# rust-store

> A single Rust core engine for multi-backend data access — GQL parsing, permission checks, computed columns, command planning and SQL dialect translation, exposed to Node.js and Python through native bindings.

`rust-store` is the engine behind [`nodejs-store`](https://github.com/coenddt/nodejs-store) and [`py-store`](https://github.com/coenddt/py-store). It holds **no database driver and performs no IO**: it compiles GQL (and writes) into backend-agnostic MongoDB command JSON for the host to execute, and translates those commands into parameterized SQL for MySQL / PostgreSQL / SQLite. Same semantics in Node.js and Python by construction.

- Crates: [`rust-store-core`](https://crates.io/crates/rust-store-core) (pure logic), `rust-store-node` (napi-rs → npm `rust-store-node`), `rust-store-py` (PyO3 → PyPI `rust-store-py`)
- Backends: MongoDB, MySQL, SQLite, PostgreSQL
- License: MIT

## When to use it

- Building, debugging or extending the `nodejs-store` / `py-store` hosts — this is where GQL parsing, permissions and dialect translation live
- Using the same query dialect from a Rust service (`cargo add rust-store-core`)
- Building a host/binding for another language (the core is language-agnostic; `core-node` and `core-py` are worked examples of the JSON command contract)
- Guaranteeing parity between a Node service and a Python service
- Moving permission and computed-column logic out of application code into the schema
- Building an AI / natural-language query layer (planning is a pure `gql → plan` function)

## Docs

- [Documentation site](https://coenddt.github.io/rust-store/): the scenario walkthroughs as individual, indexable pages — the fastest way in
- [crates.io: rust-store-core](https://crates.io/crates/rust-store-core): the pure-logic Rust library — `cargo add rust-store-core`; use it directly from a Rust service
- [docs.rs: rust-store-core](https://docs.rs/rust-store-core): auto-generated API reference for every public function
- [README](https://github.com/coenddt/rust-store#readme): architecture, API reference, GQL capabilities, dialects, permission model, parity testing, gotchas
- [Use cases](https://github.com/coenddt/rust-store/tree/main/doc/use-cases): six scenario walkthroughs — embed the core in a Rust service, natural language → query compiler, one dialect across four databases, Node.js / Python parity, reuse the permission model, SQL pushdown limits by dialect
- [Chinese README](https://github.com/coenddt/rust-store/blob/main/README.zh-CN.md): 中文文档
- [CHANGELOG](https://github.com/coenddt/rust-store/blob/main/CHANGELOG.md)

## Concepts

- [Core principles](https://github.com/coenddt/rust-store#what-it-is): no IO / no clock / no randomness; explicit over silent; JSON in, JSON out
- [API reference](https://github.com/coenddt/rust-store#api-reference): `Registry` plan methods, permissions, computes, dialect translation, federation
- [GQL capabilities](https://github.com/coenddt/rust-store#gql-capabilities): relations, `$group` / `$having`, relation aggregate predicates (semi/anti-join), computed columns
- [Backends and dialects](https://github.com/coenddt/rust-store#backends-and-dialects): the four targets and what is Mongo-only
- [Permission model](https://github.com/coenddt/rust-store#permission-model): schema / field / relation / computed-column level, `creator`, system context
- [Boundaries and gotchas](https://github.com/coenddt/rust-store#boundaries-and-gotchas): error prefixes, `__present` sentinel, U1–U4 rules, federation `degraded`

## Optional

- `text-to-query` skill: natural language → GQL + params
- [nodejs-store](https://github.com/coenddt/nodejs-store) / [py-store](https://github.com/coenddt/py-store): the two production hosts