Performance Recommendations¶

Practical guidance — what to actually do. For the methodology and measured numbers behind these recommendations, see Performance.

Let source generation do its job¶

Compono's construction plans are generated at compile time specifically so composing a type at runtime costs a dispatch to generated code, not reflection (docs/adr/0001-source-generation-first.md). You don't need to do anything to get this benefit for an ordinary composed type — it's the default, not an opt-in, for any type reached through a recognized call site (Create<T>(), CreateMany<T>(), a [Compose] theory parameter). See The Composition Model for what makes a call site recognized.

Keep collection sizes proportional to what the test actually needs¶

A larger composed collection means more nested composition work, multiplied by every test that composes it. Set an explicit, smaller size for a specific member with For<T>().Member(x => x.Y).WithCollectionSize(n) rather than inflating the suite-wide default to accommodate one test's unusual need — see Collections for the global-vs-per-member distinction.

Don't reach for [Shared] to avoid recomposition¶

[Shared] exists to make a test assert against or configure a specific instance, not as a performance optimization — composing an ordinary value is cheap by design (see "Let source generation do its job" above). Reaching for [Shared] where independence was actually intended trades a real correctness property (two composed values of the same type are independent unless told otherwise) for a performance concern that generated composition doesn't actually have.

Measure before optimizing further¶

If a composed test suite's runtime genuinely becomes a bottleneck, profile it rather than guessing — the overwhelming majority of a test suite's wall-clock time in practice is I/O (a real HTTP call, a real database), not composition. benchmarks/Compono.Benchmarks (BenchmarkDotNet) is the same tool this project's own maintainers use to validate Compono's own construction performance, and the same discipline (measure the specific claim, don't extrapolate from a different one) applies to your own suite.

Next¶

• The measured numbers and methodology behind "generated construction avoids reflection overhead" → Performance.
• Collection composition in depth → Collections.
• Applies at scale → Large Test Suites.