a very basic django app to run "digital exhibition websites"
We use docker to make development easier:
cp docker/.env.example docker/.env
edit the ./docker/.env
file using a proper database name and change the password; then
copy the ./example.env
file to .env
and fill the fields using the same database name and password.
This second step is needed because the environment variable names are different in
docker and in miller.
SECRET_KEY=*****
DEBUG=True
MILLER_DATABASE_NAME=your db name
MILLER_DATABASE_USER=your db user
MILLER_DATABASE_PASSWORD=your db pass
MILLER_DATABASE_HOST=localhost
then start the development docker with:
make run-dev
This will install all images (redis, postgres...) and build locally celery and miller for you.
Watchdog
takes care of restarting miller and celery when a py file change in the codebase.
For the first time, of whenever a new migration is available, make sure the db is aligned with:
make run-migrate
then test that everything works as expected:
make run-test
To create a new superuser
docker exec -it docker_miller_1 python manage.py makemigrations
The Visual Editor is our favorite way to handle themes and documents in Miller. It is a React app that connects flawlessy with the Miller JSON based api, and a few Configuration are needed to make the connection
Create a new Application instance in Miller admin. This
will be the URL if you run with the development docker compose. Fill with Client Type
set to Public
and Authorization Grant Type to Resource owner password-based
.
Put the given Client Id inside the .env file of the VisualEditor along with the relative URL of the JSON schema to validate documents:
REACT_APP_DOCUMENT_SCHEMA=
REACT_APP_MILLER_CLIENT_ID=
For a more traditional approach, you can use pipenv to install the dependencies and run the server. We still recommend to run docker image for running Postgres and Redis:
docker run -it --rm --name miller_postgres \
-e POSTGRES_PASSWORD=mysecretpassword \
-e POSTGRES_USER=miller \
-e PGDATA=/var/lib/postgresql/data/pgdata \
-v "$PWD/docker/postgres-data:/var/lib/postgresql/data" \
-p 54320:5432 \
postgres:14.1
In this case, use the sae POSTGRES_PASSWORD and POSTGRES_USER in the env file and we set the volume path to the one we defined earlier for our docker development stack.
Copy the ./env.example
file to ./.development.env
, then edit the values accoring to your system.
An example of a ./.development.env
file:
SECRET_KEY=*****
DEBUG=True
MILLER_DATABASE_NAME=your db name
MILLER_DATABASE_USER=your db user
MILLER_DATABASE_PASSWORD=your db pass
MILLER_DATABASE_HOST=localhost
MILLER_DATABASE_PORT=54320
These values replace the default values in ./miller/settings.py
thanks to
the method get_env_variable
:
DATABASES = {
'default': {
'ENGINE': 'django.db.backends.postgresql_psycopg2',
'NAME': get_env_variable('MILLER_DATABASE_NAME'),
'USER': get_env_variable('MILLER_DATABASE_USER'),
'PASSWORD': get_env_variable('MILLER_DATABASE_PASSWORD'),
'HOST': get_env_variable('MILLER_DATABASE_HOST', 'localhost'),
'PORT': get_env_variable('MILLER_DATABASE_PORT', '54320'),
}
}
All values in get_env_variable
can be replaced with the one of your choice
(e.g; postgres or redis port number can be changed)
Complete the intallation:
ENV=development pipenv run ./manage.py migrate
Run with:
ENV=development pipenv run ./manage.py runserver
In parallel, launch the celery tasks manager:
ENV=development pipenv run celery -A miller worker -l info
From v3.0.3 we added a convenient docker-compose file to run Postgres and Redis using the same env file configuration as pipenv:
ENV=development make run-pipenv
Then in another terminal, run:
ENV=development pipenv run ./manage.py runserver
Behind the scenes, it simply run the docker-compose file docker/docker-compose.pipenv.yml
.
Use test runner without DB:
ENV=development pipenv run ./manage.py test --testrunner=miller.test.NoDbTestRunner
Miller uses the external lib imagemagick to create thumbnails of your resources and Postgres database. Though we recommend that you use the docker image for development, sometimes you just need to work the old way.
To install the correct version of python, you can follow the doc at https://hackernoon.com/reaching-python-development-nirvana-bb5692adf30c
pyenv installs 3.8.0
pyenv local 3.8.0
In order to install pipenv using the correct version of python,
use the pip
module that is shipped with local python version:
python -m pip install --upgrade pip
python -m pip install pipenv
Install the library imagemagick6
according to your OS, then install requirements:
pipenv install
Configure in Three steps:
2. get valid JSON schema files for the data instance and the data payload in your MILLER_SCHEMA_ROOT folder
for instance, given this structure
docker/data/
├─ private/
│ ├─ my-google-service-account.json
│ ├─ my-schema/
│ │ ├─ document/
│ │ │ ├─ instance.json
│ │ │ ├─ payload.json
-
your MILLER_SCHEMA_ROOT variable will be MILLER_SCHEMA_ROOT=/private/my-schema
as /private is mounted on your the docker/data/private of your local installation.
3.run using the env variables MILLER_SCHEMA_ROOT, GOOGLE_SPREADHSEEET_ID and GOOGLE_SPREADHSEEET_SERVICE_ACCOUNT_KEY
MILLER_SCHEMA_ROOT=/private/my-schema \
GOOGLE_SPREADHSEEET_ID=xYz \
GOOGLE_SPREADHSEEET_SERVICE_ACCOUNT_KEY=/private/my-google-service-account.json \
make run-dev
then:
GOOGLE_SPREADHSEEET_ID=xYz make run-import-from-google
where=
is a new param that accepts a JSON string and mimics the behaviour of the django Q
loopkup.
As this is quite powerful, we coupled with a strict JSON schema validation.
The where=
url param espects two types: either a list of lookup objects, e.g [{ "type": "entity"}]
; or an object defining complex operation, enabling "Op.or"
, "Op.and"
and "Op.not"
for a list of lookup objects, e.g. {"Op.not":[{ "type": "entity"}]}
.
Operators can be nested, so that very complex filters can be achieved:
/api/document/?where={"Op.not": [{ "Op.or": [{ "type": "entity"}, {"data__type": "drawing" }] }]}
resulting Q filter:
(NOT (AND: (OR: ('type', 'entity'), ('data__type', 'drawing'))))
In this case, the api returns all documents not being of "type":"entity"
or having "data__type":"drawing"
Note that in absence of the operator the concatenation is of type AND:
/api/document/?where=[{ "type": "entity"}, {"data__type": "drawing" }]
resulting Q filter:
(AND: ('type', 'entity'), ('data__type', 'drawing'))
In settings.py
file many variables allow fine tuning of the search engine.
Asu usual, variables in settings file can be overriden by environment variables.
SOLR_ENABLED=True \
SOLR_URL=http://localhost:8983/solr/miller \
pipenv run ./manage.py runserver
Don't forget to enrich your own docker-compose
file with the solr service and the same environment variables.
The solr endpoint is exposed in the api if SOLR_ENABLED=True
and SOLR_URL
is set.
By default, no action is being perdformed by the solr endpoint, but you can override the default behaviour by creating a ./miller/api/solr.py
file in your project root folder and implements the SolrViewset
class.
You should use a docker volume to override the default ./miller/api/solr.py
file:
volumes:
- ./miller/api/solr.py:/miller/miller/api/solr.py