ocean on my new map.