This repo is the Marble backend implementation:
- 1 single go app
- Postgres DB
Disclaimer
This repository’s README file is intended for internal use by our development team. The documentation provided here is specifically designed for setting up and running the project on macOS.
While external contributions and interest are appreciated, please note that we do not officially support setups on other operating systems. If you encounter issues outside of the macOS environment, support may be limited.
For general documentation and user-facing guides, please refer to our main repository: Marble Documentation.
Install mise-en-place or alternatively install Go manually on your laptop (see the version in go.mod).
Create your own .env file based on .env.example. You can customize it with your own values, but it should work locally with the default values (except for third-party service functionalities).
Launch the Postgres DB and other services used by the backend:
docker compose up -dNB: It creates a
marble-backend_postgres-dbvolume to store PG data.
Install the Firebase tools:
# Install the Firebase CLI
curl -sL https://firebase.tools | bash
# Check the version is at least 13.16.0
firebase --version
# Login to firebase cli
firebase loginThen copy the ./contrib/firebase-local-data.example folder to ./firebase-local-data. This folder will be used to store the local data of the Firebase emulator. It is ignored by git.
cp -r ./contrib/firebase-local-data.example ./firebase-local-data
Then start it using (replace [GOOGLE_CLOUD_PROJECT] with the value from your .env file):
firebase --project [GOOGLE_CLOUD_PROJECT] emulators:start --import=./firebase-local-data --export-on-exitNB: The
--importflag is used to import the local data into the emulator. The--export-on-exitflag is used to export the data when the emulator is stopped so you don't lose your changes.
Export your .env file (e.g mise) and run the root of the project:
The backend project is made of five discrete components:
- The API server
- The background worker
- The scheduled executor
- The data ingestion worker
- The pending webhook handler
Depending on which feature you need while developing, you should run one or more of those services. The last three are one-off commands that are usually run in cron and do not need to run in the background. The worker, though, handles all asynchronous background tasks the API needs (such as index creation) and might be requireed for some of the API functionnality to work properly.
The docker compose of this repository only contains the dependencies required to run the backend service, but does not start the services themselves. It is assumed the developer will run them themselves.
mise exec -- go run . --migrations --server
mise exec -- go run . --workerAlternatively without mise, export your .env file (e.g. using direnv) :
go run . --migrations --server # To start the API
go run . --worker # To start the workerIf you need to run the one-off components (for example if you are working on background data ingestion or on scheduled scenario execution), run them directly from your editor or the terminal when required:
mise exec -- go run . --scheduled-executer
mise exec -- go run . --send-pending-webhook-events
mise exec -- go run . --data-ingestionAlternatively, without mise :
go run . --scheduled-executer
go run . --send-pending-webhook-events
go run . --data-ingestionUsing VSCode, you can also run the
Migrate and Launch (.env)task in the "Run and debug" tab. This will load your env file, migrate the DB and start the server. Other components can also be started with the appropriately-named tasks.
The application can be run with the following flags:
--migrations: run the migrations--server: run the server--scheduled-executer: execute scheduled scenario job--data-ingestion: run data ingestion job--cron-scheduler: background job that automatically runs the cron jobs--send-pending-webhook-events: retry sending failed webhooks
NB:
.vscode/launch.jsoncontains the configuration to run the app with these flags.
The routing of the application is defined inside api/routes.go.
For further information on the API, you can also refer to the following resources:
- our API docs for public facing reference
- the Open API Specification defined in the frontend repository here.
The application uses goose to manage migrations.
Migrations are located in the repositories/migrations folder.
Execute the program with flags -migrations to run migrations.
To create a new migration, you can use the following command from within the repositories/migrations folder
goose create add_some_column sqlIt happens that the migrations end up being misordered. This happens if two people pushed new migrations A and B, B having a timestamp greated than A, but B B is commited first to main. The issue can occur when pushing to the main branch, or when pulling changes from remote main to the local branch. In this case, you may need to roll back a few local migrations before you can migrate up again.
The easiest way of doings this is by installing the goose cli with brew (brew install goose), configuring the goose environment variables
(typically export GOOSE_DRIVER=postgres GOOSE_DBSTRING="user=postgres dbname=marble host=localhost password=marble" GOOSE_MIGRATION_DIR="./repositories/migrations" should work),
and then running goose down as many times as needed from the repository root folder. See also the goose doc.
There is a recommended extensions list in the .vscode/extensions.json file.
All required configuration settings are already included inside the .vscode/settings.json file.
Recommended settings are in the .vscode/.user-settings.sample.json file. Cherry-pick them to your user config file.
- Run firebase emulator with paramater:
--export-on-exit - Add user, change options...
- Exit the emulator
NB: The data will be saved in the
./firebase-local-datafolder. If you want to share the data, you can copy it to./contrib/firebase-local-data.exampleand commit it.
docker volume rm marble-backend-postgres-db deletes the PG volume, useful to reset the app to a known state
In practice, this single-line will delete the stack and create a new one:
docker compose down && docker volume rm marble-backend-postgres-db && docker compose up -d