Skip to content
gr

Roadmap

What gr 1.0 ships, what is deferred to post-1.0, and what is outside the project's scope.

The full roadmap is in ROADMAP.md in the repository. This page is a summary.

What v1.0 ships

  • Embedded Go library (github.com/tamnd/gr) with openCypher read and write.
  • gr CLI: interactive shell, one-shot query runner, bulk importer, backup/restore, schema inspection, EXPLAIN/PROFILE.
  • gr serve: Bolt v4.4/v5.0 wire protocol and HTTP JSON API, compatible with Neo4j drivers.
  • Stable library API (148 exported names), frozen configuration surface (51 knobs), versioned file format, documented wire protocol.
  • WAL journaling, per-page checksums, crash recovery, hot backup.
  • Asynchronous read-replica WAL shipping.
  • Honest TCK conformance statement and deviations registry.

Post-1.0 deferrals

These are deferred, not declined — each has a seam already in place.

Concurrent writers (high priority). The 1.0 MVCC model serializes write transactions. Post-1.0, WithConcurrentWriters(true) enables parallel writes through the existing MVCC version model.

Replication maturity (high priority). v1.0 ships async read-replica WAL shipping. Post-1.0: synchronous/quorum replication, writable replicas, automatic failover.

Broader language coverage (medium–high). v1.0 covers a hardened subset with an honest deviations registry. Post-1.0: remaining TCK scenarios and a measured path toward GQL.

Graph-algorithms library (medium). Post-1.0: PageRank, community detection, and centrality metrics as CALL-able procedures.

Compiled/JIT execution (low–medium). Post-1.0: a compiled or JIT path, likely LLVM-backed or WebAssembly-targeted.

Richer authorization (low–medium). Post-1.0: fine-grained attribute-level authorization and multi-tenant isolation.

Richer import sources (low). Post-1.0: Arrow, JSON-lines, and Avro in addition to CSV.

Non-goals

These are outside the project's scope by identity:

  • Clustering and sharding. gr is a single-node embedded database. Horizontal sharding and distributed query execution are a different product.
  • SPARQL and RDF. gr stores a labeled-property graph and queries it with Cypher. RDF's triple model and SPARQL are a different design.
  • SQL. The query language is Cypher.
  • Mandatory schema. gr is schema-optional by design.
  • Adjacent-shape stores. Document, key-value, time-series, and vector-only databases are not gr and do not become gr.

Tracking progress

The TCK deviations registry (tck/deviations.yaml) records every scenario skipped with a reason code. As post-1.0 features land, their scenarios move to the passing corpus and CONFORMANCE.md updates. Releases follow semantic versioning: minor versions add features additively; a major version changes a contract.

2 min read