Author Archives: bbbscarter

About bbbscarter

Coder, designer and general game developer. Co-creator of Fable. Billionaire philanthropist. Serial liar.

UberLogger – Redux

Having not touched  UberLogger for a while, two users contacted me this week with some very decent feature requests (thanks Steve and Gavin!). What should have been a fairly quick set of changes turned into an extensive rewrite. But good news – version 2 is now live!


The core of the work was around  Unity’s GUI ‘layout’ system which takes care of placing GUI widgets. Instead of specifying positions for controls, you express your layout in terms of ‘vertical’ and ‘horizontal’ sections and flexible spacers and let Unity do the work of translating that into pixel positions.

Unfortunately I hit the point where it became more trouble than it was worth. Specifically:

  • I wanted to add ‘auto scroll’ to the logging list, so the logging window will automatically move to the end of the list. In order to do this you need to know where the end of the scroll window is… and when you use layouts you don’t know where the controls will end up. I tried working things out by myself, but ultimately just ended up fighting with the layouting.
  • Internally, when doing the layouting, Unity calls OnGui multiple times, first to get the control positions and then to render; in order for this to work it needs the results to be consistent between runs. This is normally fine, but I wanted to optimise the log view list, which gets slow when you’re adding 10k elements to it every time OnGui is run. The obvious optimisation is to not add elements that you know are outside of the scroll view (by calculating the positions of elements by hand) but this confused the layouting system – things were different between runs, and Unity complained.

Laying things out by hand is more work, obviously, but wasn’t actually as much trouble as I was expecting, and doing so allowed me to remove some fairly hacky bits of code and make the rendering several orders of magnitude faster.

After that, there was a fight with GUI skinning; as Gavin pointed out, UberLogger didn’t work nicely with the dark ‘Pro’ skin. Unity’s internal GUI skinning system is nice and flexible, but when you’re writing an add-on that’s meant to fit in seamlessly with the existing editor GUI skin you don’t want to create a new skin, you need to harvest what’s already being used. ‘EditorStyles’ is your friend here, and contains all sorts of useful styles you can reuse (like ‘toolbarButton’, for instance).

Unfortunately, it doesn’t contain all the styles the editor uses, and the console ‘log line’ styling is notably absent. UberLogger originally bodged around this by grabbing a slider bar control and hand-modifying some of the textures and colours – which is why it didn’t work on the Pro skin.

After poking around a bit, I realised that you can find all the GUI styles by iterating over; spitting out the names, I spotted ‘CN EntryBackEven’ and ‘CN EntryBackOdd’, and harvested their organs for my own logger styles.

Finally, an old and somewhat poor design decision came back and bit me. When I started writing UberLogger I initially had difficulty extracting all the information I needed from log calls made via Unity’s own Debug.Logxxx methods – in particular, you don’t get a useful callstack from them. My initial approach to fixing this was to spoof the Debug namespace with my own static Debug class and thus capture all calls to Debug.Logxxx. This works, but:

  1. Doesn’t fix the problem with extracting callstacks from errors fired deep from within the bowels of Unity itself. This was fixed a while ago by wrangling the log message Unity gives to you.
  2. You have to spoof *everything* in the Debug namespace, even stuff you don’t care about. This is messy and guaranteed to break if ever Unity add something to the namespace.
  3. You can’t spoof everything – as Steve pointed out, among other things Debug.isDebugBuild is a property, and you can’t have properties in static classes in C#. Unfortunately this was breaking the Unity Test Tools, which is a bad thing.

Thankfully, since 1. was fixed ages ago, the proxying wasn’t needed anyway. So I’ve now moved all UberLogger debug commands from Debug to UberDebug, and renamed them from ULogxxx to just Logxxx. Obviously this is a breaking change, but I think it’s a legitimate one.

So – version two features!

  • Much, much faster rendering when there are lots of elements in the log window.
  • Pro skin supported.
  • No longer conflicts with Unity Test Tools.
  • A new ‘Collapse’ button – works like Unity’s collapse button, but grouping similar messages together.
  • A new ‘Follow’ button, which makes the log window scroll to follow new messages.

As ever, you can get UberLogger from here.

Let me know if it works for you, or if you have any problems!


UberAudio – an improved audio workflow for Unity

A few months ago I decided to rewrite/tidy-up and open-source a bunch of Unity add-ons I’ve written and regularly use. First up was UberLogger, a drop-in replacement for Unity’s debug console. Next up is something slightly bigger – UberAudio.

Much like UberLogger, UberAudio is designed to take something in Unity that’s good and make it better. Unity’s built-in audio system is very powerful, but I found the workflow fell a little short of my needs; but, as ever, the marvellous thing about Unity is that it’s so easy to extend. Specific areas of audio workflow that I wanted to improve include:

  • Keeping AudioSource settings out of scenes and prefabs. If the settings of an AudioSource aren’t correct, you shouldn’t have to hunt around for it and check out a scene or a prefab. All your audio should be in one place, and not in a scene.
  • Similarly, if you want an audio designer to do a pass on the audio in your game, they should be able to do so without touching levels, prefabs or code. They should be able to drop new audio into the game by themselves.
  • Loading. You should be able to load up audio data without having to load a new scene. UI audio, audio variations, dropping audio for things you aren’t using to save memory, even swapping out audio sets entirely – you should have control over what’s loaded and when, without changing any game logic.
  • Flexible audio triggering. It should be trivial to play a random sound from a selection. Similarly, if you’ve got a variety of different creature types and they share some sounds and not others, it should be easy to create defaults and override the ones you need, without touching code, scenes or prefabs.
  • Lifetime management. If a creature is playing some audio when it dies, most of the time you don’t want the audio to die immediately with it; it should finish playing first. Likewise, if you have something that plays a looping sound (a buzzing bee, for example) and it dies, the loop should be released and then end naturally. This should be the default behaviour.

UberAudio solves all of these problems, and several others, and does so without getting in your way – you can always get a handle on your raw AudioSources if you need to. It’s based on the audio systems I wrote for Dungeon Keeper and Fable, so the core design has been tested in fairly intensive production environments and historically has scaled well.

The workflow is very simple:

  • Create an audio bank.
  • Add audio to your bank.
  • Mount the audio bank in your scene.
  • In code, trigger audio ‘events’ on GameObjects. These events have an intelligent lookup system (see the readme) so you can easily create default sounds and specialise them later on in your audio banks.

Editing audio banks in Unity looks like this:


As you can see, you have exactly the same level of control over how your audio behaves as with normal AudioSources but it’s all grouped together, with some extra fields for the new functionality.

Obviously this isn’t a competitor for great tools like wWise, but if you’re looking for something simple, flexible and free, UberAudio might be what you want.

Take a look and let me know if you find it useful. Feature requests, bug reports and pull requests are always welcome!


UberLogger – a replacement Unity console and logging framework

I’ve been using Unity for a few years now and have developed a deep and unhealthy love for it. As a game development framework it’s easily the most expressive and flexible system I’ve ever used, in 20 years of making games.

However, as with most things, there’s the odd bit of friction that could be smoothed over, and one of the joys of Unity is that it allows you to address them when you want to. I’ve developed a couple of such things over the years, which I intend on releasing freely to the community.

First up is a replacement logging system for Unity. Unity’s built-in logging system and console are great – being able to click on logs to select game objects, viewing callstacks, etc are incredibly handy features. But I found myself wanting more. Specifically:

  • Debug channels. When you’re trying to ship a game debug messages can get very spammy, and make it difficult to sift out the signal from the noise; at the same time those spammy messages can often be useful for tracking down rare bugs. Debug channels allow you to categorise your messages as ‘loading timings’, ‘weapon debugging’, etc, and limit your view to a particular channel, allowing you to keep all your messages without being overwhelmed by them.
  • Filtering out stack frames from the callstack. It’s not uncommon to want to create your own debug layer to manipulate your logging before it hits Unity. Unfortunately, by default your new methods appear in the callstack, and the source of the error is shown to be your own debug method, which is annoying – you want some way to remove elements from the callstack.
  • Inline source view. You see a bug and you want to quickly see the source code around one part of the stack frame, without necessarily jumping into your text editor.
  • A modular system to allow different backends to process logs. I want to be able to write my own file logger, or my own in-game console, and have full access to the callstack, etc.
  • Timestamps! I want to see when a log happened.
  • A more compressed view; Unity’s console uses space somewhat inefficiently, using two lines of text for every log.
  • An in-game console. Sure, much of the time I’m working in the Unity editor, but sometimes I want to see my messages on my target device.

UberLogger addresses all of these issues and some more. It’s a drop-in replacement for the default logging framework, so no code changes are needed – Debug.Log, etc, all just work, though if you want to use features like channels there are some new methods.

In the editor it looks like this:


And in game it looks like this:


It uses an MIT license, so if you’ve got any features to add send me a message (or a pull request). And if you find it useful, let me know!


Compiling Unity projects from within Emacs

I’m often asked how I get Emacs to compile up Unity projects*, and I thought it would be worthwhile writing a proper post on what I’ve done to make this a relatively painless process.

Once you’ve followed this post you should get syntax highlighting, in-Emacs compiles and on-the-fly syntax checking, as well as autocompletion and refactoring. I should say that this *doesn’t* stop Unity re-compiling your files when you switch back to it – I’ve not made any attempt to unify them, as I suspect it’s more trouble than it’s worth.

The key elements are:
– A ‘make’ script.
– Flycheck
– CSharp-mode.el
– Omnisharp-emacs

The make script

My make script is something I hacked together in python a couple of years ago, worked out from examining the compile commands Monodevelop emits, and it has grown to support various features. You can download it from here – be warned it’s not particularly pretty, and flits between about 3 different coding standards…

Drop it into the root of your Unity project and change it so it captures your source files. The arguments the make script expects are ‘fast/slow’ (explained in a moment), the working directory of the compile (usually the root of the Unity project) and, optionally, a file to exclude and a file to add (again, explained below).

The ‘fast/slow’ bit is something I added to allow for faster compiles. Once you’ve got a Unity project that includes lots of generic game libraries (your own utility functions, NGui, etc) compile times can slow down quite a bit. It therefore makes sense to partition your code into ‘stuff that’s rarely changed’ and ‘actual game stuff’, and only compile the rarely changed stuff… well, rarely.

If you look at the code of the make file there’s one function to get the source files for the rarely change library, and another function to get the source for the main game. If you pass down the ‘fast’ option, it the make script won’t bother compiling the library; if you pass down ‘slow’ it will. In this way it’s possible to keep your general compile times down in Unity to a couple of seconds, even for large projects.

Triggering a compile from within Emacs is fairly straightforward.

(defun unity-compile-game ()
(let ((cmd (concat "python " (project:project-root project:active-project) " fast " (project:project-root project:active-project))))
(compile cmd)))

(defun unity-recompile-game ()
(let ((cmd (concat “python ” (project:project-root project:active-project) “ slow ” (project:project-root project:active-project))))
(compile cmd)))

As you can see, all it does is create a command line of ‘python my-project-root/ my-project-root slow’ (or ‘fast’) and passes that down to the built in ‘compile’ command. I’ve got a custom project system I use in Emacs (I should probably move over to using something like projectile in the future) which I use to find the project root from a given source file – you should switch in whatever system you’re using instead.


I use the excellent flycheck to check my builds as I work. Because we’re able to keep our compile times nice and short it provides very fast feedback on code issues as you type, which I personally find invaluable. And, unlike alternatives like Monodevelop’s built in live ‘syntax checking’, it runs an actual build of the game so it’s always completely accurate.

Setting it up is fairly simple:
(require 'flycheck)
(flycheck-define-checker csharp-unity
"Custom checker for Unity projects"
:modes (csharp-mode)
:command ("python" (eval (concat (project:active-project-root) "")) "fast" (eval (project:active-project-root)) source-original source)
:error-patterns((warning line-start (file-name) "(" line (zero-or-more not-newline) "): " (message) line-end)
(error line-start (file-name) "(" line (zero-or-more not-newline) "): " (message) line-end)))

Again, switch in whatever you need to get the root of your project. Flycheck works by saving out a temporary version of the file being edited, which can then be switched for the original file in the build – this is what the ‘exclude’ and ‘extra’ parameters of the make script are for. Thus the command line looks something “python SomeFolder/ SomeFolder SomeSourceFile.cs SomeSourceFileFlycheck.cs”.

I should add that the error patterns I’ve given to flycheck are probably monodevelop specific and have only been tested on my OSX machines.


CSharp-mode provides C# syntax highlighting and other things for Emacs. It looks like it has been abandoned, but it’s complete enough that I’ve not hit any major issues with it.

I tend turn off csharp-mode’s (dodgy) imenu support and turn on it’s brace matching:

(setq csharp-want-imenu nil)
(local-set-key (kbd "{") 'csharp-insert-open-brace)

Omnisharp for Emacs
Completing the picture is Omnisharp for Emacs, which I’ve blogged about before. It’s awesome, and gives you auto-completion, refactoring and all sorts of other loveliness. I personally use it in tandem with company-mode.

Oh, one more thing. Get annoyed on OSX because double clicking a file in Unity doesn’t open it at the right line number in Emacs? Some smart chap worked out how OSX handles line numbers (the command line isn’t good enough for some operating systems, apparently) and wrote Sublime Proxy to intercept the appropriate events. I’ve (badly) hacked it to work with Emacs. Get it here and make it less bad if you want.

That’s it for now. I hope someone finds it useful, let me know if something doesn’t work.


* Okay, not really that often.

‘var’, ELDoc and Omnisharp

A couple of weeks ago in the office we were having a conversation about the pros and cons of using ‘var’ in C#. If you’re not aware of it, ‘var’ is the C# equivalent of ‘auto’ in C++11; it tells the compiler work out the type of a variable declaration for you, based on the type of its initialising value. So for example:

List<GameObject> someList = new List<GameObject>();

can be replaced with:

var someList = new List<GameObject>();

The compiler can automatically work out that someList should be a list of GameObjects based on the initial assignment. Moreover because it does so at compile-time, there’s no runtime overhead, unlike with ‘real’ dynamic typing. Over time this can save a fair amount of typing, and reduces silly typo-based compiler errors.

It’s also a useful tool when refactoring. Consider the following:

foreach(SomeType loopVar in ImportantList)

This can be replaced with with:

foreach(var loopVar in ImportantList)

Now, if you change the type of ImportantList, the loop will ‘just work’ (as long as the loop body still contains valid code, obviously).

The concern in our office was regarding code readability. The more you use ‘var’, the harder it can be to work out what’s going on. For example:

var inventory = GetInventory();
var slot = inventory.GetSlot();
foreach(var item in slot.GetItems())
{ ... }

If you’re new to this code it’s hard to know what slot or item is without jumping to another file. Although I’m personally happy with the tradeoff, the concern is a valid one and I wanted to try and develop a tool to address the issues.

Thankfully a dash of Emacs, a bit of ELDoc and a giant pinch of OmniSharp provided a pretty comprehensive solution. ELDoc shows information about the symbol at point in the echo area. It didn’t take long to get OmniSharp and ELDoc working together, and the result is very handy even if you don’t use var.

Some examples:


Here we can clearly see the type that has been assigned to myTween in the echo area, even though we’ve declared it with var.


And here we can see the complete function signature and return type of, which makes understanding the code around the call easier, especially in the presence of heavily overloaded methods.

And that’s it! ELDoc support is in my OmniSharp fork on github, and it will hopefully get pulled across to main soon. I hope someone finds it useful.

My new +3 helm of swooping

I’ve been working on a post about game design. I’ve recently (re)learned lots of fun things about game prototyping, the importance of identifying and maintaining a game’s design ‘promise’, and how critical it is to keep the design/implementation/feedback loop tight. In short, it’s a post that’s not about Emacs.

This is not that post.

I’ve recently discovered helm-swoop. Helm-swoop is an Emacs helm extension that displays a list of matches for a search term in a popup helm buffer. You can navigate the list of matches with the cursor keys, and point will jump around in the buffer being searched to the relevant line. You can also continue typing to narrow down the search term further.

I’ve nabbed a gif from helm-swoop’s github page to make it clearer:


I’ve always been quite satisfied with the various incremental searching systems in Emacs; i-search and evil-mode’s vim-style search are excellent, best-in-class systems. But I love the way that helm-swoop shows you all the matches to your search, with some context, without you having to move point. You start typing, look to see if you get the hit you want, change your mind, retype. Once you’ve got something close to what you want you move down the list of hits, keeping your eye on the main buffer, and hit enter when you’re there. It’s very slick.

On top of this, you can live-edit the helm-swoop buffer, in a manner reminiscent of Emacs 24’s occur-edit-mode. Search for ‘color’, C-c C-e to go into edit mode, change the occurrences you care about to ‘colour’, C-c C-s, and the original buffer gets your changes and proper spelling. Fantastically powerful stuff.

i-search and the evil-equivalent still have their place, of course, and I use them regularly, but I’m using helm-swoop more and more.

That’s it! Try out helm-swoop, and I promise my next post will be something a bit less Emacsy.

Somehow I managed to make that sound like a threat…

C# autocompletion in Emacs

Just a quickie. The family are in Cornwall, so I was going to take the opportunity to knock up some kind of C# autocompletion – intellisense is the only thing I miss from my days of working with Visual Studio (though it’s not enough to make me want to embrace the horrors of Monodevelop). I was intending on glueing SublimeText’s CompleteSharp binary into Emacs with a bit of elisp and unicorn dust when I noticed on Reddit that someone had created Omnisharp, which integrates NRefactory, the code analysis and completion backend for Monodevelop, into Vim. Rather splendidly, they’d implemented it via an frontend-agnostic http server, allowing other editors to call to use it.

When I originally checked there were integrations for Vim and SublimeText, so I started working out how to wrangle http POST requests in lisp, but by the end of the week I noticed that someone had beaten me to it and created the rather wonderful Omnisharp-Emacs. Fortune was on my side.

I’ve put in a bit of work integrating it with my preferred completion system, company-mode, and now the full glory of C# completion is mine. Couple of screenies:

autocompletion popup

Here you can see the popup menu with the relevant completions, and more detailed information in the minibuffer.
autocompletion parameters

And here we’ve selected a candidate and company-mode is allowing us to fill in the parameters by tabbing through them and typing in text.

The company-mode integration is sitting in a fork in github, and I’m hoping to get it pulled upstream soon.

Omnisharp has lots of other tricks too, such as showing help documentation, renaming variables, jumping to definitions and lots of other fun stuff. I’m probably going to be using this on a daily basis.

There is a minor downside to all this, which is that NRefactory works in terms of Mono/Visual Studio solution files and their associated project files, and I’ve moved my workflow entirely over to my own, Emacsy, concept of projects which doesn’t require manual registration of files and things like that. I don’t like having to add files manually, but nor do I like having two systems to maintain, so I’m going to have to work out a way of getting them to play nicely together.

Besides that, very minor, wrinkle, I’m a happy man, and I’m hugely grateful to the authors of Omnisharp, Omnisharp-Emacs and company-mode – all three are fantastic.

Which leaves the rest of the weekend free to decide whether to buy an Ergodox keyboard with Cherry Blues or Cherry Browns, and then I can wear my geek badge with pride.


After a busy weekend of github-based collaboration, company-mode is now properly integrated into omnisharp-emacs. Head on over to the github page to have a look at the awesome features we’ve piled in – return type information, documentation buffers, tabbing through method parameters, smart completion detection, it’s all there.