Oxiland

Oxiland is a safe Rust reimplementation of the Redland librdf object model, powered by Oxigraph.

The project is working toward measurable Redland compatibility without carrying manual C memory management into its Rust API. Oxigraph provides the RDF and SPARQL engine; Oxiland supplies Redland-oriented concepts, behavior, migration paths, and—later in the 0.x series—a separately audited C compatibility layer.

[!IMPORTANT] Oxiland 0.1 provides the trusted core model surface. It does not yet provide complete Redland API accounting, differential behavioral parity, source, or ABI compatibility. See the parity ledger and 0.1 report.

Why Oxiland?

• Familiar Redland concepts with Rust ownership and error handling.
• Oxigraph-backed RDF terms, datasets, parsing primitives, and SPARQL.
• In-memory operation by default, with optional Fjall persistence.
• No unsafe code in the primary crate.
• Compatibility claims backed by an API inventory and differential tests rather than an unqualified percentage.
• A staged path toward existing C consumer support.

Current capabilities

“Available” means the current public workflow is implemented and tested. It does not imply full subsystem parity with Redland.

Requirements

• Rust 1.87 or newer

Oxiland pins Oxigraph 0.5.9 for the current release so compatibility testing is performed against a known engine version.

Installation

Add Oxiland to your Cargo.toml:

[dependencies]
oxiland = "0.1.0"

Quick start

use oxiland::terms::{Literal, NamedNode, Triple};
use oxiland::{Model, Query, QueryResults};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let model = Model::new()?;

    model.add(Triple::new(
        NamedNode::new("https://example.com/alice")?,
        NamedNode::new("https://example.com/name")?,
        Literal::new_simple_literal("Alice"),
    ))?;

    let result = Query::new("ASK { ?s ?p ?o }").execute(&model)?;
    assert!(matches!(result, QueryResults::Boolean(true)));

    Ok(())
}

Model::add returns true when it inserts a new statement and false when the same statement already exists.

Run this example from the repository with:

cargo run --example quick_start

Contexts and pattern matching

Redland contexts map to Oxigraph graph names:

use oxiland::terms::{GraphName, Literal, NamedNode, Triple};
use oxiland::{Model, StatementPattern};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let model = Model::new()?;
    let subject = NamedNode::new("https://example.com/alice")?;
    let graph = NamedNode::new("https://example.com/people")?;

    model.add_to_graph(
        Triple::new(
            subject.clone(),
            NamedNode::new("https://example.com/name")?,
            Literal::new_simple_literal("Alice"),
        ),
        GraphName::NamedNode(graph),
    )?;

    let matches = model
        .find(StatementPattern {
            subject: Some(subject.as_ref().into()),
            ..StatementPattern::default()
        })
        .collect::<Result<Vec<_>, _>>()?;

    assert_eq!(matches.len(), 1);
    Ok(())
}

Model::find returns a streaming StatementMatches iterator over a store snapshot (ADR-005).

The complete example is runnable with cargo run --example contexts.

Persistent storage

Model::open stores quads in a Fjall keyspace and keeps an Oxigraph in-memory working set for querying:

use oxiland::Model;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let model = Model::open("./data/oxiland-store")?;
    assert!(model.is_empty()?);
    Ok(())
}

On-disk compatibility and migration guarantees are not yet stabilized. Do not treat 0.x persistent stores as archival formats without an independent export.

Relationship to Redland

Oxiland targets the public librdf 1.0.17 API and the reference manual labeled 1.0.18. Raptor and Rasqal behavior is included only when exposed through a public librdf workflow.

Compatibility is reported at separate levels:

1. Concept parity — Redland workflows have Rust equivalents.
2. Safe API accounting — every public Redland item is mapped or classified.
3. Behavioral parity — equivalent operations match native Redland fixtures.
4. C source compatibility — supported C programs compile against Oxiland.
5. C ABI compatibility — supported existing binaries can load the compatibility library.
6. Downstream compatibility — selected real consumers pass unchanged.

This distinction matters: safe Rust may replace allocation functions idiomatically while the future C layer must still reproduce their observable ownership behavior. Details live in the compatibility plan.

Architecture

The primary oxiland crate remains safe and delegates standards-heavy work to Oxigraph:

Rust application
      │
      ▼
Oxiland safe facade ──> Oxigraph RDF, storage, I/O, and SPARQL
      ▲
      │
Future oxiland-capi
      ▲
      │
Existing C application

The future oxiland-capi crate will isolate opaque handles, strings, allocators, callbacks, panic containment, and other audited unsafe code. The safe crate has #![forbid(unsafe_code)].

See the architecture plan and decision log for the full boundaries and open decisions.

Roadmap

Milestones are evidence-gated rather than date-gated. The detailed 0.x roadmap defines deliverables, dependencies, exclusions, and release criteria.

Project documentation

• Planning index
• Changelog
• Parity ledger
• 0.1 compatibility report
• 0.x roadmap
• Execution plan and current backlog
• Architecture
• Compatibility contract
• Verification and release gates
• Architecture decisions
• Risk register
• 0.1 inventory

The parity ledger describes what exists now. Planning documents describe the intended path and must not be read as implemented functionality.

Development

Run the default local checks:

cargo fmt --check
cargo clippy --all-targets -- -D warnings
cargo test
cargo doc --no-deps
python3 scripts/check-inventory.py
scripts/generate-public-api.sh check

CI runs these checks on stable and Rust 1.87.

Compatibility work should be implemented as a vertical slice: inventory mapping, public API, implementation, positive and failure tests, differential evidence where applicable, and parity-ledger updates. The execution plan defines readiness and completion.

Contributing

Contributions are welcome while the API is evolving. Before proposing a broad facade, storage, or FFI change:

1. identify the affected Redland subsystem or inventory entries;
2. read the relevant architecture and compatibility decisions;
3. define observable behavior and unsupported cases;
4. include tests through the public Oxiland API;
5. update the parity ledger and planning evidence.

The most valuable current tasks are listed in the 0.2 backlog.

License

Oxiland is licensed under either of:

• Apache License, Version 2.0
• MIT License

at your option.