Test Selection Rationale

Why these tests, and what each one reveals. For full test specifications, see specifications.md.

---

Test Overview

The suite contains 19 tests across 5 groups, each targeting a distinct performance dimension.

Entity (3 tests)

Test	Measures
create_empty	Entity allocation cost alone
create_with_components	Entity allocation with component attachment
destroy	Entity removal from a populated world

Component (4 tests)

Test	Measures
get	Field read speed on table components
set	Field write speed on table components
add	Component insertion cost (archetype transition)
remove	Component removal cost (archetype transition)

Tag (3 tests)

Test	Measures
has	Tag presence check cost
add	Tag insertion cost (structural change)
remove	Tag removal cost (structural change)

System (6 tests)

Test	Measures
throughput	Per-entity iteration cost (1 system, 100% match)
overlap	Scheduling efficiency with overlapping queries (3 systems)
fragmented	Iteration across 20 archetypes (2 systems)
chained	Data-dependency pipeline (4 systems, A → B → C → D → E)
multi_20	Per-system dispatch overhead (20 systems, 100% match)
empty_systems	Short-circuit cost with zero matching entities (20 systems)

Structural Scaling (3 tests)

Test	Measures
create	Entity creation cost with 20 registered systems
add_component	Component addition cost with 20 registered systems
destroy	Entity destruction cost with 20 registered systems

Three tests suffice because create, add, and destroy all trigger the same system-matching overhead. Additional operations (remove, create_empty, tags) would repeat the same scaling pattern without new information.

The group registers 20 background no-op systems (matching the multi_20 and empty_systems count) that query Position + Velocity — the same components the tracked entities carry. Comparing these results against the zero-system baselines in Parts 1-3 reveals how much each framework's structural cost grows with system count.