We have been working to fix all the bugs and polish our latest release 0.15 for the last 3 months, and we feel now is the time to mark it as stable. This means we are happy with how the game is performing, and that there are no major game-breaking or save-corrupting issues. 0.15 is one of the biggest updates we have ever developed, with a large number of major and minor additions, alongside countless other changes.

Note: Due to a change in the size of boilers, previous steam engine power stations will be non-functional, and will need repair.

This new resource to mine and refine opens up a new branch of possibilities to players, from highly compact and efficient power stations, to deadly atomic bombs. Processing this new resource comes with its own challenges, the new centrifuge machine and enrichment recipes, and even the mining of this valuable ore is not simple.

The introduction of 3 new science packs - Military, Production and High-tech - drastically change the progression and feel of the game. The new recipes provide a steady increase in complexity, as well as acting as better stepping stones to build a factory around.

Introduced alongside these changes is the new infinite technologies. These are long term improvements that take a greater and greater sink of science to unlock. They offer bonuses such as Worker robot speed, Turret damage, Mining productivity and much more. However they require 'Space science packs', obtained only by sending satellites to space on a rocket.

This update adds the first batch of high-resolution sprites to the game, with the majority of necessary factory entities covered: Transport belts, Inserters, Assembling machines, Mining drills, Pipes, Trains, and so on. Not all of the game entities are available in high-resolution as of yet, but work is still in-progress to update the whole game to this new quality.
Set sprite resolution to ‘High’ in the graphics options to enable high-res sprites

A new system for storing and managing your in-game blueprints. Blueprints can be transferred easily from your player inventory into the library, and can then be accessed from another map on the same system. You can also share and see other players blueprints in multiplayer, just by joining a server. Sharing with players outside of the game has also been made a lot more simple, with the ability to export and import whole blueprints and blueprint books as lines of text.

These are only the major features, and there are countless other additions and changes to the game to discover: Wave defense and PvP scenarios, new oil processing recipes, programmable speakers, new logistic network overview, mini-tutorials...

We have a plan moving forward to update and finish the game, you can check our roadmap on our forum.

We have been working to fix all the bugs and polish our latest release 0.15 for the last 3 months, and we feel now is the time to mark it as stable. This means we are happy with how the game is performing, and that there are no major game-breaking or save-corrupting issues. 0.15 is one of the biggest updates we have ever developed, with a large number of major and minor additions, alongside countless other changes.

Note: Due to a change in the size of boilers, previous steam engine power stations will be non-functional, and will need repair.

This new resource to mine and refine opens up a new branch of possibilities to players, from highly compact and efficient power stations, to deadly atomic bombs. Processing this new resource comes with its own challenges, the new centrifuge machine and enrichment recipes, and even the mining of this valuable ore is not simple.

The introduction of 3 new science packs - Military, Production and High-tech - drastically change the progression and feel of the game. The new recipes provide a steady increase in complexity, as well as acting as better stepping stones to build a factory around.

Introduced alongside these changes is the new infinite technologies. These are long term improvements that take a greater and greater sink of science to unlock. They offer bonuses such as Worker robot speed, Turret damage, Mining productivity and much more. However they require 'Space science packs', obtained only by sending satellites to space on a rocket.

This update adds the first batch of high-resolution sprites to the game, with the majority of necessary factory entities covered: Transport belts, Inserters, Assembling machines, Mining drills, Pipes, Trains, and so on. Not all of the game entities are available in high-resolution as of yet, but work is still in-progress to update the whole game to this new quality.
Set sprite resolution to ‘High’ in the graphics options to enable high-res sprites

A new system for storing and managing your in-game blueprints. Blueprints can be transferred easily from your player inventory into the library, and can then be accessed from another map on the same system. You can also share and see other players blueprints in multiplayer, just by joining a server. Sharing with players outside of the game has also been made a lot more simple, with the ability to export and import whole blueprints and blueprint books as lines of text.

These are only the major features, and there are countless other additions and changes to the game to discover: Wave defense and PvP scenarios, new oil processing recipes, programmable speakers, new logistic network overview, mini-tutorials...

We have a plan moving forward to update and finish the game, you can check our roadmap on our forum.

• Train stop text angle is now configurable in the graphics settings. Default value is 30 degrees.
  

• Fixed that resizing the game window was very slow on Linux. more
  
• Fixed rendering of turret ranges on map. more
  
• Fixed "Disable listed mods" in minimal mode would disable all mods including Base if mod-list.json didn't exist. more
  
• Fixed that pressing escape in the "mods error" GUI would close it and leave the game in a broken state. more
  
• Fixed possible crash when loading game. more
  
• Car and Tank now make a sound when deconstructed. more
  
• Fixed that using the color command with no arguments would set your color to black.
  
• Fixed a crash when deleting a blueprint book while the label is being edited in the blueprint library. more
  
• Fixed crash when closing the game in the Generate Map GUI. more
  
• Fixed that the generate map GUI would show incorrect values for some of the enemy expansion settings in some cases. more
  
• Fixed that some blueprints would always have to be reuploaded after connecting to a server, even if they weren't changed. more
  
• Fixed that the map seed field wouldn't be used when given in a map-gen-settings.json file through the command line. more
  
• Fixed that missing controls in "autoplace_controls" for map gen settings would get filled with default values instead of disabling unlisted controls.
  
• Fixed that blueprints would stop transferring if the game was loaded from a map that included some transfers in progress. more
  
• Fixed missing font for Thai language.
  
• Changed the hazard concrete/concrete tile transition so it behaves predictively.
  

• Fixed layered icons would render incorrectly in some cases. more
  

• Fixed a crash when using LuaPlayer::disable_all_prototypes() and opening the technology GUI. more
  
• Fixed that research bonus could be set to negative. more
  
• Fixed that items in the character trash slots would get lost on reducing inventory size instead of spilling the items on the ground. more
  
• Fixed validation for pickup_position and drop_position. more
  
• Exposed internal buffer of fluid turret to Lua as its last fluidbox.
  
• Added LuaEntityPrototype::fluid_capacity read.
  

You can get experimental releases by selecting the 'experimental' beta branch under Factorio's properties in Steam.

• Train stop text angle is now configurable in the graphics settings. Default value is 30 degrees.
  

• Fixed that resizing the game window was very slow on Linux. more
  
• Fixed rendering of turret ranges on map. more
  
• Fixed "Disable listed mods" in minimal mode would disable all mods including Base if mod-list.json didn't exist. more
  
• Fixed that pressing escape in the "mods error" GUI would close it and leave the game in a broken state. more
  
• Fixed possible crash when loading game. more
  
• Car and Tank now make a sound when deconstructed. more
  
• Fixed that using the color command with no arguments would set your color to black.
  
• Fixed a crash when deleting a blueprint book while the label is being edited in the blueprint library. more
  
• Fixed crash when closing the game in the Generate Map GUI. more
  
• Fixed that the generate map GUI would show incorrect values for some of the enemy expansion settings in some cases. more
  
• Fixed that some blueprints would always have to be reuploaded after connecting to a server, even if they weren't changed. more
  
• Fixed that the map seed field wouldn't be used when given in a map-gen-settings.json file through the command line. more
  
• Fixed that missing controls in "autoplace_controls" for map gen settings would get filled with default values instead of disabling unlisted controls.
  
• Fixed that blueprints would stop transferring if the game was loaded from a map that included some transfers in progress. more
  
• Fixed missing font for Thai language.
  
• Changed the hazard concrete/concrete tile transition so it behaves predictively.
  

• Fixed layered icons would render incorrectly in some cases. more
  

• Fixed a crash when using LuaPlayer::disable_all_prototypes() and opening the technology GUI. more
  
• Fixed that research bonus could be set to negative. more
  
• Fixed that items in the character trash slots would get lost on reducing inventory size instead of spilling the items on the ground. more
  
• Fixed validation for pickup_position and drop_position. more
  
• Exposed internal buffer of fluid turret to Lua as its last fluidbox.
  
• Added LuaEntityPrototype::fluid_capacity read.
  

You can get experimental releases by selecting the 'experimental' beta branch under Factorio's properties in Steam.

Hello,
I can't believe that we have been able to produce a post every Friday for 200 weeks without missing a single one. To be honest I'm not sure if this isn't the right time to pause for a while, to avoid being this kind of show that gets worse and worse over time until it is so bad that you want to take your intestines and strangle yourself with them. But people in the office convinced me with arguments like "FFF is the only good thing we have", so we probably have to continue for a little longer.

Our 0.16 trello card list is long, probably too long, it contains 80 cards. Many of them have a similar history as this one:



The main reason of showing this is, that that it is almost certain, that some of these tasks will be pushed again to next version, or away completely. It depends how fast will the work on the more important things progress.

The bigger things are:

• Artillery train
• GUI/UX improvements
• Mod portal improvement (rewrite)
• Map generation improvement (related to the updated high res terrains and decoratives)

Then there are optimisations, we want to tackle mainly these:

• Train pathfinding and collision checking optimisations
• Additional smoke related optimisations
• Lamp related optimisations
• Item stack optimisation
• Optimisation of electric network update logic
• Optimisation of logistic robot movement
• Group multiple item icons to one sprite in order to optimize rendering of saturated belts

There are overall tweaks planned like:

• Find a way to show connected walls + pipes in a blueprint
• Confirm blueprint deletion
• Fluid squashing in pipes
• Map interaction improvements

And there are these semi-bugs or small things like:

• Train block its own path with chain signals
• Make programmable speaker 'Global playback' only apply for people on that force
• Electric network UI weird with lots of accumulators.

Lets hope we can finish at least half of the stuff we plan for 0.16.

The current terrain generation tends to make a world that looks the same everywhere.With the goal of making exploration more rewarding we've been working on a branch called "mapgen-fixes",which introduces some new biomes, new decoratives, and changes to the way things are distributed.I've been tasked with fixing up the obviously screwy parts(next on the to-do list is to deal with an overabundance of biters)so that this can finally get merged to master and others can start testing and giving their feedback.

On the one hand this kind of work is fun because you have a bit of artistic freedom.There is no 'right answer' to what the Factorio world should look like,so I can play with settings and try to create a result that I think is interesting.On the other hand, not having a 'right answer' means you can't just write a unit test to tell if your job is done.I needed a way to look at results from the map generator and compare changes across versions that would be more efficient and easier to automate than loading up the entire game and poking around in the map editor, waiting for chunks to load,and then trying to remember what it looked like before and asking myself if this is better or worse. So I created a map preview generator.

Map Preview Generator
In the interest of being able to compare against maps generated with the current versions of Factorio,I got the map preview generator merged early. In fact it's already available on 0.15;you can use it from the command-line like so (adjust path to Factorio binary as needed):



The above map is 900x900 pixels and has a resolution of 4 meters per pixel, so is 3.6km on a side.To give some sense of scale, a medium-sized factory with mining outposts could easily extend to the edge,but to cover the entire area would be rather impressive.

If you make a very zoomed-out map (scale=32 is the maximum supported for 0.15.x)you may notice that it takes a long time to generate the preview.On 0.16 this will have been taken care of by refactoring the terrain generation system to operate at arbitrary scales instead of always generating 32x32-tile chunks.To understand why this works, you need to understand that map generation is done in 2 phases:

• Calculating the value of several variables (elevation, temperature, humidity, and a few others) for every point on the map. This is done by sampling a coherent noise function (whose basis is similar but not identical to Perlin noise). The value at each point can be calculated solely based on the point's coordinates and the map seed, making this part 'embarassingly parallelizable' and easy to do at any scale (and with a bit of further refactoring, by arbitrary transformations of the inputs).
• One chunk at a time, 'correcting' the results of the first step, making sure entities don't overlap, replacing combinations of tiles that we don't have good transition graphics for, etc. This is where we enforce that the only thing that can be adjacent to water is grass, because that's the only transition we have graphics for (which happens to be something the art guys are working to rectify).

For the sake of making the map preview generator fast and simple,only the first step is done.This is a good enough approximation for my purposes,though if you look closely you may notice that shorelines are off by a tile or two in some places.

Programmable Noise
One of the things that I disliked about how Factorio's map generation works even before I was hired to improve it, was that once you've expanded your base to a certain size,exploration of the map ceases to be interesting.At an extreme scale this uniformity is exceedingly obvious,but it is also apparent at realistic factory scales.There are no especially large lakes to work your way around,and different parts of the map don't really have any unique character.

Take for instance this super zoomed-out map (32 tiles per pixel):



In part, this is due to the underlying noise functions we use not being very flexible.The function to calculate elevation, for example,is defined by a very small number of variables, including seed,number of octaves,and persistence.Persistence itself is varied over the map so that in some areas you get smooth coastlines and in others very noisy ones.This simple specification does a pretty good job of generating random-looking worlds,but it doesn't give as much control as I would like in order to make different parts of the world look different.

In order to efficiently calculate terrain,a lot of the low-level code is dedicated to caching noise values and vectorizing calculations on them.This is great, but means the high-level logic for how noise is generated is rather buried in a lot of implementation details,and is therefore difficult to reason about and even harder to change.But abstractly, the noise function is just a functional program that adds and multiplies values from the basis function (at different scales) together.In order to make this logic more obvious I've created an assembly-like language for specifying the steps to produce noise.e.g. Take basis noise at a certain scale, divide it by persistence, add in basis noise at another scale, and so on.This approach is super flexible, but the assembly language is not very nice to read or write.A longer-term solution will be to allow building noise functions from Lua code,which will have the side-effect of making the system hackable by modders.

Abstracting high-level aspects of terrain generation out of the C++ code opens up a lot of possibilities.Some ideas:

• Larger scale features that are modulated by distance so that the starting area remains relatively flat and predictable, but you find seas and larger continents far away.
• 'Ridge' some large-scale noise to produce linear features like mountain ranges or rivers.
• Altering the input to some noise layers with the output of another noise function. This can warp landforms in a way that looks a bit like the result of plate tectonics.

Related to this, we are also working to make the world seem a bit less flat, but that is probably better left for another FFF.

As always, leave us any feedback or comments over on our forum

Hello,
I can't believe that we have been able to produce a post every Friday for 200 weeks without missing a single one. To be honest I'm not sure if this isn't the right time to pause for a while, to avoid being this kind of show that gets worse and worse over time until it is so bad that you want to take your intestines and strangle yourself with them. But people in the office convinced me with arguments like "FFF is the only good thing we have", so we probably have to continue for a little longer.

Our 0.16 trello card list is long, probably too long, it contains 80 cards. Many of them have a similar history as this one:



The main reason of showing this is, that that it is almost certain, that some of these tasks will be pushed again to next version, or away completely. It depends how fast will the work on the more important things progress.

The bigger things are:

• Artillery train
• GUI/UX improvements
• Mod portal improvement (rewrite)
• Map generation improvement (related to the updated high res terrains and decoratives)

Then there are optimisations, we want to tackle mainly these:

• Train pathfinding and collision checking optimisations
• Additional smoke related optimisations
• Lamp related optimisations
• Item stack optimisation
• Optimisation of electric network update logic
• Optimisation of logistic robot movement
• Group multiple item icons to one sprite in order to optimize rendering of saturated belts

There are overall tweaks planned like:

• Find a way to show connected walls + pipes in a blueprint
• Confirm blueprint deletion
• Fluid squashing in pipes
• Map interaction improvements

And there are these semi-bugs or small things like:

• Train block its own path with chain signals
• Make programmable speaker 'Global playback' only apply for people on that force
• Electric network UI weird with lots of accumulators.

Lets hope we can finish at least half of the stuff we plan for 0.16.

The current terrain generation tends to make a world that looks the same everywhere.With the goal of making exploration more rewarding we've been working on a branch called "mapgen-fixes",which introduces some new biomes, new decoratives, and changes to the way things are distributed.I've been tasked with fixing up the obviously screwy parts(next on the to-do list is to deal with an overabundance of biters)so that this can finally get merged to master and others can start testing and giving their feedback.

On the one hand this kind of work is fun because you have a bit of artistic freedom.There is no 'right answer' to what the Factorio world should look like,so I can play with settings and try to create a result that I think is interesting.On the other hand, not having a 'right answer' means you can't just write a unit test to tell if your job is done.I needed a way to look at results from the map generator and compare changes across versions that would be more efficient and easier to automate than loading up the entire game and poking around in the map editor, waiting for chunks to load,and then trying to remember what it looked like before and asking myself if this is better or worse. So I created a map preview generator.

Map Preview Generator
In the interest of being able to compare against maps generated with the current versions of Factorio,I got the map preview generator merged early. In fact it's already available on 0.15;you can use it from the command-line like so (adjust path to Factorio binary as needed):



The above map is 900x900 pixels and has a resolution of 4 meters per pixel, so is 3.6km on a side.To give some sense of scale, a medium-sized factory with mining outposts could easily extend to the edge,but to cover the entire area would be rather impressive.

If you make a very zoomed-out map (scale=32 is the maximum supported for 0.15.x)you may notice that it takes a long time to generate the preview.On 0.16 this will have been taken care of by refactoring the terrain generation system to operate at arbitrary scales instead of always generating 32x32-tile chunks.To understand why this works, you need to understand that map generation is done in 2 phases:

• Calculating the value of several variables (elevation, temperature, humidity, and a few others) for every point on the map. This is done by sampling a coherent noise function (whose basis is similar but not identical to Perlin noise). The value at each point can be calculated solely based on the point's coordinates and the map seed, making this part 'embarassingly parallelizable' and easy to do at any scale (and with a bit of further refactoring, by arbitrary transformations of the inputs).
• One chunk at a time, 'correcting' the results of the first step, making sure entities don't overlap, replacing combinations of tiles that we don't have good transition graphics for, etc. This is where we enforce that the only thing that can be adjacent to water is grass, because that's the only transition we have graphics for (which happens to be something the art guys are working to rectify).

For the sake of making the map preview generator fast and simple,only the first step is done.This is a good enough approximation for my purposes,though if you look closely you may notice that shorelines are off by a tile or two in some places.

Programmable Noise
One of the things that I disliked about how Factorio's map generation works even before I was hired to improve it, was that once you've expanded your base to a certain size,exploration of the map ceases to be interesting.At an extreme scale this uniformity is exceedingly obvious,but it is also apparent at realistic factory scales.There are no especially large lakes to work your way around,and different parts of the map don't really have any unique character.

Take for instance this super zoomed-out map (32 tiles per pixel):



In part, this is due to the underlying noise functions we use not being very flexible.The function to calculate elevation, for example,is defined by a very small number of variables, including seed,number of octaves,and persistence.Persistence itself is varied over the map so that in some areas you get smooth coastlines and in others very noisy ones.This simple specification does a pretty good job of generating random-looking worlds,but it doesn't give as much control as I would like in order to make different parts of the world look different.

In order to efficiently calculate terrain,a lot of the low-level code is dedicated to caching noise values and vectorizing calculations on them.This is great, but means the high-level logic for how noise is generated is rather buried in a lot of implementation details,and is therefore difficult to reason about and even harder to change.But abstractly, the noise function is just a functional program that adds and multiplies values from the basis function (at different scales) together.In order to make this logic more obvious I've created an assembly-like language for specifying the steps to produce noise.e.g. Take basis noise at a certain scale, divide it by persistence, add in basis noise at another scale, and so on.This approach is super flexible, but the assembly language is not very nice to read or write.A longer-term solution will be to allow building noise functions from Lua code,which will have the side-effect of making the system hackable by modders.

Abstracting high-level aspects of terrain generation out of the C++ code opens up a lot of possibilities.Some ideas:

• Larger scale features that are modulated by distance so that the starting area remains relatively flat and predictable, but you find seas and larger continents far away.
• 'Ridge' some large-scale noise to produce linear features like mountain ranges or rivers.
• Altering the input to some noise layers with the output of another noise function. This can warp landforms in a way that looks a bit like the result of plate tectonics.

Related to this, we are also working to make the world seem a bit less flat, but that is probably better left for another FFF.

As always, leave us any feedback or comments over on our forum

Hello, version 0.15.30 was released today, and we consider it to be our first 0.15 stable version candidate, as long as no other big bugs appear, we will have 0.15 stable in a week.As the teasing never ends, it is time to show some ongoing work on 0.16 already.

The transport belt optimisations explained in FFF #176 are now merged in our 0.16 developer branch. I tested it on a megabase factory and the improvement was quite noticeable. The UPS of the whole factory was increased from 25 to 38, the belt update itself runs 5 times faster now. I hope that we can make the game run 2 times faster again before 0.16 is out :)

There is a basic problem with the way grass and water terrain interact in Factorio.

The first one is related to the fact that the transition is on the tile of the grass. This means that the whole transition is considered to be a grass tile, which leads to weird looking things like this:

The solution is quite simple, we will just define that the transition is drawn on the water tile instead, something like this:

It will make structures like this impossible:

as the minimum width of one tile of land will make this shape:

This is simple so far, but it creates new problem, as suddenly we need to solve situations like this:

it looks like this in 0.15:

but in next version, we would need to have graphics like this:

Note, that this is just one combination, we would need all the other shapes like "O", "U", "L", and mainly the graphics of the transition suddenly wouldn't have the space of the whole tile, as a possible transition to another terrain can be on the other side.
The solution of the problem is being investigated now by making some graphical experiments, but the possible solutions we see now is:

• Just make the transition smaller to make it work.
• Make 2-3 layers of the transition, so they can be combined together nicely.
• Use our 'tile correction logic' that we hate so much, which would just make this configuration impossible to exist.

As some people requested it, we made subtitles of the second part of the interview with me. It is more focused on the technical part of Factorio which might not be entertaining for so many people, but some might still find it interesting.


As always, let us know if you have any thoughts or feedback on our Forum.

Hello, version 0.15.30 was released today, and we consider it to be our first 0.15 stable version candidate, as long as no other big bugs appear, we will have 0.15 stable in a week.As the teasing never ends, it is time to show some ongoing work on 0.16 already.

The transport belt optimisations explained in FFF #176 are now merged in our 0.16 developer branch. I tested it on a megabase factory and the improvement was quite noticeable. The UPS of the whole factory was increased from 25 to 38, the belt update itself runs 5 times faster now. I hope that we can make the game run 2 times faster again before 0.16 is out :)

There is a basic problem with the way grass and water terrain interact in Factorio.

The first one is related to the fact that the transition is on the tile of the grass. This means that the whole transition is considered to be a grass tile, which leads to weird looking things like this:

The solution is quite simple, we will just define that the transition is drawn on the water tile instead, something like this:

It will make structures like this impossible:

as the minimum width of one tile of land will make this shape:

This is simple so far, but it creates new problem, as suddenly we need to solve situations like this:

it looks like this in 0.15:

but in next version, we would need to have graphics like this:

Note, that this is just one combination, we would need all the other shapes like "O", "U", "L", and mainly the graphics of the transition suddenly wouldn't have the space of the whole tile, as a possible transition to another terrain can be on the other side.
The solution of the problem is being investigated now by making some graphical experiments, but the possible solutions we see now is:

• Just make the transition smaller to make it work.
• Make 2-3 layers of the transition, so they can be combined together nicely.
• Use our 'tile correction logic' that we hate so much, which would just make this configuration impossible to exist.

As some people requested it, we made subtitles of the second part of the interview with me. It is more focused on the technical part of Factorio which might not be entertaining for so many people, but some might still find it interesting.


As always, let us know if you have any thoughts or feedback on our Forum.

• Fixed crash related to empty player blueprint shelf. more
  
• Fixed crash related to handling focused state of widgets.
  
• Fixed possible crash when using font with size 0. more
  
• Fixed focus error preventing to access GUI when the game is paused in multiplayer.
  
• Fixed a crash when the map can't be saved to disk due to permission errors when joining MP games. more
  

• Added optional "hide_resistances" to entity prototype to control whether resistances should be hidden in description for friendly forces. Default is true.
  

You can get experimental releases by selecting the 'experimental' beta branch under Factorio's properties in Steam.

• Fixed crash related to empty player blueprint shelf. more
  
• Fixed crash related to handling focused state of widgets.
  
• Fixed possible crash when using font with size 0. more
  
• Fixed focus error preventing to access GUI when the game is paused in multiplayer.
  
• Fixed a crash when the map can't be saved to disk due to permission errors when joining MP games. more
  

• Added optional "hide_resistances" to entity prototype to control whether resistances should be hidden in description for friendly forces. Default is true.
  

You can get experimental releases by selecting the 'experimental' beta branch under Factorio's properties in Steam.