Database models and migrations for the Local News Lab's article recommendation system
Project description
This branch houses the effort to move our DB code to a SQLAlchemy-based stack and is under active development in parallel with the move to Google Analytics 4. For the current production code, see the main
branch.
article-rec-db
Database models and migrations for the Local News Lab's article recommendation system.
Usage
A note before continuing: A lot of the commands you'll see below are wrapped inside Poe tasks defined in pyproject.toml
. Poe is installed as a dev-dependency; run poe --help
to get a list of tasks and their descriptions. It's entirely possible to run commands without using Poe, but if you decide to use Poe, make sure to read through the tasks to understand what they do.
As a package
We use SQLModel, a layer on top of SQLAlchemy with Pydantic, to define our tables. This is useful because we can import this package to interact with the tables AND have Pydantic objects in Python that correspond to a row in the table.
To install the package from PyPi, run: pip install article-rec-db
. Check existing versions
here.
Database management
We use Terraform to manage cluster entities such as databases, roles and extensions. The code is in the terraform
directory. Stages (dev and prod) are represented as different databases. To make changes to an existing database,
- Make changes inside
terraform/modules
. - Run
poe terraform [stage] plan
to see the changes that will be applied to the corresponding database. - At this point, if you're happy, you can run
poe terraform [stage] apply
yourself, but we prefer a CI/CD approach. Merging a PR to thedev
branch will trigger a plan to be applied to thedev
database, and the same for theprod
branch. We always merge todev
first, then do another merge fromdev
toprod
.
Table and column migrations
So you made some changes to what tables there are, what columns there are, indices, etc. and you'd like to update the databases. This is what alembic is for! (And notice the difference between Terraform and alembic: Terraform manages database entities that are not specific to a database, like roles and extensions, while alembic manages database entities that are specific to a database, like tables and columns.)
To generate a new revision after you've updated the models:
- Run this from the root of the project:
DB_CONNECTION_STRING='postgresql://user:password@host:port/db_name' alembic revision --autogenerate -m "message"
. (There's a Poe task for this: runpoe rmtdiff -d db_name -m "message"
) - Check the
/alembic/versions/
directory for the new revision and verify that it does what you want it to. RunTYPE=alembic poe test
to test models against a local DB initialized via Alembic, and resolve issues as needed. - Run this from the root of the project:
DB_CONNECTION_STRING='postgresql://user:password@host:port/db_name' alembic upgrade head
. Note that you only need to generate the revision file (step 1) once because we want the same content in each environment's database, but you do need to run theupgrade head
command once for each database (change the DB_NAME to the desired target). (There's a Poe task for this: runpoe rmtupgrade -d db_name
)
Similar to database management, we let our CI/CD handle Step 3.
Note to LNL devs: Our automated deployment process will run the Terraform changes first, then the Alembic changes. So, for example, using Terraform to create a new database and Alembic to create a new table in that database will work in just one PR, but creating a new table with Alembic and using Terraform to grant a role access to that table won't. Best to divide changes into atomic units, each handled by a single PR.
Development
This project uses Poetry to manage dependencies. It also helps with pinning dependency and python versions. We also use pre-commit with hooks for isort, black, and flake8 for consistent code style and readability. Note that this means code that doesn't meet the rules will fail to commit until it is fixed.
We use mypy for static type checking. This can be run manually,
and the CI runs it on PRs to the main
branch. We also use pytest to run our tests.
This can be run manually and the CI runs it on PRs to the main
branch.
Setup
- Install Poetry.
- Run
poetry install --no-root
- Run
source $(poetry env list --full-path)/bin/activate && pre-commit install && deactivate
to set uppre-commit
You're all set up! Your local environment should include all dependencies, including dev dependencies like black
.
This is done with Poetry via the poetry.lock
file.
Run Code Format and Linting
pre-commit run --all-files
runs isort, black, and flake8 all in one go, and is also run on every commit.
poe format
does what pre-commit run --all-files
does and also formats the Terraform code.
Run Static Type Checking
To manually run mypy, simply run mypy
from the root directory of the project. It will use the default configuration
specified in pyproject.toml
.
Update Dependencies
To update dependencies in your local environment, make changes to the pyproject.toml
file then run poetry update
from the root directory of the project.
To update Terraform dependencies, make changes to versions.tf
files as necessary.
Run Tests
To manually run rests, you need to have a Postgres instance running locally on port 5432. One way to do this is to run a Docker container, then run the tests while it is active.
- (If you don't already have the image locally) Run
docker pull ankane/pgvector:v<version used in your remote db>
- Run
docker run --rm --name postgres -e POSTGRES_PASSWORD=postgres -e POSTGRES_HOST_AUTH_METHOD=trust -p 127.0.0.1:5432:5432/tcp postgres
- Run
pytest tests
from the root directory of the project. Explore thepytest
docs (linked above)
Note that if you decide to run the Postgres container with different credentials (a different password, port, etc.) or via a different method, you will likely need to update the test file to point to the correct Postgres instance.
Additionally, if you want to re-run the tests, you want to make sure you start over from a fresh Postgres
instance. If you run Postgres via Docker, you can simply ctrl-C
to stop the image and start a new one.
Steps 2 and 3 can be combined into one Poe task: poe test
, which also stops the container after the tests are done, even if tests fail. In addition, you can also run poe lclstart
to just start the container, and poe lclstop
to stop it whenever you're done. poe lclconnect
will connect you to the container via psql
so you can poke around.
poe test
, by default, is equivalent to TYPE=sqlmodel poe test
, which tests the models against a local DB initialized via SQLModel. You can also run TYPE=alembic poe test
to test the models against a local DB initialized via Alembic. The first is more convenient and is good for development, the second reflects the production environment more closely and is good for testing Alembic revisions once you're about to submit a PR.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for article_rec_db-0.0.11-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | d706ec6cec3a5f7a90449a4abfe27b436d09423195abccd12cee42605d11d8fd |
|
MD5 | 08de2e44abb01b1a9e49b4c2df5b4a01 |
|
BLAKE2b-256 | 6400735aec05da1dc9ee6fbdb1f7735be62145e2663da73c439453aba3b4286e |