When a (Deep) Warm Ocean or (Deep) Frozen Ocean biome is close to a temperate landmass, the world generation create jaggy, repeating transitions, which tend to follow a North-South or East-West axis.
Notice how the water color changes in the provided screenshots, and look at the biome map given by Amidst. (The world seed is -7748547350291891684, and the screenshots below are at -1282 1772 and -2560 1530)
Note:
This change seems to be related to a 5x slowdown between 18w15a and 18w16a when generating the biome map with Amidst, which use the GenLayer class of the Minecraft .jar to calculate biome data. It is likely that this slowdown is also present in-game when generating chunks.
Note bis: When diffing the MCP-decompiled code for 18w15a and 18w16a, I found this change, which may be the culprit: https://pastebin.com/fQfT5BUW
Attachments
Comments 6
Still reproducible (sorry about removing the affected version; that was an edit conflict). Here's another picture, from 1.13-pre3 at /execute in overworld run tp @s -1272.43 105.87 1769.37 0.00 90.00, and with the biome blend set to 0 (much more visible that way):
The relevant code still exists in 1.13-pre3 too (cdo).
For these frozen ocean biome boundaries, the biome boundaries are identical to the temperate biomes 32 blocks away, and offset to chunks, so they are offset +32, +16, 0, -16 and -32 blocks. Rivers are counted as temperate biomes.
For an example, create a world using the seed -5 and examine the boundaries of the frozen ocean near spawn at -1024,512. This can be done using Amidst.
That's how some fractal based generation works, it basically uses the same base pattern with some randomness thrown in. In this case, perhaps the base pattern used is too jagged and perhaps could use a bit of smoothing out, and the repeating patterns could be fixed a bit by increasing the random values that are introduced to the base pattern. Of course, all of this is based on observation and prior knowledge on the subject, and what I described may not be the method they used.