// Build a JSON array of sources for test/bytediff.mjs or bench/throughput.mjs from the
// pinned OSS corpus (vendor/oss/<project>).
//
// Usage:
//   node scripts/oss-sample.mjs bytediff [project]        -> corpus/oss-bytediff-input.json
//     all parse-clean .ts/.mts/.cts files (whole corpus, or one project).
//   node scripts/oss-sample.mjs throughput [--mb=20] [--seed=N]
//     a seeded ~20MB sample of non-.d.ts (runtime) parse-clean files, drawn evenly
//     across projects so one large repo can't dominate -> corpus/oss-throughput-sample.json
import fs from "node:fs";
import path from "node:path";
import { projects, ossRoot, walkFiles, parseClean } from "./oss-corpus.mjs";

const root = path.resolve(import.meta.dirname, "..");
const mode = process.argv[2];
const arg3 = process.argv[3] && !process.argv[3].startsWith("--") ? process.argv[3] : undefined;
const mb = Number((process.argv.find((a) => a.startsWith("--mb=")) ?? "--mb=20").slice(5));
const seed0 = Number((process.argv.find((a) => a.startsWith("--seed=")) ?? "--seed=1").slice(7));

function mulberry32(a) {
  return () => {
    a |= 0; a = (a + 0x6d2b79f5) | 0;
    let t = Math.imul(a ^ (a >>> 15), 1 | a);
    t = (t + Math.imul(t ^ (t >>> 7), 61 | t)) ^ t;
    return ((t ^ (t >>> 14)) >>> 0) / 4294967296;
  };
}

if (mode === "bytediff") {
  const ps = arg3 ? [arg3] : projects();
  const out = [];
  for (const name of ps) {
    const dir = path.join(ossRoot, name);
    if (!fs.existsSync(dir)) continue;
    for (const f of walkFiles(dir)) {
      const c = parseClean(f);
      if (c) out.push(c.src);
    }
  }
  const dest = path.join(root, "corpus/oss-bytediff-input.json");
  fs.writeFileSync(dest, JSON.stringify(out));
  console.log(`oss-sample: ${out.length} parse-clean files -> ${dest}`);
} else if (mode === "throughput") {
  const rand = mulberry32(seed0);
  const ps = projects();
  const perProject = [];
  for (const name of ps) {
    const dir = path.join(ossRoot, name);
    if (!fs.existsSync(dir)) continue;
    const files = [...walkFiles(dir)].filter((f) => !f.endsWith(".d.ts"));
    // Fisher-Yates shuffle with the seeded RNG so the sample is reproducible.
    for (let i = files.length - 1; i > 0; i--) {
      const j = Math.floor(rand() * (i + 1));
      [files[i], files[j]] = [files[j], files[i]];
    }
    perProject.push({ name, files });
  }
  const budget = mb * 1e6;
  let used = 0;
  const out = [];
  // Round-robin across projects so the sample isn't dominated by whichever repo sorts first.
  let idx = 0, remaining = perProject.map((p) => p.files.length).reduce((a, b) => a + b, 0);
  while (used < budget && remaining > 0) {
    const p = perProject[idx % perProject.length];
    idx++;
    if (p.files.length === 0) { remaining = perProject.reduce((a, b) => a + b.files.length, 0); if (remaining === 0) break; continue; }
    const f = p.files.pop();
    remaining--;
    const c = parseClean(f);
    if (!c) continue;
    out.push(c.src);
    used += Buffer.byteLength(c.src);
  }
  const dest = path.join(root, "corpus/oss-throughput-sample.json");
  fs.writeFileSync(dest, JSON.stringify(out));
  console.log(`oss-sample: ${out.length} files, ${(used / 1e6).toFixed(1)}MB -> ${dest}`);
} else {
  console.error("usage: node scripts/oss-sample.mjs <bytediff|throughput> [project] [--mb=20] [--seed=1]");
  process.exit(1);
}