Rust

Rust client for ChronDB, a time-traveling key/value database built on Git architecture.

Requirements

• Rust 1.56+ (2021 edition)

• libclang (for bindgen to generate FFI bindings)

Stack Size (Handled Automatically)

ChronDB uses GraalVM native-image with Apache Lucene and JGit, which require deep call stacks (~64MB).

This is handled automatically. The Rust binding spawns a dedicated worker thread with a 64MB stack for all FFI operations. You do not need to configure RUST_MIN_STACK, ulimit, or any other stack settings.

The worker thread architecture also provides:

• Consistent stack size across all platforms

• No special CI configuration required

• Thread-safe operation via message passing

Installation

Add chrondb from crates.io:

cargo add chrondb

Or add it directly to your Cargo.toml:

[dependencies]
chrondb = "*"
serde_json = "1"

Native shared library

The crate requires the libchrondb native library at runtime. Download it from the latest GitHub release:

macOS (Apple Silicon):

curl -L https://github.com/avelino/chrondb/releases/download/latest/libchrondb-latest-macos-aarch64.tar.gz | tar xz

Linux (x86_64):

curl -L https://github.com/avelino/chrondb/releases/download/latest/libchrondb-latest-linux-x86_64.tar.gz | tar xz

Configure the runtime library path

The shared library must be discoverable at runtime:

Linux:

export LD_LIBRARY_PATH=/path/to/chrondb-rust/lib:$LD_LIBRARY_PATH

macOS:

export DYLD_LIBRARY_PATH=/path/to/chrondb-rust/lib:$DYLD_LIBRARY_PATH

Library path (advanced)

To override the library location at build time, set CHRONDB_LIB_DIR:

export CHRONDB_LIB_DIR=/path/to/dir/with/libchrondb

Quick Start

use chrondb::ChronDB;
use serde_json::json;

fn main() -> chrondb::Result<()> {
    // Single path (preferred)
    let db = ChronDB::open_path("./mydb")?;

    // Save a document
    db.put("user:1", &json!({"name": "Alice", "age": 30}), None)?;

    // Retrieve it
    let doc = db.get("user:1", None)?;
    println!("{}", doc); // {"name":"Alice","age":30}

    Ok(())
    // db is automatically closed via Drop
}

Legacy API (deprecated): ChronDB::open("/tmp/data", "/tmp/index") still works but is deprecated. Use open_path instead.

API Reference

ChronDB::open_path(db_path) -> Result<ChronDB>

Opens a database connection using a single directory path. Data and index are stored in subdirectories automatically.

Parameter	Type	Description
db_path	&str	Path for the database (data and index stored inside)

Returns: Result<ChronDB>

Errors: IsolateCreationFailed, OpenFailed(reason)

---

ChronDB::open(data_path, index_path) -> Result<ChronDB> (deprecated)

Deprecated. Use open_path instead. This method still works but will be removed in a future release.

Opens a database connection with separate data and index paths.

Parameter	Type	Description
data_path	&str	Path for the Git repository (data storage)
index_path	&str	Path for the Lucene index

Returns: Result<ChronDB>

Errors: IsolateCreationFailed, OpenFailed(reason)

---

ChronDB::builder(data_path, index_path) -> ChronDBBuilder

Creates a builder for opening a database with custom options like idle timeout.

Parameter	Type	Description
data_path	&str	Path for the Git repository (data storage)
index_path	&str	Path for the Lucene index

Returns: ChronDBBuilder

ChronDBBuilder::idle_timeout(duration) -> Self

Sets the idle timeout for the GraalVM isolate. When no operations happen for the specified duration, the isolate is automatically closed to free CPU and memory. The next operation transparently reopens it.

Parameter	Type	Description
duration	Duration	Max idle time before suspending the isolate

ChronDBBuilder::open() -> Result<ChronDB>

Opens the database with the configured options.

Errors: IsolateCreationFailed, OpenFailed(reason)

Example

use chrondb::ChronDB;
use std::time::Duration;

let db = ChronDB::builder("/tmp/data", "/tmp/index")
    .idle_timeout(Duration::from_secs(120))
    .open()?;

// Use normally — isolate suspends after 120s of inactivity,
// reopens transparently on next operation.
db.put("user:1", &json!({"name": "Alice"}), None)?;

---

put(&self, id, doc, branch) -> Result<serde_json::Value>

Saves a document.

Parameter	Type	Description
id	&str	Document ID (e.g., "user:1")
doc	&serde_json::Value	Document data as JSON value
branch	Option<&str>	Branch name (None for default)

Returns: The saved document as serde_json::Value.

Errors: OperationFailed(reason)

---

get(&self, id, branch) -> Result<serde_json::Value>

Retrieves a document by ID.

Parameter	Type	Description
id	&str	Document ID
branch	Option<&str>	Branch name

Returns: The document as serde_json::Value.

Errors: NotFound, OperationFailed(reason)

---

delete(&self, id, branch) -> Result<()>

Deletes a document by ID.

Parameter	Type	Description
id	&str	Document ID
branch	Option<&str>	Branch name

Returns: Ok(()) on success.

Errors: NotFound, OperationFailed(reason)

---

list_by_prefix(&self, prefix, branch) -> Result<serde_json::Value>

Lists documents whose IDs start with the given prefix.

Parameter	Type	Description
prefix	&str	ID prefix to match
branch	Option<&str>	Branch name

Returns: JSON array of matching documents (empty array if none).

---

list_by_table(&self, table, branch) -> Result<serde_json::Value>

Lists all documents in a table.

Parameter	Type	Description
table	&str	Table name
branch	Option<&str>	Branch name

Returns: JSON array of matching documents (empty array if none).

---

history(&self, id, branch) -> Result<serde_json::Value>

Returns the change history of a document.

Parameter	Type	Description
id	&str	Document ID
branch	Option<&str>	Branch name

Returns: JSON array of history entries (empty array if none).

---

query(&self, query, branch) -> Result<serde_json::Value>

Executes a query against the Lucene index.

Parameter	Type	Description
query	&serde_json::Value	Query in Lucene AST format
branch	Option<&str>	Branch name

Returns: Results object with results, total, limit, offset.

Errors: OperationFailed(reason)

---

execute_sql(&self, sql, branch) -> Result<serde_json::Value>

Executes a SQL query directly against the database without needing a running server.

Parameter	Type	Description
sql	&str	SQL query string
branch	Option<&str>	Branch name (None for default)

Returns: Result object with type, columns, rows, count.

Errors: OperationFailed(reason)

---

last_error(&self) -> Option<String>

Returns the last error message from the native library, if any.

---

Drop

The ChronDB struct implements Drop. When it goes out of scope, the database connection is closed and the GraalVM isolate is torn down automatically.

Document ID Convention

Documents use the format table:id:

db.put("user:123", &json!({"name": "Alice"}), None)?;
db.put("order:456", &json!({"total": 99.90}), None)?;

Internally, user:123 is stored as user/user_COLON_123.json in the Git repository.

Use list_by_table("user") to retrieve all documents in the user table.

Error Handling

ChronDBError Enum

pub enum ChronDBError {
    IsolateCreationFailed,   // GraalVM isolate could not be created
    OpenFailed(String),      // Database failed to open (with reason)
    CloseFailed,             // Database failed to close
    NotFound,                // Document does not exist
    OperationFailed(String), // Operation failed (with reason)
    JsonError(String),       // JSON serialization/deserialization error
}

All variants implement Display and std::error::Error.

The crate also provides a type alias:

pub type Result<T> = std::result::Result<T, ChronDBError>;

Conversion

serde_json::Error is automatically converted to ChronDBError::JsonError via From.

Example

use chrondb::{ChronDB, ChronDBError};

fn main() {
    let db = ChronDB::open_path("./mydb").unwrap();

    match db.get("user:999", None) {
        Ok(doc) => println!("Found: {}", doc),
        Err(ChronDBError::NotFound) => println!("Document does not exist"),
        Err(e) => eprintln!("Error: {}", e),
    }
}

Examples

Full CRUD

use chrondb::ChronDB;
use serde_json::json;

fn main() -> chrondb::Result<()> {
    let db = ChronDB::open_path("./mydb")?;

    // Create
    db.put("user:1", &json!({"name": "Alice", "email": "[email protected]"}), None)?;
    db.put("user:2", &json!({"name": "Bob", "email": "[email protected]"}), None)?;

    // Read
    let alice = db.get("user:1", None)?;

    // Update
    let mut updated = alice.clone();
    updated["age"] = json!(30);
    db.put("user:1", &updated, None)?;

    // Delete
    db.delete("user:2", None)?;

    // List by table
    let users = db.list_by_table("user", None)?;
    println!("Users: {}", users);

    // List by prefix
    let matched = db.list_by_prefix("user:1", None)?;
    println!("Matched: {}", matched);

    Ok(())
}

Query

use chrondb::ChronDB;
use serde_json::json;

fn main() -> chrondb::Result<()> {
    let db = ChronDB::open_path("./mydb")?;

    db.put("product:1", &json!({"name": "Laptop", "price": 999}), None)?;
    db.put("product:2", &json!({"name": "Mouse", "price": 29}), None)?;

    let results = db.query(&json!({
        "type": "term",
        "field": "name",
        "value": "Laptop"
    }), None)?;

    println!("Total: {}", results["total"]); // 1

    Ok(())
}

History (Time Travel)

use chrondb::ChronDB;
use serde_json::json;

fn main() -> chrondb::Result<()> {
    let db = ChronDB::open_path("./mydb")?;

    db.put("config:app", &json!({"version": "1.0"}), None)?;
    db.put("config:app", &json!({"version": "2.0"}), None)?;

    let entries = db.history("config:app", None)?;
    println!("History: {}", entries);

    Ok(())
}

SQL Queries

Execute SQL queries directly without needing a running server:

use chrondb::ChronDB;
use serde_json::json;

fn main() -> chrondb::Result<()> {
    let db = ChronDB::open_path("./mydb")?;

    db.put("user:1", &json!({"name": "Alice", "age": 30}), None)?;
    db.put("user:2", &json!({"name": "Bob", "age": 25}), None)?;

    let result = db.execute_sql("SELECT * FROM user", None)?;
    println!("Columns: {}", result["columns"]);
    println!("Count: {}", result["count"]);

    let result = db.execute_sql("SELECT * FROM user WHERE name = 'Alice'", None)?;
    println!("Rows: {}", result["rows"]);

    Ok(())
}

Using with Drop (automatic cleanup)

use chrondb::ChronDB;
use serde_json::json;

fn do_work() -> chrondb::Result<()> {
    let db = ChronDB::open_path("./mydb")?;
    db.put("temp:1", &json!({"data": "value"}), None)?;
    Ok(())
    // db is dropped here, closing the connection
}

Idle Timeout (long-running services)

ChronDB loads a GraalVM native-image shared library whose internal threads consume CPU even when no operations are in flight. For long-running services with sporadic database access, use idle_timeout to automatically suspend the isolate when idle:

use chrondb::ChronDB;
use serde_json::json;
use std::time::Duration;

fn main() -> chrondb::Result<()> {
    // Isolate suspends after 2 minutes of inactivity
    let db = ChronDB::builder("/tmp/data", "/tmp/index")
        .idle_timeout(Duration::from_secs(120))
        .open()?;

    // Normal usage — no change in API
    db.put("audit:1", &json!({"action": "login"}), None)?;
    let doc = db.get("audit:1", None)?;

    // After 120s without operations, the GraalVM isolate is torn down.
    // The next call to put/get/query transparently reopens it.

    Ok(())
}

When to use:

• Long-running daemons that write to ChronDB intermittently (e.g., audit logging, MCP servers)

• Services where memory/CPU usage matters during idle periods

When NOT to use:

• Short-lived CLI tools (just use ChronDB::open)

• High-throughput services with constant database access (the isolate would never go idle)

Building from Source

# 1. Build the shared library (requires Java 11+ and GraalVM)
cd chrondb/
clojure -M:shared

# 2. Build the Rust binding
cd bindings/rust/
CHRONDB_LIB_DIR=../../target cargo build

# 3. Run tests
CHRONDB_LIB_DIR=../../target \
  LD_LIBRARY_PATH=../../target \
  cargo test

build.rs Behavior

The build script (build.rs):

1. Reads CHRONDB_LIB_DIR (defaults to ../../target)

2. Configures rustc-link-search and rustc-link-lib=dylib=chrondb

3. If libchrondb.h and graal_isolate.h exist in that directory, generates FFI bindings via bindgen

4. Otherwise, uses stub bindings (src/ffi_stub.rs) to allow compilation without the native library