#!/usr/bin/env Rscript
#
#  make_samples.R — Metrika
#
#  Author:  Simon-Pierre Boucher
#  Contact: contact@spboucher.ai
#  Copyright © 2026 Simon-Pierre Boucher. All rights reserved.
#
#  Generates the sample datasets bundled with the app (loaded via the
#  `sysuse` command). Deterministic: fixed seed, values rounded so the
#  CSVs are bit-stable across regenerations.
#

out <- file.path(
  dirname(dirname(normalizePath(sub("--file=", "", grep("--file=", commandArgs(FALSE), value = TRUE))))),
  "MetrikaKit", "Sources", "ZQEngine", "Resources", "Samples"
)
dir.create(out, recursive = TRUE, showWarnings = FALSE)

# sales — the CLAUDE.md flagship example: firm-year panel of revenue and
# price with regional effects. 200 firms × 5 years.
set.seed(2026)
firms <- 200
years <- 2019:2023
n <- firms * length(years)
firm_id <- rep(seq_len(firms), each = length(years))
year <- rep(years, times = firms)
region <- rep(sample(1:4, firms, replace = TRUE), each = length(years))
firm_effect <- rep(round(rnorm(firms, 0, 0.3), 6), each = length(years))
price <- round(runif(n, 5, 25) + 0.4 * (year - 2019), 4)
log_revenue <- 4.2 - 0.07 * price + 0.12 * (region - 1) +
  0.05 * (year - 2019) + firm_effect + round(rnorm(n, 0, 0.2), 6)
sales <- data.frame(
  firm_id = firm_id,
  year = year,
  region = region,
  price = price,
  revenue = round(exp(log_revenue), 4),
  employees = pmax(1, round(exp(2 + 0.5 * firm_effect + rnorm(n, 0, 0.4))))
)
write.csv(sales, file.path(out, "sales.csv"), row.names = FALSE, quote = FALSE)

# mtcars — the R classic (public dataset), with the car name as a string
# column so string handling has a sample too.
data(mtcars)
cars <- cbind(model = rownames(mtcars), mtcars)
write.csv(cars, file.path(out, "mtcars.csv"), row.names = FALSE)

cat("wrote", file.path(out, "sales.csv"), "and mtcars.csv\n")