quack_vs_postgres.yaml

meta:
  name: "Quack vs Postgres: Client-Server Head-to-Head"
  description: >
    The apples-to-apples question the execution-modes experiment (capsule
    3e2fe152) deliberately avoided: if you want CLIENT-SERVER analytics,
    is DuckDB-over-Quack (beta, pushdown mode) already competitive with
    the incumbent client-server database? Three engines, identical data
    and SQL (analytical_wall), stock defaults, cold cache per query:
    duckdb in-process (reference floor), quack_pushdown (DuckDB
    client-server), postgres (the incumbent).
    Caveat for cross-engine fairness, disclosed up front: on macOS,
    Postgres runs inside Docker's VM and pays filesystem-virtualization
    tax that the native engines do not. Treat Postgres numbers here as
    macOS-Docker numbers; a Linux run is the clean comparison.

dataset:
  source: sql_benchmarks.plugins.data_sources.declarative_gen
  tables:
    analytical_data:
      rows: rows
      columns:
        - name: id
          provider: sequence
          primary_key: true
        - name: region
          provider: choice
          options: ["North", "South", "East", "West"]
        - name: category
          provider: choice
          options: ["A", "B", "C"]
        - name: price
          provider: random_float
          min_value: 10.0
          max_value: 500.0
        - name: quantity
          provider: random_int
          min_value: 1
          max_value: 100
        - name: discount
          provider: random_float
          min_value: 0.0
          max_value: 0.30

definitions:
  rows:
    tiny:   1_000
    small:  100_000
    medium: 1_000_000
    large:  10_000_000

execution:
  test_suite: analytical_wall
  engines:
    - duckdb
    - quack_pushdown
    - postgres
  replication: 5
  matrix:
    rows:
      - tiny
      - small
      - medium
      - large