Rig.TUnit.Databases.Sql.Oracle 0.1.0-beta.2

Rig.TUnit.Databases.Sql.Oracle

Testcontainers-backed Oracle fixture using gvenzl/oracle-free:23.5-slim-faststart and Oracle.EntityFrameworkCore.

What this package is

An Oracle Free integration fixture. OracleFixture spins the gvenzl/oracle-free:23.5-slim-faststart image — the fastest-to-boot Oracle container variant (~60–90 s warm, considerably more on first pull) — and exposes a working connection string plus an EF Core extension (UseOracle) for the Oracle.EntityFrameworkCore provider.

Integrates with Rig.TUnit.Databases.Sql's family contract so the same semantic assertions run across MySql / Postgres / Oracle / SqlServer / Sqlite.

When to use it

• Integration tests targeting Oracle-specific features (sequences, PL/SQL, MERGE, RETURNING INTO).
• Multi-engine parity testing where Oracle is a required cell.
• Verifying behaviour under the OCP licence — Oracle EF Core is free.
• Not for: unit tests. Oracle's session setup is the slowest in the family; prefer SQLite in-memory for fast feedback loops.

Prerequisites

• .NET 10 SDK
• Docker Desktop / Colima (first pull ~2 GB)
• Oracle.EntityFrameworkCore transitively included via this package.

Quick start

using Microsoft.EntityFrameworkCore;
using Rig.TUnit.Databases.Sql.Oracle.Extensions;
using Rig.TUnit.Databases.Sql.Oracle.Fixtures;

await using var fx = new OracleFixture();
await fx.InitializeAsync();

var opts = new DbContextOptionsBuilder<TestDb>()
 .UseOracle(fx.ConnectionString)
 .Options;

Options

Fixture + helper APIs

• Rig.TUnit.Databases.Sql.Oracle.Fixtures.OracleFixture
• Rig.TUnit.Databases.Sql.Oracle.Options.OracleFixtureOptions
• Rig.TUnit.Databases.Sql.Oracle.Builder.OracleRigBuilder
• UseOracle(RigBuilder, …) extension
• UseOracle(DbContextOptionsBuilder, string) — EF wiring

Per-test isolation

Oracle uses schema-per-test via IsolationKey-derived user names. Tests create CREATE USER {iso} IDENTIFIED BY … in Arrange and DROP USER {iso} CASCADE in teardown. Because CREATE USER requires session privs, the fixture pre-provisions a test-DBA schema.

Parallelism + performance

• First-run container pull: ~3–5 min (large image).
• Warm startup: ~60–90 s.
• Per-test schema create: ~1–2 s (Oracle's session setup is the slowest in the SQL family).
• Under parallel execution: cap Iterations at 2–4 — more will exhaust Oracle's default processes limit (150).

Troubleshooting

• ORA-00020: maximum number of processes exceeded — reduce Iterations in your ParallelIsolationContract subclass, or configure the container with -e ORACLE_PROCESSES=500.
• Startup timeout — gvenzl/oracle-free:23.5-slim-faststart is the fastest variant; oracle-free:23.5-slim is 2× slower; the full oracle-free:23.5 is 5× slower.

See .

Provider quirks + edge cases

• Oracle identifiers are case-sensitive ONLY when quoted. EF Core quotes by default — expect "Orders" vs orders distinctness.
• Max identifier length 30 chars (23c extends to 128 but only when COMPATIBLE=12.2 is disabled; Free ships with it enabled).
• NUMBER with no precision maps to decimal(38,0) by default — specify HasPrecision(…) in OnModelCreating to avoid silent truncation.
• TIMESTAMP WITH LOCAL TIME ZONE behaviour differs from WITH TIME ZONE — Oracle stores UTC for the former, offset for the latter.

Benchmarks

See ; baseline in benchmarks/baseline-005.json. Oracle's per-test cost is the largest in the SQL family — tracked closely.

Related docs

• Family base:

License

MIT. See .