Embedded JSON document database powered by a Rust core, with language wrappers via FFI or subprocess.
Zero external dependencies. No MySQL, PostgreSQL, SQLite, or PDO required. Just your filesystem and raw speed.
AnvilDB was born out of a real need: working in environments where you have no installation permissions on the operating system. Without being able to install MySQL, PostgreSQL, or additional extensions, I needed a lightweight database with zero external dependencies that could run on the filesystem alone.
The goal was to have something fast for scaffolding projects and prototyping ideas without friction — an implementation that is often temporary, but needs to work from the very first moment with no configuration or infrastructure. Just copy, use, and move on.
- Rust-powered core compiled as a native shared library (
.so/.dylib/.dll) - Joins — INNER and LEFT joins across collections via hash join (O(n+m))
- Write buffering — batched disk writes with configurable threshold and timer
- Compression — all data compressed on disk (deflate), transparent to the API
- Encryption at rest — optional AES-256-GCM, per-file nonce, key as hex string
- Lazy loading — collections loaded on first access, not at startup
- In-memory indexes (hash, unique, and range/BTreeMap) for sub-millisecond lookups
- Aggregations — sum, avg, min, max, count with optional group_by
- CSV export/import for data portability
- LRU cache with automatic invalidation on writes
- Atomic writes via temp file + rename to prevent corruption
- Schema validation to enforce document structure
- Bulk operations for efficient batch inserts
- Cross-platform precompiled binaries (Linux, macOS, Windows / x86_64, aarch64)
The core exposes a C API (anvildb.h) and a JSON-over-pipes protocol (anvildb-server) — any language can integrate via FFI or subprocess.
| Language | Package | Status |
|---|---|---|
| PHP | anvildb-php | Available |
Want to create a wrapper for another language? See the Wrapper Development Guide.
graph TD
App[Application] --> Wrapper[Language Wrapper]
Wrapper -->|FFI · direct| Core[Rust Core Engine]
Wrapper -->|Subprocess · stdin/stdout| Server[anvildb-server] --> Core
Core --> FS[("Filesystem (.anvil)")]
Core internals: Write Buffer · LRU Cache · In-memory Indexes (Hash, Unique, Range) · Query Engine (filter, join, sort, paginate) · Schema Validation · Codec (deflate + AES-256-GCM) · Atomic Storage
10,000 records | PHP 8.4 | Linux x86_64
| Operation | Time | Throughput |
|---|---|---|
| Bulk insert (10x1000) | 204ms | ~49k docs/s |
| Read all (10k docs) | 22ms | ~454k docs/s |
| Filter query | 4.6ms | — |
| Filter + sort + limit | 3.7ms | — |
| Count with filter | 0.2ms | — |
With compression, encryption, atomic writes, and schema validation active.
Full benchmark history: BENCHMARKS.md
anvildb/
├── core/ # Rust core engine
│ ├── src/ # Engine, FFI, storage, query, indexing
│ └── tests/ # Rust integration tests
├── wrappers/php/ # PHP FFI wrapper
│ ├── src/ # AnvilDb, Collection, Query, FFI, Exception
│ ├── tests/ # PHPUnit tests
│ └── docs/ # PHP-specific documentation
├── docs/ # Core documentation
├── benchmarks/ # Benchmark scripts
└── Cargo.toml # Workspace root
Full documentation at docs/index.md — development, architecture, C API reference, error codes, testing, CI/CD, wrapper development.
See CONTRIBUTING.md for guidelines.
AnvilDB is built on top of these excellent open-source projects:
| Crate | Description | License |
|---|---|---|
| serde / serde_json | Serialization framework & JSON support | MIT / Apache-2.0 |
| uuid | UUID generation (v4) | MIT / Apache-2.0 |
| miniz_oxide | Pure Rust deflate compression | MIT / Apache-2.0 |
| aes-gcm | AES-256-GCM authenticated encryption | MIT / Apache-2.0 |
| log | Logging facade | MIT / Apache-2.0 |
| env_logger | Log output to stderr (optional, dev only) | MIT / Apache-2.0 |
| getrandom | OS-level random number generation | MIT / Apache-2.0 |