A project from DailyDrip.
Fly is an OTP application you can use to transform files on the fly with various workers.
It's very much in-progress at present! As in, it doesn't work at all like you want yet! Ultimately, you will be able to use it to generate URLs that specify Just-in-Time transformations of a given file.
For now, the API will certainly change as it's extremely early in development.
There's an example application here.
In general, the setup is trivial. First, add the dependency and start the fly
application:
def application do
[
mod: {YourAppHere, []},
applications: [
# ...
:fly,
]
]
end
defp deps do
[ # ...
{:fly, path: "../fly"},
]
end
Then you'll want to configure Fly, in config/config.exs
. Here's an example
configuration:
config :fly, :workers,
%{
static: {Fly.Worker.StaticText, %{}},
pngify: {Fly.Worker.Pngify, %{}},
resize: {Fly.Worker.Resize, %{}}
}
Finally, forward requests to Fly.Plug
. In web/router.ex
:
defmodule YourAppHere.Router do
use YourAppHere.Web, :router
# ...
forward "/fly", Fly.Plug
# ...
end
Now, Fly
is ready to take your requests. You can create a URL that will route
a request through Fly.Plug
:
<img src="/fly/resize?url=http://www.rd.com/wp-content/uploads/sites/2/2016/04/01-cat-wants-to-tell-you-laptop.jpg&size=400x" />
Now if you load the page, you'll see the image in that URL returned, resized to
400 px wide and respecting its aspect ratio, using imagemagick to do the
resizing (but this is just a function of the resize
worker you configured
earlier!)
You can look at the Fly.Worker.Resize worker to see an example of building a worker. You configure them in your application configuration. An example is in this project's config.exs.
You can run workers directly:
Fly.run(:resize, input, %{size: "100x"})
About DailyDrip
This code is part of Elixir Drips, a daily continous learning website where you can just spend 5 minutes a day to learn more about Elixir (or other things!)
Here are the items I plan to add, in no particular order:
- Support workers that don't care about URLs at all - really, the workers should be responsible for fetching the URL, rather than the plug.
- Cache layer
- Add an LRU Cache
- This cache would be in our supervision tree, of course.
- Output file caching
- We need to support caching output files, so that when someone requests a job that's already been worked, they get it from the cache rather than generating it via compute again.
- Input file caching
- I'd like to add an additional caching layer for the input files. This way, if we add something like an image filter generator, it only downloads the source file one time.
- Image filter worker
This project is licensed under the MIT License.