Benchmarks

Running

task benchmark          # Standard Go benchmarks (pglike/SQLite, :memory:)
task testBenchmark      # Standalone PG benchmarks (starts/stops Podman container)

Structure

benchmark_test.go           # Standard Go benchmarks (in-process, :memory:)
cmd/bench-*/                # Standalone programs for real PostgreSQL
benchmarks/
  analysis/                 # Narrative analysis per topic (purpose, findings, AI summary)
  reports/                  # Timestamped raw reports with data tables and SQL queries

Each benchmark topic has up to three source files (-purpose.md, -analysis.md, -ai-summary.md) which are combined into a single HTML page per topic by task docs:bench-combine.

Reports in benchmarks/reports/ include full SQL queries, schema, raw output, and results tables (TPS).

Writing Benchmarks

• Use :memory: for in-process benchmarks (no disk I/O noise)
• b.ResetTimer() after setup
• Sub-benchmarks (b.Run(...)) to vary parameters
• l.EnsureInterestAccounts() during setup for interest-related benchmarks
• Standalone PG benchmarks go in cmd/bench-*, wired into task testBenchmark

Benchmark Index

In-process (pglike/SQLite)

Standalone (real PostgreSQL)