How to run and package Threepenny apps with Electron and Electron Packager.
For reference, a minimal working example is available here.
Normally when running a Threepenny app we execute our Haskell, with stack exec
or otherwise, which starts a local server and we open our browser on a certain
port to view our app. However this has a few subtle drawbacks.
One drawback is that most browsers are designed with remote servers in mind and have security features which don't make much sense for local server architectures like Threepenny's. Take file selection for example. When a user selects a file in a browser, the browser only exposes the file name and contents to the server, not allowing the server to receive the full file path. However for local server architectures there isn't much point in sending the entire file contents to the server since the server is on the same file system and could read the file directly, if only it had the full file path. Electron displays our app in a Chromium instance with many of these security features removed.
Another drawback is that the user has to run the app from the command line. Using electron-packager we can package native apps for Linux, macOS and Windows.
To run a Threepenny app with Electron we need an Electron main process. We provide this one: electron.js. It runs the following on startup:
- selects a free port to run on
- executes our Threepenny app binary, passing the port to run on as an argument
- waits for Threepenny's server to start accepting connections
- opens an Electron window which loads the URL of our Threepenny app
To get started with the linked electron.js
first add
this package.json to your project root directory.
You'll need Node installed and npm
on your PATH, confirm by running which npm
. Now run npm install
from the project root directory to install the
necessary dependencies.
The linked electron.js
executes the Threepenny app binary, passing the port to
run on as an argument. This of course means your Threepenny app needs to take
the port as an argument. We suggest also setting stdout to be line buffered, at
least while developing. Altogether it should look something like this:
module Main where
import System.Environment (getArgs)
import System.IO
import YourApp (start)
main :: IO ()
main = do
hSetBuffering stdout LineBuffering
[port, otherArg1, otherArg2] <- getArgs
start (read port)
Now copy the linked electron.js
to your project root directory. You'll have to
edit the defined constants: relBin
, which is the relative path from
electron.js
to your Threepenny application binary; and binArgs
, which
contains any additional arguments to pass to the binary. If you're not sure
about the relative path to your application binary, and you're using Stack, see
the next section.
Now run your app with Electron: ./node_modules/.bin/electron electron.js
relBin
is the relative path from electron.js
to your Threepenny application
binary. This might change depending on which tool or platform you are building
with and thus can be a pain to set manually. If you are using Stack you can
easily build your application binary to an explicit location, possibly a build
directory:
stack install --local-bin-path build
Now you can simply set relBin
to ./build/your-app-exe
.
This section assumes the app is already setup to run with Electron based on the above instructions.
First install electron-packager: npm install electron-packager
Optionally edit the "name" field in package.json
to set the name of the
packaged app. Then to package the app for the current platform, simply:
./node_modules/.bin/electron-packager .
This is the most basic way to package the app, it will copy the current
directory to the packaged app. However you'll likely want to avoid copying some
source files, which can be achieved with the --ignore
flag. You might end up
using:
./node_modules/.bin/electron-packager . --ignore=app --ignore=src
If you are using Stack and building your application binary to an explicit
location, as explained above, then you might want
to also ignore .stack-work/
. An icon can be set by passing the icon path to
--icon
, note that the icon format
depends on the platform.
For more options use the --help
flag.