Skip to content

serverlesspub/pandoc-aws-lambda-binary

Repository files navigation

Pandoc for AWS Lambda

Scripts to compile Pandoc for AWS Lambda instances powered by Amazon Linux 2.x, such as the nodejs10.x runtime, and the updated 2018.03 Amazon Linux 1 runtimes.

Usage

Absolutely the easiest way of using this is to pull it directly from the AWS Serverless Application repository into a CloudFormation/SAM application, or deploy directly from the Serverless Application Repository into your account, and then link as a layer.

The pandoc binary will be in /opt/bin/pandoc after linking the layer to a Lambda function.

For more information, check out the pandoc-lambda-layer application in the Serverless App Repository.

For manual deployments and custom builds, read below...

Prerequisites

  • Docker desktop
  • Unix Make environment
  • AWS command line utilities (just for deployment)

Compiling the code

  • start Docker services
  • make all

There are two make scripts in this project.

  • Makefile is intended to run on the build system, and just starts a Docker container matching the AWS Linux 2 environment for Lambda runtimes to compile Pandoc using the second script.
  • Makefile_Pandoc is the script that will run inside the container, and actually compile binaries.

The output will be in the result dir.

Configuring the build

By default, this compiles a version expecting to run as a Lambda layer from /opt. You can change the expected location by providing a TARGET variable when invoking make.

The default Docker image used is lambci/lambda-base-2:build. To use a different base, provide a DOCKER_IMAGE variable when invoking make.

Modify the versions of libraries or Pandoc directly in Makefile_Pandoc.

Experimenting

  • make bash to open an interactive shell with all the build directories mounted

Compiled info

pandoc 2.7.2
Compiled with pandoc-types 1.17.5.4, texmath 0.11.2.2, skylighting 0.7.7

Deploying to AWS as a layer

Run the following command to deploy the compiled result as a layer in your AWS account.

make deploy DEPLOYMENT_BUCKET=<YOUR BUCKET NAME>

configuring the deployment

By default, this uses pandoc-layer as the stack name. Provide a STACK_NAME variable when calling make deploy to use an alternative name.

example usage

An example project is in the example directory. It sets up two buckets, and listens to file uploads on the first bucket to convert and generate HTML files from markdown. You can deploy it from the root Makefile using:

make deploy-example DEPLOYMENT_BUCKET=<YOUR BUCKET NAME>

For more information, check out:

Author

Gojko Adzic https://gojko.net

License

About

pandoc precompiled binary for AWS Lambda with a Node.js wrapper

Resources

License

Stars

Watchers

Forks

Packages

No packages published