Page 2 of 2
Re: please do not use floating point value in Factorio
Posted: Fri Jun 04, 2021 3:46 pm
by boskid
Ambaire wrote: ↑Fri Jun 04, 2021 3:28 pm
Would it be possible to override that for modded games? Specifically, Pyanodons, and the seablock mod for it? Early game, I'm slowly producing liquid glass from quartz at a rate of 10 fluid units per 10 seconds or so, and with the underground pipe length I was using, I was experiencing 75% fluid loss thanks to that 'feature'.
Even with short range transport from a producer to a consumer, I was losing liquid glass over time at a rate of 0.1 units every minute or so. I could tell since the amount in the destination went from a whole number to .9, .8, etc.
Also, since pipes can only have one fluid type now without exploits, there is zero reason for any automatic flush to ever occur. If a manual flush is desired, it's easy enough to connect a pump to one end of the pipe system.
I am not sure if you are playing on the latest version, but the flush condition when all pipes have less than 0.05 of fluid was removed (
97338). Also the hard fluid mixing prevention was replaced with soft fluid mixing prevention where only manual placement of pipes by a player is checked against fluid mixing. Hard mixing prevention was causing too many corner cases to handle.
Re: please do not use floating point value in Factorio
Posted: Fri Jun 04, 2021 3:57 pm
by MiniHerc
boskid wrote: ↑Fri Jun 04, 2021 3:46 pm
Ambaire wrote: ↑Fri Jun 04, 2021 3:28 pm
Would it be possible to override that for modded games? Specifically, Pyanodons, and the seablock mod for it? Early game, I'm slowly producing liquid glass from quartz at a rate of 10 fluid units per 10 seconds or so, and with the underground pipe length I was using, I was experiencing 75% fluid loss thanks to that 'feature'.
Even with short range transport from a producer to a consumer, I was losing liquid glass over time at a rate of 0.1 units every minute or so. I could tell since the amount in the destination went from a whole number to .9, .8, etc.
Also, since pipes can only have one fluid type now without exploits, there is zero reason for any automatic flush to ever occur. If a manual flush is desired, it's easy enough to connect a pump to one end of the pipe system.
I am not sure if you are playing on the latest version, but the flush condition when all pipes have less than 0.05 of fluid was removed (
97338). Also the hard fluid mixing prevention was replaced with soft fluid mixing prevention where only manual placement of pipes by a player is checked against fluid mixing. Hard mixing prevention was causing too many corner cases to handle.
Oh. Sorry. Perhaps I was hallucinating it.
Re: please do not use floating point value in Factorio
Posted: Mon Jul 18, 2022 8:57 am
by jooe
Sorry to necro an old thread, but I am experiencing a very similar problem.
Having a long pipe connection (> 500 pipes or more) while the pipes are almost empty causes an extremely high loss of fluids.
In production graphics, while producing ~100 units/second, consumption was staying around 1 units/second (= 99% loss). Direct connection from producer to consumer, getting even worse when adding tanks or pumps in between.
The fact that tanks make it worse points to the float precision problem mentioned before, so there is probably not much to do about it, but I thought I'd mention it anyways.
I was able to improve the system a little by using a train to decrease the pipe lenghts, now losses are down to around 75%. Shorter connections are not really possible (highly modded game with extreme number of different products, cannot produce them all in one place ...)
Re: please do not use floating point value in Factorio
Posted: Mon Jul 18, 2022 1:00 pm
by FuryoftheStars
jooe wrote: ↑Mon Jul 18, 2022 8:57 am
Sorry to necro an old thread, but I am experiencing a very similar problem.
Having a long pipe connection (> 500 pipes or more) while the pipes are almost empty causes an extremely high loss of fluids.
In production graphics, while producing ~100 units/second, consumption was staying around 1 units/second (= 99% loss). Direct connection from producer to consumer, getting even worse when adding tanks or pumps in between.
The fact that tanks make it worse points to the float precision problem mentioned before, so there is probably not much to do about it, but I thought I'd mention it anyways.
I was able to improve the system a little by using a train to decrease the pipe lenghts, now losses are down to around 75%. Shorter connections are not really possible (highly modded game with extreme number of different products, cannot produce them all in one place ...)
This sounds like it might be more of a pressure issue than floating point loss:
https://wiki.factorio.com/Fluid_system#Transport
Re: please do not use floating point value in Factorio
Posted: Mon Jul 18, 2022 1:56 pm
by Zavian
Whilst a pressure issue is one possibility, the other common cause of very low fluid levels in pipes is demand outstripping supply.
If the output fluid boxes in the supply machines are empty, then that sounds like a case of demand exceeds supply.
Re: please do not use floating point value in Factorio
Posted: Mon Jul 18, 2022 2:57 pm
by FuryoftheStars
I will note that a pressure issue will cause exactly that as well, and considering they mentioned it's over a >500 pipe segment section, that's where I'm leaning.
Re: please do not use floating point value in Factorio
Posted: Mon Jul 18, 2022 6:16 pm
by atomizer
I did a trivial test and can confirm fluid loss:
viewtopic.php?f=7&t=102921
Re: please do not use floating point value in Factorio
Posted: Tue Jul 19, 2022 12:55 am
by FuryoftheStars
atomizer wrote: ↑Mon Jul 18, 2022 6:15 pm
Setup: about 600 pipes between two assemblers
Reproduction:
- add 1 barrel into unbarreling machine
- open the pipe
- observe fluid system contents drop over time, destroying fluid (increase game speed as needed)
The output machine is required, a standalone pipe does not lose fluid.
Using small amounts of fluid (like 1 barrel's worth), you are correct, it appears that the system does experience some kind of loss. Though it does settle at a point and stops losing.
Using as much as 5 barrel's worth, however, and setting the output inserter of the barreler assembler to insert into the unbarreler assembler (creating a closed circuit) does not experience loss.
This demonstrates that the loss only occurs when the pipes have extremely low fluid content per pipe.
Edit: Apologies, missed the fact that it only happens when the barreling assembler is attached. The fluid amounts flowing from the pipes to the assembler must be so low that floating point or something else is kicking in to cause the loss. (The original below should still be valid, though.) /Edit
That said, this should have no bearing on jooe's report as their system is producing ~100 fluid per second.
Re: please do not use floating point value in Factorio
Posted: Tue Jul 19, 2022 4:25 pm
by atomizer
FuryoftheStars wrote: ↑Tue Jul 19, 2022 12:55 am
That said, this should have no bearing on jooe's report as their system is producing ~100 fluid per second.
Indeed, at 100/s flow rate any pipe that can handle it (i.e. up to 2.4km) seems to not have any problems.
On a hunch, I also tested 100/m input flow and got the following:
Code: Select all
L | outflow (/m)
1 | 100
220 | 73.3
416 | 14.7
1k | 7
2k | 5.5
I didn't reach the reported 99% loss, but it's reasonable to assume there's a configuration that does.
Re: please do not use floating point value in Factorio
Posted: Wed Jul 20, 2022 11:48 am
by jooe
FuryoftheStars wrote: ↑Tue Jul 19, 2022 12:55 am
That said, this should have no bearing on jooe's report as their system is producing ~100 fluid per second.
I'm sorry for being confused myself, production was only 100/m, so that is well within the range of the tests atomizer did.
The 99% loss might have been when there were still some tanks in the circuit, as their fluid box is much larger, the floating point rounding issue probably kicks in a few magnitudes earlier.
I did solve the problem more or less by just increasing output. "We have 99% wasted production? Well, then let's produce 100 times more so it doesn't matter ..."