Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/fabienarcellier/spike-orm-python-typescript

use graphql service to expose and manipulate data from a database
https://github.com/fabienarcellier/spike-orm-python-typescript

Last synced: 25 days ago
JSON representation

use graphql service to expose and manipulate data from a database

Awesome Lists containing this project

README 

        
## Use graphql service to expose and manipulate data from a database



![CI](https://github.com/FabienArcellier/spike-orm-python-typescript/workflows/CI/badge.svg)



the goal of this spike is to validate a backend architecture based on GraphQL API and ORM for postgresql.



* python stack : ariadne + sqlalchemy

* typescript stack : apollo server + prisma



I have used [sportsdb](http://sportsdb.org/sd) as datasource to populate postgresql database.



![](docs/datastore_view.png)



## references



## step 1 : Run the datastore and import sql dumps



I have used the convention on `postgresql` docker image. We can put sql dump in `/docker-entrypoint-initdb.d` directory. The container will load

the image when it start first.



* https://hub.docker.com/_/postgres : see section `Initialization scripts`



## step 2 : Ensure the datastore can be running in github action



I have used `docker-compose` to package this spike environment.

*Github Action* can load docker-compose definition on the image `ubuntu`.



* take a look on [github action definition](.github/workflows/main.yml)



I develop a script to check the datastore is up and running on `Github Action` before moving next

step.



* take a look on [wait_datastory.py](ci_tests/ci_tests/wait)



## step 3 : Implement ORM stack with sqlalchemy and ariadne



There is an automatic mapping between the SQLAlchemy item, Ariadne and the schema.

I don't have to write api entity for every class. I just have to declare the query and mutation.

An automatic cast is done when we change the type from the database.



* see [query.py](sportsdb_backend_python/sportsdb_backend/api/query.py)



The `uvicorn server` is not resilient on query error. I have to investigate it. If the query is not done properly on the database, the server is not able to

respond to any other query. I think the session of sqlalchemy has been corrupted. I have to investigate a way to monitor and renew the sqlalchemy session.



I have used `sqlacodegen` to build [entities](sportsdb_backend_python/sportsdb_backend/entities.py) from the database.

The result is ok, but I think the automapper approach is more interesting.



### step 3.1 : implement probes and utility url with ariadne



I didn't find a way to add custom route on ariadne. In `asgi.py`, there is a direct mapping on the application and the handler for http request.



Two scenarios in this case :



* use `starlette` as webapplication over `ariadne` : https://ariadnegraphql.org/docs/starlette-integration

* use a webserver (`nginx`) as reverse proxy in front to route the call to graphql based on url mapping as `/api`.



## step 4 : Implement ORM stack with Prisma and Apollo Server



The generation of client on prisma est native. It works fine. There is an automatic mapping between Prisma and

Apollo Server and the schema. I don't have to write api entity for every class. I just have to declare the query and mutation.

An automatic cast is done when we change the type from the database.



Prisma client in my IDE is not useful because there is no autocomplete on the column, only on the object. I don't have autocomplete

on the CRUD operation of Prisma. According to what I read on internet, I should have it.



To override the connection string of database, I had to rewrite the variable `DATABASE_URL`. I can't inject it in Prisma Client.

I lost lot of time finding resources about that on internet.