Friday Facts #361 - Train stop limit, Tips and tricks

[FactorioBot]

Here it is! (beep boop)

https://factorio.com/blog/post/fff-361

[boskid]

As the original content for the train stop limit was decided to be too dense, technical and with not enough images, the original text for those interested is here:

technical wall of text about train stop limit

This is a tiny story about my trains related side project for a feature that was requested quite often (80501, 81682).

In large rail networks, there is quite often need to deliver a single item type to multiple places (let it be iron plates). Standard approach to solve this would be to have multiple target train stops with different names and dedicated trains. This approach avoids the case where too many trains would go to a single target jamming all the intersections since the station was not designed for that many trains.

Another approach is to have target train stops which belong to the same station. In this case, multiple trains will have the same schedule. To avoid all the trains from going to a single train stop and making other targets to starve, train stop can be disabled by circuit network when the request is already fulfilled. This approach has however one huge downside: trains horde. When a single train stop is enabled, a lot of trains may rush to it trying to deliver their items, first train is unloaded and rest of the trains after traveling half of the rail network have to go back since train stop is back disabled. And this is only a 1-to-N (or N-to-1) case. For the N-to-M delivery network (like multiple mining outposts delivering ore to multiple smelters) it was nearly impossible or would require crazy circuit networks setups that would selectively disable some train stops and would send limited amount of trains into rail network.

Suggested idea to limit amount of trains that could go to a single train stop was kind of simple and i could see how it would solving all of those cases. For this i decided to first make a prototype which would be as simple to implement as possible, but first i need a way to hook this mechanic into existing trains logic.

When a train wants to go to a station (which is just a name which can be assigned to multiple train stops), there are couple of things that need to happen:
- List of possible target rails is collected. Train stops have a rail to which they are connected to and a related inbound direction. In case of temporary train stops, they are simply a rail pointer which counts as 2 possible targets (one for each direction train could get to it).
- Current position of train is passed with the list of targets to a trains pathfinder, which finds a path to one of the provided train stops.
- Trains pathfinder returns a path which describes all the intermediate rails train has to visit to get to the destination.

Simplest and most obvious way to hook was to prevent given train stop from being added to the possible targets list when a train stop is full. With this in mind, first prototype came into life: trains limit was set by writing to an exposed trains_limit variable using console commands, and every time train was requested to find a path, i would go through list of all the trains, look to which train stops they are going to, convert it into amount of trains going for each train stop, and if given candidate train stop would happen to be full, it would not be added to the list of possible targets. That way pathfinder would not be able to find a path and consequently train would not be able to go to a full train stop.

With this prototype i was able to play with this feature (some details are in 80501). Quite quickly i noticed some issues with this approach. Save file i was using for testing had stacker with waypoint train stops in it.

high_ore_usage_output_stacker.png (1.64 MiB) Viewed 31656 times

When a train is moving, it has a braking point which is ahead of the train and is used to decide when to start breaking. Without it, if train would not be able to reserve a signal, it would stop immediately at the signal making bad experience. The same braking point is used with waypoint train stops: when a braking point reaches a waypoint, path is immediately expanded to point to the next target. Officialy (when looking into the locomotive) train is still heading for the waypoint, but internally it already has a path that goes beyond and ends on a train stop from next schedule record. When a train reaches a waypoint, official current goal is updated.

The issue was that the train when going to the waypoint, would keep that train stop from being overrun (when a train has a path to train stop, no more trains can go here, and when a train repaths, it temporarily clears its target which makes that train stop not full so it is able to path to the same train stop), but when the braking point passed that waypoint, the waypoint would no longer be considered as full. Other train far inside of the network would find a path and make the waypoint to be full. For now this is all fine, but when the first train's braking point reached a signal it was unable to reserve, repath was performed. Doing a repath when existing path has waypoints clears the waypoints and forces train to find a path to the waypoint again - and since it is already full because of other train, this train is now stuck in a NO_PATH state.

To fix this issue, when counting trains going for each train stop, i had to also count the waypoints into the usage counters. This guarantees that in case of a repath with full waypoint, clearing path will make the waypoint counter to be 1 less than full and train will be able to find a path to it.

At this point i also decided to add the circuit network input signal. This was a much better way to provide a limit value and would allow making more complex contraptions, like on ore outpost, setting limit of trains to a value related to the amount of ore in chests which could be loaded without the train waiting for more resources.

Another issue that was noticed with the prototype, was the backpressure: When all train stops from a given station were full, some trains when doing a repath would have no valid target making them go into the NO_PATH state. In general this approach was nice since it made trains at the ore outpost to wait before there was a free spot in iron smelter. The flaw was that when a train stop was no longer full, no repath event was generated. There was no good way to solve this in the current prototype since it would require recounting train stop usage counters every time any train would repath.

At this point the prototype was frozen for couple of months, from time to time i was looking into it to decide if it could be improved.

Some day (around 2 months before 1.0 release) i found a nice solution: i need a reliable reference counter. When a train gets a path to a train stop, in train stop the counter is simply incremented, and when the reference is cleared (because repath made train go to a different train stop) i could quickly check if new value is below current limit (keep in mind that limit was able to change) and if it was, train stop was marked and at the end of current tick, all trains in NO_PATH state heading for a station that has train stop marked would be forced to repath.

fff361-train-stop.png (84.66 KiB) Viewed 31656 times

This reference counter logic was nicely wrapped and added in 3 places: in a train (when a train is stopped at train stop, it keeps it reserved), in a train's path (when a train is heading to a train stop it keeps it reserved) and in path waypoints. Counter itself was directly available and i decided to expose it as a signal. At this point it was also possible to implement the hard trains limit logic: when a limit was decreased i could mark the train stop as to be updated and when it was found there are too many trains, all the trains heading for that train stop were collected, sorted by the distance remaining and the farthest trains would be forced to repath so the limit is enforced.

This approach was working quite nicely for me, however kovarex looked at it and decided it is a bad solution: when a train would be forced to go away, it could happen all other train stops are also full so the train could stop in the middle of the rail network complaining about the NO_PATH and blocking rails. It was far better if a train already has a reservation, it should be able to reach that train stop. For this i had to change the hook into trains logic: before the repath happens, old path is deleted. At this point reservation for the old target (or first waypoint) is stashed, and when possible targets list is created, train stop even when being full will still be returned as valid target if the train has a stashed reservation for it. This guarantees that train will not stop in the middle of the rail network because of the limit decreased. At the same time, this logic made the hard limit logic to fail, since trains over the limit were always able to find a path to a full train stop. Hard limit logic is now deleted, but if it would be highly requested by the mod authors, i could add it back.

Another complaint was the ability to specify the limit only using the circuit network: it was convenient and simple for me to implement it that way, but it was not user-friendly. In most cases the limit would be used as constant. For this i looked at the inserter's stack size override gui code and made similar thing in the train stop. Unfortunately there was no good place in the train stop gui due to the sliders to set a train stop color. When the train stop gui was reworked a little, there was a place for the simple trains limit text box that was not using the circuit control GUI.

There was one last thing to decide and solve. When all train stops are full, target list would be empty and train would enter a NO_PATH state. One possible solution was to handle this case the same way as station with all train stops disabled: train would skip such station. Personally i did not like that approach since it would not be possible to make a schedule for a train that has to visit all the intermediate stations (train that collects multiple different ingredients and then goes to final "crafting" station). Well, it would be possible using fake unreachable train stop, but then it would need to not have any limits and trains could overflow all the rails close to it, or the limit would be also set here and the problem would again appear when even this train stop was full. For this i decided the NO_PATH state is way better as it made the back pressure work nicely: train will not leave previous train stop it currently stands until it finds a valid target and is able to make a reservation. The NO_PATH message is however quite misleading. Technically it is a NO_PATH but i want to avoid players being confused when a train says there is no path while all the rails and signals are properly placed. To fix this, when collecting targets when any train stop was rejected because it was full, and in the end it happend the train has no valid targets, the train is told to show the "Destination full" message instead.

fff361-destination-full.gif (1.64 MiB) Viewed 31656 times

[glee8e]

I haven't come up with a practical use of this, but I can't help but ask this one thing: can we set the train limit on station via circuit network? Or, can we set the limit without physically walk to that station?

[mmmPI]

[Jürgen Erhard]

I haven't come up with a practical use of this, but I can't help but ask this one thing: can we set the train limit on station via circuit network? Or, can we set the limit without physically walk to that station?

Setting the limit via circuit network is explicitly mentioned.

As to walking there: I copy/paste what I want to modify to whereever I am, wait for bots to construct, change it, then copy/paste back. This assumes having radar coverage of what I want to change, of course (but I have radar coverage of everything anyway, I weird that way).

[ManaUser]

Nice. Just yesterday I was thinking about how I miss getting updates on what you guys are up to. I understand if every week was too much, but I'm glad to hear we'll be getting occasionally updates again.

As for the content, the train limit sounds great. I don't often reach a scale where this is a big problem, but I've come close enough to see there was a feature gap. This should be a significant improvement.

[Nosferatu]

Thanks for the FF!
Thanks for the railstop limit!!!

[glee8e]

Setting the limit via circuit network is explicitly mentioned.

As to walking there: I copy/paste what I want to modify to whereever I am, wait for bots to construct, change it, then copy/paste back. This assumes having radar coverage of what I want to change, of course (but I have radar coverage of everything anyway, I weird that way).

Somehow missed that paragraph. Perhaps I'm overhyped with the return of FFF XP

[Kalixt]

Noice!

[gtament]

Please fix spidertron doesn't build without player issue.

[conn11]

Glad to get some updates (and some more in the weeks to follow) on development.
Limiting the train number per station should be interesting. I’m excited to play around with that.
Together with some more map icons, logistics for spidertron and the upgrade for tip and tricks as well as the tutorials will make the supposedly small 1.1 quite the enhancement.

It seems logical not to bloat the information pages, but probably having links to the wiki on all the entities is something to be investigated.

[LiquidInsight]

It really made my night to see my RSS feed light up with a new FFF! So happy to read all of the upcoming changes. I always love reading about how the devs are fighting for UPS or continually perfecting new features. Speaking of, the new train-stop limits seem like they'll be a lifesaver for those not using LTN. You guys rock!

[Theamora]

imo as cool as the "tips and tricks" thing is, I genuinelly and personally think this ruins the game in a few aspects of simply not letting the player learn these kinds of things their selves. After 1000+ hours of factorio, still being able to figure out more efficent ways to do the same thing is something that makes you still want to comeback because a lot of the first few hundred hours is simply learning about these kinds of techniques and how to use them effectively.

I know you could probably argue that it saves a lot of time for newer players, but just having the ability to skip a lot of the learning process seems like you're removing a lot of the fun learning process the game has and knowing that someone else, with way less hours than you, can just know way more than you seems less encouraging than just learning this stuff yourself.

Idk, I think there's a fine line on keeping the best of both worlds rather than just ramming in this kind of information. The game is already stupidly easy as it is with how easy it is to give yourself limitless space, even with the enemies. The actual difficult part of the game is making something as efficent as possible or as tight as possible rather than just getting something to work.

[KatherineOfSky]

The changes looks AMAZING! Really excited to see the Train Stop Limit implemented in game.

The Tips looks great, though I feel VERY STRONGLY that you do not hide tips from the user. If anything, add a little dot or animated icon next to the pertinent ones, or ones most recently related to progress in the game. However, I ask you most sincerely not to hide information from the player.

Let me give you an example of a "smart program" that I used. This was a graphics program that I used many years ago to make images for my web page. I wanted to do something very simple: add a drop shadow to an object. I had my image ready, and I was searching that damn menu for HOURS trying to find the Drop Shadow selection. (I knew I had seen it before, but where the heck did it go???) Finally, on another day, I *selected the object* and THEN looked... it was right there, right where it should have been....

But the program didn't grey out the selection, it removed it from the menu completely, giving you no idea that it was there in the first place. After so much wasted time, I decided never to use that graphics program again and switched to a different suite of sproducts.

I prefer to see an option, even if I can't use it, and I think most Factorians would too. We are ravenous for information, and I can definitely see a new player watching a YT video and hearing about construction bots and wanting to check the tips section for that... even if they have just started their first game.

And while people wouldn't quit playing the game over it, I doubt they would ever open the Tips menu again, thinking that 5 topics is all they get for the whole game. (That's what I would do: there's no reason I would think that anything else is forthcoming. It would just look like a pitiful 5-point Tips sheet)

Anyway, hiding information is NEVER good, IMO. (Just ask me how I feel about the new YT Studio interface, which is maddeningly frustrating in its hiding of essential information, requiring mouseovers, clicks, and extra menus, when it used to be persistently displayed).

Please give people the information they are thirsting for, right out of the gate!


As a side note, it was lovely to read about what the team is doing! I've sorely missed the FFFs, and I know so many others in the community have as well! I hope you are all doing well!

[skelgaard]

About train limits, it should skip the stop if limit is changed while in route, if there are other stops where it can fit if there is limit in them.

that way you can do a depot where you can have more trains waiting, if no ore is ready/needed

[Lubricus]

Train stop limit is huge! It will change megabasing a lot! It sounds great. I personally think it should be hard/complex to scale up big. So I would have gone with the approach to somehow enable the possible to send a train to a specific train-station set by the circuits. As now there is no feasible way to do that in vanilla to se what is needed look at the LTN in vanilla project. https://www.reddit.com/r/factorio/comme ... 0_finally/

[TheEnemy42]

Weee, new FFF and great changes too!

As a translator I urge you to find a way where we get the newly translated strings into the new release. 1.0 still has the spidertron and new crashed spaceship wreck texts in English in all localisations. Especially I expect the new tips and tricks to be new strings heavy and if all of the new help texts are not translated for new players it'll be a dent in quality IMHO.

[Klonan]

Weee, new FFF and great changes too!

As a translator I urge you to find a way where we get the newly translated strings into the new release. 1.0 still has the spidertron and new crashed spaceship wreck texts in English in all localisations. Especially I expect the new tips and tricks to be new strings heavy and if all of the new help texts are not translated for new players it'll be a dent in quality IMHO.

Sure, the translations will be included as we start doing the deploys again, which automatically push and pull from crowdin.
There is no way to update the translations in 1.0 though.

[IronCartographer]

But the program didn't grey out the selection, it removed it from the menu completely, giving you no idea that it was there in the first place.
...
I prefer to see an option, even if I can't use it, and I think most Factorians would too.

Good point, I like the idea. In fact, the tech tree design referenced by the FFF is itself an example of highlighting the relevant techs but not making the later ones invisible!

I recently had a similar thought about the personal logistic requests UI. It would be great if it allowed you to use the same layout as the crafting menu (everything displayed) and then highlighted the active requests (currently the setup panel highlights in red the ones that already exist so you don't try to create a duplicate). Then there could be a toggle for hiding inactive requests, but the UI would be unified and quicker to customize (rather than opening a separate panel for selecting and configuring each item).

logistic requests UI.png (275.56 KiB) Viewed 31302 times

Relevant discussion about potential logistic requests improvement from a much older FFF thread: viewtopic.php?p=420541#p420541

[imTheSupremeOne]

Idea for a tip:

This is a problem that I had, and saw it in a few streamers that were new to the game — new players don't even consider that it's possible to put productivity modules into labs, or even science packs recipe; even if they are already using them for blue circuits.