Skip to main content

Devtools

OpenFn/Devtools is a set of tools for writing & testing job expressions, managing OpenFn projects, and developing new adaptors. It's how most people work with OpenFn from their own command lines, outside of OpenFn.org, Microservice, or Lightning.

Are you a developer?

The Devtools repo is a collection of bash and Node scripts, as well as a suggested (but not necessary) directory structure for working with OpenFn jobs and adaptors.

To run OpenFn jobs locally, you only need Core and at least one adaptor, e.g. language-http and you may prefer to install core globally via npm install -g @openfn/core

Up and running

  1. Install git and Node.js (version 14 or greater)

  2. Clone and install devtools to setup core, language-common, and language-http using either SSH or HTTPS:

git clone git@github.com:OpenFn/devtools.git
cd devtools
./install.sh ssh

Note: If you get a "permission denied" message when running ./install.sh, try run chmod +x ./install.sh then retry the install command.

Usage

Execute takes:

  1. -l [language-package].Adaptor: The adaptor being used
  2. -e [expression.js]: The expression being tested
  3. -s [state.json]: The message data: {...} and credential configuration: {...}
  4. -o [output.json]: The file to which the output will be written

Run a job using bash

~/devtools/core/bin/core execute \
-l ~/devtools/adaptors/language-http \
-s ./tmp/state.json \
-o ./tmp/output.json \
-e ./tmp/expression.js

More on Devtools

Install a specific adaptor version

To install specific adaptors, run ./install.sh ${ssh || https} language-${name}

When you install a new adaptor, the latest version will be enabled by default. To switch the adaptor version when running jobs locally, in the root of the adaptor directory, run:

git checkout tags/v2.4.15 (substitute 2.4.15 with the adaptor version you want)

The --test option

~/devtools/core/bin/core execute \
-l ~/devtools/adaptors/language-http \
-s ./tmp/state.json \
-o ./tmp/output.json \
-e ./tmp/expression.js \
--test

This intercepts all HTTP requests and displays the request information for debugging.

.FakeAdaptor

Adaptors may provide dummy modules for testing. language-salesforce has a built-in .FakeAdaptor which allows a user to test expressions on data without sending them to a real Salesforce server.

Instead of using -l ./language-salesforce.Adaptor, use -l./language-salesforce.FakeAdaptor to test expressions offline: ./core/bin/core execute -l ./language-salesforce.FakeAdaptor -s ./tmp/state.json -o ./tmp/output.json -e ./tmp/expression.js

Offline testing for other adaptors

For most standard adaptors which make use of HTTP requests, you can add --test to the execute command to intercept all HTTP requests and return a 200.

Hands-on with devtools and the command line

tip

Check out this example workflow for using devtools in your day-to-day.

  1. cd in the folder containing the repo you're working on.
  2. You can keep your job scripts anywhere, but store state.json and output.json in a tmp folder. In our repos we always add the tmp directory in our .gitignore file that tells Github to ignore the specified paths. Make sure you have your .gitignore file and you know what's tracked by Github and what's not. state and config may contain sensitive configuration information and project data so never upload them to Github!
  3. The devtools command is a mouthful. You can search your command line history with Ctl-r and typing core to pull it up the devtools command. Notice that it’s got line breaks and a flag for all the important bits… -l for language-package (adaptor), -s for state, -o for output, and -e for expression. You can also save your frequently used devtools commands in a document and just copy-paste.
  4. It's quick the change job names or the adaptor in the command. If you put all your adaptors in the same folder ~/devtools/adaptors/language-_________ you can quickly swap them in the command, as you can see in the video below. The Backspace key deletes characters behind your cursor, Delete deletes them in front.
  5. You can use the TAB key to auto-complete the file path as you search for a job.
  6. Once you've changed a couple of characters for the adaptor and expression (in the video state and output stayed the same because we're using the tmp convention) press enter and see the results.

devtools

Configure an OpenFn project

You can use YAML project configuration files to define projects both on the OpenFn platform and in OpenFn microservice. Running ./scripts/generate-project.js helps you build a project config YAML interactively, adding your triggers, credentials and jobs to the config. You can read more about the config file here

If you choose monolith mode, all your job code will be included in the YAML. In URI mode, you’ll get a config file with URI-s to your defined jobs.

Generate Project

Pre-Requisites

  1. Node is required to run jobs and use many of the scripts in Devtools (e.g., npm run build is required after changes to adaptors).

  2. A basic working knowledge of NodeJs, promises and asynchronous functions is essential for writing adaptors.

Scripts

Devtools comes with a collection of scripts to aid in setting up a development environment for adaptor work, and include commands to quickly clone a large number of adaptors, create tarballs of adaptors with only production dependencies included, etc.

For the kitchen sink, run:

./install ssh
./scripts/bootstrap npm-install

In order to run the scripts, ensure you have cd'd into the project directory and enter ./scripts/<script-name>

bootstrap

Installs all adaptors in repos file to the /adaptors directory and prepares the working directory. This needs to be run before running any of the other scripts. Pass npm-install to run npm install for each adaptor also.

./scripts/bootstrap npm-install - to clone, set up hooks and npm install in each ./scripts/bootstrap- to clone and set up hooks in each

generate-project.js

./scripts/generate-project.js interactively generates a YAML project configuration file that can be used both on the OpenFn platform and in OpenFn microservice to define projects.

generate-doclets

Iterates overs all language pack folder names found in the repos list and creates a doclet json file in the doclets directory.

analyse-doclets

Iterates overs all doclets found in doclets and gives a tree view of the doclet structure using jsdoc-query.

Building adaptors for platform

All adaptor releases are built inside a docker container. The importance of running the build and release process through a container is to standardize the build environment across the team. While adaptors can be built and run on lots of different operating systems and architectures, when we run the platform on Kubernetes it expects linux boxes running x86... so that's where we build these official releases.

Here's how to build and release adaptors:

  1. Reopen your package in dev-container by typing ctrl+shift+p (or cmd+shift+p on mac) and choosing Remote-Container: Rebuild and Reopen in Container.
  2. After the build is finished, open a terminal in vscode and run openfn-devtools release . to build, tag, and push to npm.
  3. Run openfn-devtools package-release . to package everything with production dependencies and push to Github.

Depending on how you've configured your local environment and your VSCode installation, you might encounter access issues preventing connections to NPM and GitHub.

Troubleshooting

There are a number of issues that you may encounter related to sharing settings that are responsible for passing ssh keys and local configurations from your host machine into the VSCode container.

Git config issues

An issue can pop up about git config not set, To solve this, you should probably set your email and name globally using the commands below:

git config --global user.email "youremail@something.com"
git config --global user.name "Your Name"

SSH key issues

You may find that you are unable to access your ssh keys from inside the container.

Error

permission denied (publickey)

To solve this, first make sure the ssh agent is up and running. In MacOS, it is running by default. On Linux you can start the agent using the command

eval $(ssh-agent -s)

Then you can add these line your ~/.bash_profile or ~/.zprofile (for Zsh) to make it run by default.

if [ -z "$SSH_AUTH_SOCK" ]; then
RUNNING_AGENT="`ps -ax | grep 'ssh-agent -s' | grep -v grep | wc -l | tr -d '[:space:]'`"
if [ "$RUNNING_AGENT" = "0" ]; then
# Launch a new instance of the agent
ssh-agent -s &> $HOME/.ssh/ssh-agent
fi
eval `cat $HOME/.ssh/ssh-agent`
fi

Next, run the command below to add your identity to the ssh agent:

  ssh-add <path-to-your-ssh-file>

Finally, configure VSCode to share your local ssh keys with the dev container. In VSCode, go to Settings, and in the search bar, type terminal.integrated.inherit. You should see the option in the image below and check it if it's unchecked.

vscode settings

Github token sharing

Our release process relies on a GH_TOKEN variable. Set up an access token in Github.

In your ~/.bash_profile or ~/.zshrc file, export the newly created token by adding this line:

export GH_TOKEN=<TOKEN>

Using a new adaptor in an OpenFn/platform instance

  1. Add your release to the scripts/install-lp script.
  2. Add the version number to priv/adaptors.json.
  3. Add the bodySchema to CredentialView.js.