Heat transfer has an apparent directional bias

[mel-odious]

When smelting multiple iron mix nodes, the nodes in the -X and -Z directions (west and south) heat faster and reach higher temperatures than the others. Not if this effect also occurs during cooking or firing pottery but those processes don't last as long so it may just be less obvious.

For a small underpowered smelter (three iron mix nodes and one charcoal block) the effect is most obvious: The node with the lowest X or Z coordinate heats the most, reaching 1800 C and smelting, while the middle node gets to around 1300 C and the opposite node stays at 1100 C. Once the hot node is finished smelting the other two heat up more, though the middle node is still hotter. Here's a screenshot where I extinguished the fires after the first node smelted and removed the top layer, north is up:

I've reproduced this on 0.3.3 and whatever branch is running on the Land of Catastrophe server currently (I'm m0dus there). I noticed from mods/climate/temperature.lua:482 in climate.air_temp_source(), the function iterates over the nearby heatable or air nodes - from some logging the ordering of the list is consistent with lower-coordinate nodes first so it hits the southwest nodes first. Then at line 510 a passed-in function parameter is modified each time the loop hits a heatable node: "temp_effect" is effectively scaled down with each pass through, starving the later nodes of heat.

A quick hack to lines 510 and 513 to declare and use a new local variable indeed leads to far more even heating, though I'm not familiar enough to know if that gives a reasonable heating rate. I don't have the repo cloned right now so you'll have to forgive the lack of PR, though I'm enjoying the game a lot so I might figure it out c:

[jeremyshannon]

Yeah at first glance, it looks like instead of using pairs and going methodically over each node in heatable with a decreasing chance, we should select a random entry up to #heatable and dump heat/cold there, then select another one until we've used up temp_effect. It might even be faster, too, if it dumps all its heat/cold in the first couple of blocks and then doesn't have to load metadata from the remaining blocks. That's probably not important, though, profiling showed that climate was not a big cpu sink. (that's the mobs.)

[jeremyshannon]

No, on second glance, that would make "chance" kind of useless, and while it makes a kind of sense, I don't think using up temp_effect was intended. A local variable, it is! This will probably buff heating again, at least a little.