A/B Testing Lambdas@Edge

Why?

As mentioned in the documentation for CloudFront Lambdas limitations

Environment variables aren't supported.

The reason I automated the functions generation was mainly to avoid human mistakes and centralize versioning of our A/B tests states.

Overview

When executed, build.sh generates 3 functions to use on AWS lambdas@edge.
Functions names match AWS default naming to avoid confusion.

The build simply loads the variables from the configuration file (conf) and replace the placeholders with the correct values.
Along with the function files, 2 other files are built:

1. build.json containing the information of the current state.
2. bookmarklet.js a js function to use as a bookmarklet to alternate between experiments.

Usage

1. Rename conf.sample to conf and update its values.
2. run ./build.sh. Functions will be automatically placed into the dist folder.
3. Create the lmabda functions with the generated code.
4. RUN TESTS!
5. Done! 🍰

Bookmarklet shield

If you fork this repository and commit your dist folder here's a shield.io badge to dynamically pull information from build.json and expose the current bookmarklet on your README.md file. fancy right?!

Considerations

• CF Origin should whitelist source header
• CF Origin should whitelist the cookie for the given cookie_name.
• Logs are region based. If you can't find them look into a closer region.
• When pointing to an S3 avoid Host on Whitelist Headers

  If the bucket has a different name than the domain name

• Did I mention creating tests?

Contributing

Contributions are welcome, and they are greatly appreciated! Every little bit helps, and credit will always be given.