Cvoya.Graph.InMemory 1.0.0

Downloadable open-source computer software from CVOYA. See the CVOYA software catalog.

Cvoya.Graph.InMemory

An in-process, in-memory provider for the CVOYA graph abstraction. It exists for two audiences:

• Application developers who want a fast unit-test double: test business logic against IGraph with no database and no containers.
• Provider implementers who want an executable specification: this provider passes the Cvoya.Graph.CompatibilityTests suite by interpreting the shared provider-neutral query model (GraphQueryModel) with LINQ-to-objects — no query language anywhere — which is the working proof that the level-1 query model is dialect-free.

Usage

using Cvoya.Graph;
using Cvoya.Graph.InMemory;

await using var store = new InMemoryGraphStore();
var graph = store.Graph;

await graph.CreateNodeAsync(new Person { Name = "Alice" });
var alice = await graph.Nodes<Person>().Where(p => p.Name == "Alice").SingleAsync();

Everything on IGraph behaves per the public contract: CRUD, LINQ querying (including traversals and path segments), transactions, streaming, and cancellation. Complex properties are stored decomposed as owned value nodes linked by property-named relationships, exactly like the Neo4j provider, so cascade and hydration semantics are real, not object aliasing. Entities round-trip through the serialization layer on every write and read: mutating an entity you hold never mutates the store.

Not a production store

• Data lives in process memory; nothing is persisted.
• Queries are unindexed scans; there is no query optimization.
• RecreateManagedIndexesAsync is a successful no-op because the provider owns no index artifacts.
• Concurrency model: transactions buffer their writes and commit atomically under a single store-wide lock (single-writer serialized commits; reads take lock-free snapshots). This is simple and correct for a test double, not a throughput design.
• Full-text search (GraphCapability.FullTextSearch) is supported at the contract floor only: naive, index-free, whole-word matching over each entity's own searchable string properties. No stemming, ranking, prefix, or substring matching.

See docs/testing-with-the-in-memory-provider.md in the repository for the full guide.