PicoLisp AWS Lambda Runtime

About

Creates a custom runtime for AWS Lambda allowing you to run PicoLisp functions in AWS Lambda. For convenience, aw’s picolisp-json and picolisp-awscurl are included.

The same variables used to configure libawscurl.l are also used in the runtime.

Thanks to PicoLisp’s small size and its speed, function invocation is nearly instant.

• [ ] Clean up code
• [ ] Better documentation
• [ ] Use Github actions to package the runtime
• [ ] More local testing of the stateful functions

Requirements

• Docker
• git
• Picolisp (tested with 19.12)
• An AWS account
• jq (for the publish.sh script)

Usage

• Build the runtime
• Upload picolisp.zip as a new Layer in AWS Lambda
• Create a new Lambda function, choosing the “custom runtime” option
• Add a new layer and select the new layer
• Write/upload PicoLisp code! Make sure your scripts have a .l extension
• For the handler, the format is: <filename>.<function name>

The Handler function takes 2 arguments: Event_Data and Headers. They are both alists.

Any kind of data can be returned by the handler, though it’s recommended to (json-encode) it.

Building

With Docker

• Check out scripts/package.sh

Manually

• Download picoLisp source code and build it
• copy runtime.l to the picoLisp folder
• copy bootstrap to the parent folder
• get json.l and libawscurl.l and copy them to folders of the same name
  • e.g.: cp path/to/json.l picoLisp/json/json.l
• zip -r picolisp.zip picoLisp/*.l picoLisp/bin picoLisp/lib/*.l picoLisp/json picoLisp/awscurl bootstrap

How it works

• The picolisp.zip file is unpacked to /opt
• AWS Lambda will find the /opt/bootstrap and execute it
• bootstrap will find the picolisp interpreter in /opt/picoLisp/bin/picolisp, load the libraries and start initializing the environment
• curl is used to obtain the event data and headers, which are parsed and turned into alists
• The handler is defined in the format <filename>.<function name>. The bootstrap parses this string, and uses the load function to load this file
• Then the function is evaluated using (eval (append (str Function) (quote Event_data) (quote Headers))), and the result is POSTed to the runtime API with curl
• In case of errors, *Err is executed, which will POST the error message to trhe runtime API
• There’s a (loop) around the above process. I’m not exactly sure why but it probably has to do with functions running in a loop and handling multiple events in a row, rather than launching a new container/process for each invocation. Without the loop, the function encounters errors, and without the POSTs in case of error or success, the function loops and processes the same Event_data until the max duration is reached as defined in the Lambda’s timeout setting.