Branch	Status
develop
master

• Table of Content
• Summary of the project
• SPEC
• Local development
  • Dependencies
    • Python3.9
    • pipenv
  • Using Postgres on local host
  • Creating the local environment
  • Setting up the local database
  • Using a local PostGres database instead of a container
  • Starting dev server
  • Running tests
    • Unit test logging
  • Linting and formatting your work
  • Using Django shell
• Migrate DB with Django
  • How to generate a db migrations script?
  • How to put the database to the state of a previous code base?
  • How to create a clean PR with a single migration script?
  • How to get a working database when migrations scripts screw up?
• Initial Setup up the RDS database and the user
• Deploying the project and continuous integration
• Docker
  • Configuration
    • General settings
    • Database settings
    • Asset Storage settings (AWS S3)
    • Development settings (only for local environment and DEV staging)
• Utility scripts
• "update_to_latest.sh" script

service-stac provides and manages access to packaged geospatial data and their metadata. It implements and extends the STAC API specification version 0.9.0 radiantearth/stac-spec/tree/v0.9.0/api-spec. Currently the STAC API has been split from the main STAC SPEC repository into radiantearth/stac-api-spec, which is under active development until the release 1.0-beta.

See SPEC

Prerequisites on host for development and build:

• python version 3.9
• libgdal-dev
• pipenv
• docker and docker-compose

If your Ubuntu distribution is missing Python 3.9, you may use the deadsnakes PPA and install it:

sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt-get update
sudo apt-get install python3.9

Generally, all modern distributions have already a pipenv package. However, the version may be dated, which could result in problems during the installation. In this case, uninstall the version from apt and install it manually like this:

pip install --user pipenv

At the time of writing, version 2022.11.30, should be working fine.

The other services that are used (Postgres with PostGIS extension for metadata and MinIO as local S3 replacement) are wrapped in a docker compose.

Starting postgres and MinIO is done with a simple

docker-compose up

in the source root folder (this is automatically done if you make setup). Make sure to run make setup before to ensure the necessary folders .volumes/* are in place. These folders are mounted in the services and allow data persistency over restarts of the containers.

If you wish to use a local postgres instance rather than the dockerised one, you'll also need the following :

• a local postgres (>= 12.0) running
• postgis extension installed (>= 3.0)

These steps will ensure you have everything needed to start working locally.

• Clone the repo

  git clone git@github.com:geoadmin/service-stac.git
  cd service-stac

• You can create and adapt your local environment variable in .env.local. This files is not under source control and if it doesn't exists during make setup it will be created from .env.default.

• Install and prepare all the dependencies (pip packages, minio, postgresql, .env.local, ...) by running

  make setup

• The command above has generated the following for you

  • python virtual environment with all dependencies (inclusive dev dependencies), you can locate the venv with pipenv --venv
  • started a minio docker container as S3 storage for the assets
  • started a PostGIS DB docker container

• You manually stop/start the minio and PostGIS DB with (see also Setting up the local database)

  docker-compose down
  docker-compose up

• Finally you can do the initial DB setup using django management commands

  # activate the python virtual environment
  pipenv shell
  # prepare the DB 
  ./app/manage.py migrate
  # Populate some test data in the DB
  ./app/manage.py populate_testdb
  # Create a superuser for the admin page
  ./app/manage.py createsuperuser

• Now you are ready to work with a full setup and 2 samples colletions

The service use two other services to run, Postgres with a PostGIS extension and S3. For local development, we recommend using the services given through the docker-compose.yml file, which will instantiate a Postgres container and a MinIO container which act as a local S3 replacement.

If you used the make setup command during the local environment creation, those two services should be already be up. You can check with

docker ps -a

which should give you a result like this :

CONTAINER ID   IMAGE                  COMMAND                   CREATED        STATUS                      PORTS                     NAMES
a63582388800   minio/mc               "/bin/sh -c '\n  set …"   39 hours ago   Exited (0) 40 seconds ago                             service-stac_s3-client_1
33deededf690   minio/minio            "/usr/bin/docker-ent…"    39 hours ago   Up 41 seconds               0.0.0.0:9090->9000/tcp    service-stac_s3_1
d158be863ac1   kartoza/postgis:12.0   "/bin/sh -c /docker-…"    39 hours ago   Up 41 seconds               0.0.0.0:15432->5432/tcp   service-stac_db_1

As you can see, MinIO is using two containers, one is the local S3 server, the other is a S3 client used to set the download policy of the bucket which allows anonymous downloads, and exits once its job is done. You should also have a postGIS container.

make setup also creates some necessary directories : .volumes/minio and .volumes/postgresql, which are mounted to the corresponding containers in order to allow data persistency.

Another way to start these containers (if, for example, they stopped) is with a simple

docker-compose up

To use a local postgres instance rather than a container, once you've ensured you've the needed dependencies, you should :

• Create a new superuser (required to create/destroy the test-databases) and a new database.

Note: the user/password and database name in the example below can be changed if required, these names reflects the one in .env.default.

sudo su - postgres
psql
# create a new user, for simplicity make it a superuser
# this allows the user to automatically create/destroy
# databases (used for testing)
psql> CREATE USER service_stac WITH PASSWORD 'service_stac';
psql> ALTER ROLE service_stac WITH SUPERUSER;
# We need a database with utf8 encoding (for jsonfield) and utf8 needs template0
psql> CREATE DATABASE service_stac_local WITH OWNER service_stac ENCODING 'UTF8' TEMPLATE template0;

The PostGIS extension will be installed automatically by Django.

Note: this is a local development setup and not suitable for production!

You might have to change your .env.local file especially the DB_PORT, if you're using this setup.

# enable first your virtual environment
pipenv shell
cd app
./manage.py runserver

./manage.py test

You can choose to create a new test-db on every run or to keep the db, which speeds testing up:

./manage.py test --keepdb

You can uses --parallel=20 which also speed up tests.

You can use --failfast to stop at the first error.

You can add the test python module to test as well

./manage.py test tests.test_asset_model

Alternatively you can use make to run the tests which will run all tests in parallel.

make test

or use the container environment like on the CI.

docker-compose -f docker-compose-ci.yml up --build --abort-on-container-exit

NOTE: the --build option is important otherwise the container will not be rebuild and you don't have the latest modification of the code.

By default only WARNING logs of the tests module is printed in the console during unit testing. All logs are also added to two logs files; app/tests/logs/unittest-json-logs.json and app/tests/logs/unittest-standard-logs.txt.

Alternatively for a finer logging granularity during unit test, a new logging configuration base on app/config/logging-cfg-unittest.yml can be generated and set via LOGGING_CFG environment variable or logging can be completely disabled by setting LOGGING_CFG=0.

In order to have a consistent code style the code should be formatted using yapf. Also to avoid syntax errors and non pythonic idioms code, the project uses the pylint linter. Both formatting and linter can be manually run using the following command:

make format

make lint

Formatting and linting should be at best integrated inside the IDE, for this look at Integrate yapf and pylint into IDE

Django shell can be use for development purpose (see Django: Playing with the API)

./manage.py shell

Logging is then redireted by default to the log files logs/management-standard-logs.txt and logs/management-json-logs.json. Only error logs are printed to the console. You can disable totally logging while playing with the shell as follow:

LOGGING_CFG=0 ./manage.py shell

NOTE: the environment variable can also be set in the .venv.local file.

For local development (or whenever you have a *-dev docker image deployed), there's shell_plus available (part of the package django_extensions), a shell on steroids that automatically pre-imports e.g. all model definitions and makes working with the Django API much easier

./manage.py shell_plus

With the Django shell ist is possible to migrate the state of the database according to the code base. Please consider following principles:

In general, the code base to setup the according state of the database ist stored here:

stac_api/migrations/
├── 0001_initial.py
├── 0002_auto_20201016_1423.py
├── 0003_auto_20201022_1346.py
├── 0004_auto_20201028.py

Please make sure, that per PR only one migrations script gets generated (if possible).

1. First of all this will only happen, when a model has changed

2. Following command will generate a new migration script:

   ./manage.py makemigrations

With the following command of the Django shell a specific state of the database can be achieved:

.manage.py migrate stac_api 0003_auto_20201022_1346

Under a clean PR, we mean that only one migration script comes along a PR. This can be obtained with the following steps (only if more than one migration script exist for this PR):

# 1. migrate back to the state before the PR
./manage.py migrate stac_api 0016_auto_20201022_1346

# 2. remove the migration scripts that have to be put together
cd stac_api/migrations && rm 0017_har.py 0018_toto.py 0019_final.py
./manage.py makemigrations

# 3. add the generated migration script to git
git add stac_api/migrations 0017_the_new_one.py

NOTE: When going back to a certain migration step, you have to pay attention, that this also involves deleting fields, that have not been added yet. Which, of course, involves that its content will be purged as well.

With the following commands it is possible to get a proper state of the database:

./manage.py reset_db
./manage.py migrate

Warning: reset_db a destructive command and will delete all structure and content of the database.

Right now the initial setup on the RDS database for the stagings dev, int and prod can be obtained with the helper script scripts/setup_rds_db.sh. The credentials come from gopass. To setup the RDS database on int, run following command:

    summon -p `which summon-gopass` -D APP_ENV=int scripts/setup_rds_db.sh

Note: The script won't delete the existing database.

When creating a PR, terraform should run a codebuild job to test and build automatically your PR as a tagged container. This container will only be pushed to dockerhub when the PR is accepted and merged.

This service is to be deployed to the Kubernetes cluster once it is merged.

The service is encapsulated in a Docker image. Images are pushed on the AWS elastic container registry (ECR). From each github PR that is merged into develop branch, two Docker images are built and pushed with the following tags:

• develop.latest (prod image)
• develop.latest-dev (dev image)

From each github PR that are merged into master, one Docker image is built an pushed with the following tag:

• master.GIT_HASH

Each images contains the following metadata:

• author
• target
• git.branch
• git.hash
• git.dirty
• version

These metadata can be seen directly on the dockerhub registry in the image layers or can be read with the following command

# NOTE: jq is only used for pretty printing the json output,
# you can install it with `apt install jq` or simply enter the command without it
docker image inspect --format='{{json .Config.Labels}}' swisstopo/service-stac:develop.latest-dev | jq

You can also check these metadata on a running container as follow

docker ps --format="table {{.ID}}\t{{.Image}}\t{{.Labels}}"d

The service is configured by Environment Variable:

These settings are read from settings_dev.py

The folder scripts contains several utility scripts that can be used for setting up local DBs, filling it with random data and the like and also for uploading files via the API.

In the main directory, there is a script named "update_to_latest.sh" which can be used to automatically update the version strings of the dependencies in the Pipfile. See the comment at the top of the script to learn how to use it.