TAtango12·Oct 5, 2018
SQL could actually be an API language without needing to be wrapped in JSON/GraphQL. Building a GraphQL backend makes you deal with a GraphQL AST, parse it, process it and so on. GraphQL backends are hard to build because naive implementations can't "optimise" the AST and resolve to making individual data calls for each node in the AST. But considering the tooling around GraphQL on the server and client side, it would have been amazing to have that API tooling around SQL! Assuming that writing an SQL parser, hooking into the AST to run the business logic or to add auth rules, and then use great client side libraries like jooq [1], massivejs [2] to be able to construct SQL queries programmatically, SQL as an API could have been a very feasible and neat idea!
The similarities between GraphQL and SQL are many on the surface. Another nice par
Read the rest
allel that the blogpost doesn't mention is the idea of query variables in GraphQL which is reminiscent of prepared statements + variables in Postgres.
We believe in a similar thesis at Hasura[3], where the idea is to give app developers GraphQL on SQL (Postgres) with minimal abstraction. However, we've had to do a lot of work to support "application level" authorization so that read/write permissions are controlled through the applications auth system and make it a more complete solution.
In fact, our initial version was basically a JSON DSL very similar to datasette, and we switched to GraphQL primarily because the client-side tooling is quite amazing.
[1] https://www.jooq.org/
[2] https://github.com/dmfay/massive-js
[3] https://github.com/hasura/graphql-engine