Go and WebAssembly: running Go programs in your browser
For a long time, Javascript was the lingua franca amongst web developers. If you wanted to write a stable, mature web app, writing in javascript was pretty much the only way to go.
WebAssembly (also called wasm) is going to change that soon. Using WebAssembly one can write web apps in any language now. In this post, we will see how to write Go programs and run them in the browser using wasm.
But first, what is WebAssembly
The webassembly.org site defines it as “a binary instruction format for a stack-based virtual machine”. That is a great definition, but let us break it down to something we can easily understand.
Essentially, wasm is a binary format; just like ELF, Mach, and PE. The only difference is that it is for a virtual compilation target, not an actual physical machine. Why virtual? Because unlike C/C++ binaries, wasm binaries are not targeted for a specific platform. So you can use the same binary in Linux, Windows and Mac without changing anything. As a result, we need another “agent” which translates the wasm instructions inside the binary into platform specific instructions and runs them. Typically, this “agent” is a web-browser, but in theory, it can just as well be anything else.
This gives us a common compilation target to build web apps using any programming language of our choice! We don’t need to worry about the target platform, as long as we compile to the wasm format. Exactly like we write a web app, but now we have the advantage of writing it in whatever language we choose.
Hello WASM
Let us start with a simple “hello world” program to get a taste of things. Ensure that your Go version is at least 1.11. We can write something like this:
|
|
Save this in a file test.go
. This just looks like a regular Go program. Now let us compile this to target the wasm platform. We need to set the GOOS
and GOARCH
for that.
$GOOS=js GOARCH=wasm go build -o test.wasm test.go
So now we have the wasm binary generated. But unlike in native systems, we need to run it inside the browser. For this, we need to throw in a few more things to accomplish this:
- A webserver which will serve our web app.
- An index.html file which contains some js glue code needed to load the wasm binary.
- And a js file which serves as the communication interface between the browser and our wasm binary.
I like to think of it just like the things required to make The PowerPuff Girls.
And BOOM, we have a WebAssembly application!
We already have the html and the js file available in our Go distribution, so we will just copy them over.
1 2 3 4 5 6 7 8 9 10 11 |
$cp "$(go env GOROOT)/misc/wasm/wasm_exec.js" . $cp "$(go env GOROOT)/misc/wasm/wasm_exec.html" . $# we rename the html file to index.html for convenience. $mv wasm_exec.html index.html $ls -l total 8960 -rw-r--r-- 1 agniva agniva 1258 Dec 6 12:16 index.html -rwxrwxr-x 1 agniva agniva 6721905 Sep 24 12:28 serve -rw-rw-r-- 1 agniva agniva 76 Dec 6 12:08 test.go -rwxrwxr-x 1 agniva agniva 2425246 Dec 6 12:09 test.wasm -rw-r--r-- 1 agniva agniva 11905 Dec 6 12:16 wasm_exec.js |
serve
is a simple Go binary that serves files in the current directory. But just about any web server will do.
Once we run this, and open our browser. We see a Run
button, clicking which, will execute our application. Then we click it and check the console:
Neat! We just wrote a program in Go and ran it inside the browser.
So far so good. But this was a simple “hello world” program. A real-world web app needs to interact with the DOM. We need to respond to button click events, take input data from text boxes, and send data back to the DOM. Now we will build a minimal image editor, which will use all of these capabilities.
DOM API
But first, to interact with the browser from Go code, we need a DOM API. We have the syscall/js
library to help us out with that. It is a very basic, but nevertheless powerful form of a DOM API, on top of which we can build our app. Let us very quickly see some of its capabilities before we move over to making our app.
Callbacks
To respond to DOM events, we declare callbacks and hook them up with events like this:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 |
import “syscall/js” // Declare callback cb := js.NewEventCallback(js.PreventDefault, func(ev js.Value) { // handle event }) // Hook it up with a DOM event js.Global().Get("document"). Call("getElementById", "myBtn"). Call("addEventListener", "click", cb) // Call cb.Release() on your way out. |
Updating the DOM
To update the DOM from inside Go, we can do-
1 2 3 4 5 |
import “syscall/js” js.Global().Get("document"). Call("getElementById", "myTextBox"). Set("value", "hello wasm") |
You can even call JS functions and manipulate native native JS Objects like FileReader
or Canvas
. Feel free to check out the syscall/js documentation for further details.
Ok, now on with building our app!
A proper web app
We will build a small app which will take an input image, then perform some manipulations on the image like brightness, contrast, hue, saturation, and finally send the output image back to the browser. There will be sliders for each of these effects, which the user can change and see the target image change in real time.
First, we need to get the input image from the browser to our Go code, so that we can work on it. To efficiently do this, we need to resort to some unsafe
tricks, the details of which I will skip here. Once we have the image, it is fully in our control and we are free to do whatever with it. Below is a brief snippet from the image loader callback, slightly simplified for brevity:
|
|
Then we take user values from any of the effect sliders, and manipulate the image. We use the awesome bild library for that. Here is a small snippet of the contrast callback:
|
|
After this, we encode the target image to jpeg and send it back to the browser. Here is the full app in action:
We load the image:
Change contrast:
Change hue:
Awesome, we are able to natively manipulate images in the browser without writing a single line of Javascript! The source code can be found here.
Note that all this is being done natively in the browser itself. There are no Flash plugins, Java Applets or Silverlight magic happening here. WebAssembly is supported natively in the browser out of the box.
Final words
Some of my closing remarks:
- Since Go is a garbage collected language, the entire runtime is shipped inside the wasm binary. Hence it is common for binaries to have large sizes in the order of MBs. This is still a sore point compared to other languages like C/Rust; because shipping MBs of data to the browser is not ideal. However, if the wasm spec supports GC by itself, then this can change.
- Wasm support in Go is officially experimental. The
syscall/js
API itself is in flux and might change in future. If you see a bug, please feel free to file an issue at our issue tracker. - Like all technologies, WebAssembly is not a silver bullet. Sometimes, simple JS is faster and easier to write. However, the wasm spec itself is very much in development, and there are more features coming soon. Thread support is one such feature.
Hopefully, this post showed some of the cool aspects of WebAssembly and how you can write a fully-functioning web app using Go. Do try it out, and file issues if you see a bug. If you need any help, feel free to drop in to the #webassembly channel.