PostgREST serves a fully RESTful API from any existing PostgreSQL database. It provides a cleaner, more standards-compliant, faster API than you are likely to write from scratch.
Demo postgrest.herokuapp.com | Watch Video
Try making requests to the live demo server with an HTTP client such as postman. The structure of the demo database is defined by begriffs/postgrest-example. You can use it as inspiration for test-driven server migrations in your own projects.
Download the binary (OS X / Linux) and invoke like so:
postgrest --db-host localhost --db-port 5432 \
--db-name my_db --db-user postgres \
--db-pass foobar --db-pool 200 \
--anonymous postgres --secure \
--port 3000TLDR; subsecond response times for up to 2000 requests/sec on Heroku free tier. (see the load test)
If you're used to servers written in interpreted languages (or named after precious gems), prepare to be pleasantly surprised by PostgREST performance.
Three factors contribute to the speed. First the server is written in Haskell using the Warp HTTP server (aka a compiled language with lightweight threads). Next it delegates as much calculation as possible to the database including
- Serializing JSON responses directly in SQL
- Data validation
- Authorization
- Combined row counting and retrieval
- Data post in single command (
returning *)
Finally it uses the database efficiently with the Hasql library by
- Reusing prepared statements
- Keeping a pool of db connections
- Using the Postgres binary protocol
- Being stateless to allow horizontal scaling
Ultimately the server (when load balanced) is constrained by database performance. This may make it inappropriate for very large traffic load. To learn more about scaling with Heroku and Amazon RDS see the performance guide.
Other optimizations are possible, and some are outlined in the Future Features.
PostgREST handles authentication (HTTP Basic over SSL) and delegates authorization to the role information defined in the database. This ensures there is a single declarative source of truth for security. When dealing with the database the server assumes the identity of the currently authenticated user, and for the duration of the connection cannot do anything the user themselves couldn't.
Postgres 9.5 will soon support true row-level security. In the meantime what isn't yet implemented can be simulated with triggers and security-barrier views. Because the possible queries to the database are limited to certain templates using leakproof functions, the trigger workaround does not compromise row-level security.
For example security patterns see the security guide.
A robust long-lived API needs the freedom to exist in multiple versions. PostgREST supports versioning through HTTP content negotiation. Requests for a certain version translate into switching which database schema to search for tables. PostgreSQL schema search paths allow tables from earlier versions to be reused verbatim in later versions.
To learn more, see the guide to versioning.
Rather than writing and maintaining separate docs yourself let the API explain its own affordances using HTTP. All PostgREST endpoints respond to the OPTIONS verb and explain what they support as well as the data format of their JSON payload.
The number of rows returned by an endpoint is reported by - and limited with - range headers. More about that.
There are more opportunities for self-documentation listed in Future Features.
Rather than relying on an Object Relational Mapper and custom imperative coding, this system requires you put declarative constraints directly into your database. Hence no application can corrupt your data (including your API server).
The PostgREST exposes HTTP interface with safeguards to prevent surprises, such as enforcing idempotent PUT requests, and
See examples of Postgres constraints and the guide to routing.
- Watching endpoint changes with sockets and Postgres pubsub
- Specifying per-view HTTP caching
- Inferring good default caching policies from the Postgres stats collector
- Generating mock data for test clients
- Maintaining separate connection pools per role to avoid "set/reset role" performance penalty
- Describe more relationships with Link headers
- Depending on accept headers, render OPTIONS as RAML or a relational diagram
- Add two-legged auth with OAuth 1.0a(?)
- ... the other issues
- Adam Baker for code contributions and many fundamental design discussions
- Nikita Volkov for writing the wonderful Hasql library and helping me use it
- Mikey Casalaina for the cool logo