alpenwerk-hr/scripts/check-schema-types.mjs

#!/usr/bin/env node
// Guards lib/types.ts against drifting away from the migrations.
//
// That file is maintained by hand (its own header explains why: no DB
// Werkzeug erzeugt sie), so a new
// column reaches the database without the TypeScript side noticing — and
// `tsc` stays perfectly happy while the app reads a field that is typed but
// absent, or writes one that exists but is not typed.
//
// Reads the migrations rather than a live database on purpose: CI then needs
// no Postgres, and the migrations are the actual source of truth for what
// gets deployed.

import { readFileSync, readdirSync } from "node:fs";
import { dirname, join } from "node:path";
import { fileURLToPath } from "node:url";

const root = join(dirname(fileURLToPath(import.meta.url)), "..");
const migrationsDir = join(root, "db", "migrations");

// Columns are compared, not their types: the hand-written file deliberately
// uses richer unions than the SQL (`EmploymentStatus` for a text column with
// a check constraint), and flagging those would be noise.
const NON_COLUMN_START =
  /^(constraint|check|primary|unique|foreign|exclude|like|inherits|partition|)$/i;

function stripComments(sql) {
  return sql.replace(/--[^\n]*/g, "");
}

/** Splits on top-level commas only, so `numeric(10,2)` stays in one piece. */
function splitTopLevel(body) {
  const parts = [];
  let depth = 0;
  let current = "";
  for (const ch of body) {
    if (ch === "(") depth++;
    if (ch === ")") depth--;
    if (ch === "," && depth === 0) {
      parts.push(current);
      current = "";
      continue;
    }
    current += ch;
  }
  if (current.trim()) parts.push(current);
  return parts;
}

function parseMigrations() {
  const tables = new Map();
  const files = readdirSync(migrationsDir).filter((f) => f.endsWith(".sql")).sort();

  for (const file of files) {
    const sql = stripComments(readFileSync(join(migrationsDir, file), "utf8"));

    for (const match of sql.matchAll(/create table (?:if not exists )?(\w+)\s*\(/gi)) {
      const name = match[1];
      // Walk from the opening paren to its match so nested parens in column
      // definitions do not end the table early.
      let depth = 0;
      let end = match.index + match[0].length - 1;
      for (let i = end; i < sql.length; i++) {
        if (sql[i] === "(") depth++;
        else if (sql[i] === ")") {
          depth--;
          if (depth === 0) {
            end = i;
            break;
          }
        }
      }
      const body = sql.slice(match.index + match[0].length, end);
      const columns = new Set();
      for (const part of splitTopLevel(body)) {
        const first = part.trim().split(/\s+/)[0] ?? "";
        if (!first || NON_COLUMN_START.test(first)) continue;
        columns.add(first.toLowerCase());
      }
      tables.set(name, columns);
    }

    for (const m of sql.matchAll(/alter table (?:if exists )?(\w+)\s+add column (?:if not exists )?(\w+)/gi)) {
      tables.get(m[1])?.add(m[2].toLowerCase());
    }
    for (const m of sql.matchAll(/alter table (?:if exists )?(\w+)\s+drop column (?:if exists )?(\w+)/gi)) {
      tables.get(m[1])?.delete(m[2].toLowerCase());
    }
    for (const m of sql.matchAll(/alter table (?:if exists )?(\w+)\s+rename column (\w+) to (\w+)/gi)) {
      const t = tables.get(m[1]);
      if (t?.delete(m[2].toLowerCase())) t.add(m[3].toLowerCase());
    }
    for (const m of sql.matchAll(/drop table (?:if exists )?(\w+)/gi)) {
      tables.delete(m[1]);
    }
  }
  return tables;
}

function parseTypes() {
  const source = readFileSync(join(root, "lib", "types.ts"), "utf8");
  const start = source.indexOf("Tables: {");
  if (start === -1) throw new Error("Tables block not found in lib/types.ts");

  const tables = new Map();
  // Each entry looks like `name: NoRelationships & { Row: { … } … }`; the Row
  // block is the one that has to match the database.
  const entry = /^ {6}(\w+): /gm;
  for (const m of source.slice(start).matchAll(entry)) {
    const rest = source.slice(start + m.index);
    const rowAt = rest.indexOf("Row: {");
    if (rowAt === -1) continue;
    let depth = 0;
    let end = rowAt + "Row: ".length;
    for (let i = end; i < rest.length; i++) {
      if (rest[i] === "{") depth++;
      else if (rest[i] === "}") {
        depth--;
        if (depth === 0) {
          end = i;
          break;
        }
      }
    }
    const body = rest.slice(rowAt + "Row: {".length, end);
    const columns = new Set();
    for (const f of body.matchAll(/(\w+)\s*\??\s*:/g)) columns.add(f[1].toLowerCase());
    tables.set(m[1], columns);
  }
  return tables;
}

const sqlTables = parseMigrations();
const tsTables = parseTypes();
const problems = [];

for (const [name, columns] of tsTables) {
  const sqlColumns = sqlTables.get(name);
  if (!sqlColumns) {
    problems.push(`Tabelle "${name}" ist in types.ts typisiert, existiert aber in keiner Migration.`);
    continue;
  }
  for (const c of columns) {
    if (!sqlColumns.has(c)) problems.push(`${name}.${c}: in types.ts typisiert, in den Migrationen nicht vorhanden.`);
  }
  for (const c of sqlColumns) {
    if (!columns.has(c)) problems.push(`${name}.${c}: in den Migrationen vorhanden, in types.ts nicht typisiert.`);
  }
}

if (problems.length > 0) {
  console.error("Schema-Drift zwischen db/migrations und lib/types.ts:\n");
  for (const p of problems.sort()) console.error("  - " + p);
  console.error(`\n${problems.length} Abweichung(en). types.ts entsprechend nachziehen.`);
  process.exit(1);
}

console.log(`Kein Schema-Drift: ${tsTables.size} typisierte Tabellen stimmen mit den Migrationen überein.`);