[16.23] Fluids have direction preference

[Wolvereness]

I noticed that when unloading petroleum, the north side of a symmetrical setup will unload first. When this same setup is rotated to point south instead of west,
the west side will unload first. This image makes it a bit more clear:



When the train arrived at the platform, both wagons were full. When the screenshot was taken, the southern wagon was still half full. From the way the pipes are set up, there shouldn't be a reason that the volume difference is so dramatic.

[Aeternus]

I have noticed this as well. My guess is that this happens with a pipe at max pressure from both sides - when the fluid propagation is done, the missing fluid in the T section gets fully satisfied from the pressure of one side.

Also, 3 pumps per tank is excessive here. Without storage to dump into there is no way in hell that the fluid output of 6 pumps is going to be squeezed through that single pump. You -really- need some storage tanks inbetween.
"Wagon -> 1 pump -> Storage tank -> Pipe -> Pump feeding to factory" would be my way to set it up. It'll fully drain a wagon in less then 3 seconds (if the tanks are empty) and with the storage tanks connected by pipe, they should drain evenly.

[chrisgbk]

In the current implementation of Factorio fluids, there is a defined update order, so in T sections of pipe, one side will always have priority. This is not so much a bug, as a side-effect of how it was designed. North and West have priority IIRC.

[Klonan]

This is just due to the update order, connections are updated from 0 upwards, so some connections will flow preferentially

In the case of a 'fluid rewrite' this may or may not be fixed,
For now it won't fix it, and isn't a bug or some combination of 'we won't fix it and its also not a bug'

I will put it on the list of features we would like in a 'new fluid system', but if that will ever occur nobody knows

[impetus maximus]

i would put a storage tank for each fluid wagon. direct from a single pump without a pipe.
taking the pipe out of the equation, the transfer should be pretty rapid.

[Wolvereness]

I wonder how technically in-depth it would be to reverse the update order each tick, which would theoretically solve this without processing overhead.

[chrisgbk]

I wonder how technically in-depth it would be to reverse the update order each tick, which would theoretically solve this without processing overhead.

That wouldn't change the behaviour or theoretically solve anything; if you built all the pipes in the reverse order, that would be exactly the same as reversing the update order.

[mrvn]

I wonder how technically in-depth it would be to reverse the update order each tick, which would theoretically solve this without processing overhead.

That wouldn't change the behaviour or theoretically solve anything; if you built all the pipes in the reverse order, that would be exactly the same as reversing the update order.

What he means is that on even tick you do it one way and on odd ticks you do it the other way. So over time they would balance out.

Problem is there are 4 directions, not two. So at minimum it would be "tick % 4".