#!/bin/bash
#
#  bench.sh — Metrika
#
#  Author:  Simon-Pierre Boucher
#  Contact: contact@spboucher.ai
#  Copyright © 2026 Simon-Pierre Boucher. All rights reserved.
#
#  Performance benchmark harness (CLAUDE.md §8). Generates a synthetic
#  dataset and times the core pipeline through metrika-cli. CI compares
#  against the budget table; regressions >10% fail the build.
#
#  Usage: Tests/Bench/bench.sh [rows]   (default 1,000,000)
#
set -euo pipefail
export LC_ALL=C

ROWS="${1:-1000000}"
ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
DATA="$(mktemp -d)/bench.csv"
trap 'rm -rf "$(dirname "$DATA")"' EXIT

echo "generating $ROWS rows…"
awk -v n="$ROWS" 'BEGIN {
  srand(42)
  print "revenue,price,region,firm_id"
  for (i = 1; i <= n; i++) {
    price = 5 + 15 * rand()
    region = (i % 3) + 1
    revenue = exp(4 - 0.08 * price + 0.1 * region + 0.15 * (rand() - 0.5))
    printf "%.6f,%.4f,%d,%d\n", revenue, price, region, (i % 500) + 1
  }
}' > "$DATA"

cd "$ROOT/MetrikaKit"
swift build -c release --product metrika-cli >/dev/null
CLI=".build/release/metrika-cli"

bench() { # bench <label> <command...>
  local label="$1"; shift
  local start end
  start=$(python3 -c 'import time; print(time.time())')
  "$CLI" "$@" >/dev/null
  end=$(python3 -c 'import time; print(time.time())')
  printf "%-28s %8.3f s\n" "$label" "$(echo "$end - $start" | bc)"
}

bench "use csv ($ROWS rows)" -e "use $DATA"
bench "use + summarize"       -e "use $DATA" -e "summarize"
bench "use + reg (robust)"    -e "use $DATA" -e "gen log_rev = ln(revenue)" \
                              -e "reg log_rev price, robust"
bench "use + bootstrap(1000)" -e "use $DATA" -e "keep in 1/100000" \
                              -e "gen log_rev = ln(revenue)" \
                              -e "bootstrap, reps(1000) seed(42): reg log_rev price"