MetonaSqlark/tests/v080-query-layer.test.ts

/**
 * v0.8.0 回归套件 —— 查询层缺陷根治（A22/A23/A25/A26/A27/A29/A30/A36）
 * ============================================================================
 * 本套件锁定 PLAN-v0.7.5.md §5 缺陷总账中查询层的 8 项修复。每一项都先给出
 * **修复前的实测错误输出**，再断言正确结果 —— 这样即使将来重构执行器，
 * 失败的断言能直接告诉后来者"当初错在哪里"。
 *
 * 参考数据（表 t）：
 *   id=1 g='a' n=10      id=3 g='b' n=30
 *   id=2 g='a' n=20      id=4 g='b' n=40
 */
import { MetonaSqlark } from '../src/core';
import { rows as rowsOf } from './helpers/assertions';
import type { DatabaseConfig } from '../src/constants';

const ENGINES: Array<[string, DatabaseConfig['mode'], Partial<DatabaseConfig>]> = [
  ['memory', 'memory', {}],
  ['disk', 'disk', {}],
  ['hybrid', 'hybrid', {}],
  ['aria', 'aria', { diskEngine: 'memory' }],
];

describe('[v0.8.0] 查询层缺陷根治', () => {
  describe.each(ENGINES)('%s 引擎', (label, mode, extra) => {
    let db: MetonaSqlark;

    beforeEach(async () => {
      db = await MetonaSqlark.create({
        name: `ql-${label}-${Math.random().toString(36).slice(2)}`,
        mode,
        ...extra,
      });
      await db.defineTable('t', {
        id: { type: 'string', primaryKey: true },
        g: { type: 'string' },
        n: { type: 'number' },
      });
      await db.query("INSERT INTO t VALUES ('1','a',10),('2','a',20),('3','b',30),('4','b',40)");
    });

    afterEach(async () => {
      await db.close();
    });

    // -------------------------------------------------------------------
    //  A22: GROUP BY 可引用 SELECT 别名
    // -------------------------------------------------------------------
    it('A22 GROUP BY 引用 SELECT 别名', async () => {
      // 修复前：抛 `COLUMN_NOT_FOUND Unknown column "g" in SELECT list`
      // （GROUP BY grp 取 row['grp'] 得 undefined → 全部行并成一组 →
      //  投影阶段又发现 g 不在输出行里，报出一个和真正原因无关的错误）
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g AS grp, COUNT(*) AS c FROM t GROUP BY grp ORDER BY grp'),
      );
      expect(rows).toEqual([{ g: 'a', c: 2 }, { g: 'b', c: 2 }]);
      // 输出键与"按基列分组"完全一致（同一查询的两种写法必须等价）
      const byBase = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g, COUNT(*) AS c FROM t GROUP BY g ORDER BY g'),
      );
      expect(rows).toEqual(byBase);
    });

    it('A22 GROUP BY 引用别名时不泄漏别名键', async () => {
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g AS grp FROM t GROUP BY grp ORDER BY grp'),
      );
      // 输出的键必须是基列名 g（SELECT 列表里 g AS grp 的投影结果），
      // 而不是内部的分组键 grp —— 否则同一查询在"有别名/无别名"两种写法下
      // 行形状不同。
      expect(rows).toEqual([{ g: 'a' }, { g: 'b' }]);
    });

    // -------------------------------------------------------------------
    //  A23: HAVING 可引用未出现在 SELECT 里的聚合
    // -------------------------------------------------------------------
    it('A23 HAVING 引用未选中的聚合（SUM）', async () => {
      // 修复前：返回 []（SUM(n) 从未被求值 → HAVING 的键在行里不存在 → UNKNOWN）
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g FROM t GROUP BY g HAVING SUM(n) > 25 ORDER BY g'),
      );
      expect(rows).toEqual([{ g: 'a' }, { g: 'b' }]);
    });

    it('A23 HAVING 引用未选中的聚合（MAX）并正确过滤', async () => {
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g FROM t GROUP BY g HAVING MAX(n) > 25 ORDER BY g'),
      );
      expect(rows).toEqual([{ g: 'b' }]);
    });

    it('A23 HAVING 的辅助聚合不得出现在输出行里', async () => {
      // 为 HAVING 计算的 SUM(n) 是内部键：输出必须只有 g 一列。
      // （修复过程中曾泄漏成 {g, 'SUM(n)'} —— 多出用户没要求的输出列。）
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g FROM t GROUP BY g HAVING SUM(n) > 0 ORDER BY g'),
      );
      expect(rows).toEqual([{ g: 'a' }, { g: 'b' }]);
      expect(Object.keys(rows[0])).toEqual(['g']);
    });

    it('A23 HAVING 同时引用别名与未选中聚合', async () => {
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g AS grp, COUNT(*) AS c FROM t GROUP BY grp HAVING SUM(n) > 25 ORDER BY grp'),
      );
      expect(rows).toEqual([{ g: 'a', c: 2 }, { g: 'b', c: 2 }]);
    });

    // -------------------------------------------------------------------
    //  A25: 带表前缀的聚合参数
    // -------------------------------------------------------------------
    it('A25 COUNT(t.n) / SUM(t.n) 按列取值而非恒 0', async () => {
      // 修复前：COUNT(t.n) → 0，SUM(t.n) → null（行键是 n，直接取 row['t.n'] 得
      // undefined，再被"过滤 NULL"剔除，且**不报错**）
      expect(rowsOf(await db.query('SELECT COUNT(t.n) AS c FROM t'))).toEqual([{ c: 4 }]);
      expect(rowsOf(await db.query('SELECT SUM(t.n) AS s FROM t'))).toEqual([{ s: 100 }]);
      expect(rowsOf(await db.query('SELECT AVG(t.n) AS a FROM t'))).toEqual([{ a: 25 }]);
      expect(rowsOf(await db.query('SELECT MIN(t.n) AS lo FROM t'))).toEqual([{ lo: 10 }]);
      expect(rowsOf(await db.query('SELECT MAX(t.n) AS hi FROM t'))).toEqual([{ hi: 40 }]);
    });

    it('A25 分组聚合的带前缀参数', async () => {
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT g, SUM(t.n) AS s FROM t GROUP BY g ORDER BY g'),
      );
      expect(rows).toEqual([{ g: 'a', s: 30 }, { g: 'b', s: 70 }]);
    });

    it('A25 COUNT(DISTINCT t.n) 带前缀且类型安全去重', async () => {
      expect(rowsOf(await db.query('SELECT COUNT(DISTINCT t.n) AS c FROM t'))).toEqual([{ c: 4 }]);
      // 数值去重不得被编码串扰（曾把 encodeValueKey 的结果 Number() 回读成 NaN）
      expect(rowsOf(await db.query('SELECT SUM(DISTINCT t.n) AS s FROM t'))).toEqual([{ s: 100 }]);
    });

    it('A25 聚合参数引用不存在的列 → COLUMN_NOT_FOUND（不静默计 0）', async () => {
      await expect(db.query('SELECT COUNT(t.nope) AS c FROM t')).rejects.toMatchObject({
        code: 'COLUMN_NOT_FOUND',
      });
    });

    // -------------------------------------------------------------------
    //  A26: UNION 尾部 ORDER BY / LIMIT 作用于整个复合结果
    // -------------------------------------------------------------------
    it('A26 UNION 尾部 ORDER BY 作用于复合结果', async () => {
      // 修复前：只对右侧 SELECT 排序 → [{1},{2},{4},{3}]
      const rows = rowsOf<Record<string, unknown>>(
        await db.query("SELECT id FROM t WHERE g = 'a' UNION SELECT id FROM t WHERE g = 'b' ORDER BY id DESC"),
      );
      expect(rows).toEqual([{ id: '4' }, { id: '3' }, { id: '2' }, { id: '1' }]);
    });

    it('A26 UNION 尾部 LIMIT 作用于复合结果', async () => {
      // 修复前：只截断右侧 → 返回 4 行
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT id FROM t UNION SELECT id FROM t LIMIT 3'),
      );
      expect(rows).toHaveLength(3);
    });

    it('A26 UNION LIMIT 只应用一次（不得双重截断）', async () => {
      const limited = rowsOf<Record<string, unknown>>(
        await db.query('SELECT id FROM t UNION ALL SELECT id FROM t LIMIT 5'),
      );
      expect(limited).toHaveLength(5);
    });

    it('A26 UNION 去重 + OFFSET', async () => {
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT id FROM t UNION SELECT id FROM t ORDER BY id LIMIT 2 OFFSET 1'),
      );
      expect(rows).toEqual([{ id: '2' }, { id: '3' }]);
    });

    it('A26 UNION ORDER BY 引用不存在的列 → COLUMN_NOT_FOUND', async () => {
      await expect(db.query('SELECT id FROM t UNION SELECT id FROM t ORDER BY nope')).rejects.toMatchObject({
        code: 'COLUMN_NOT_FOUND',
      });
    });

    // -------------------------------------------------------------------
    //  A27: DISTINCT 作用于输出列（投影之后）
    // -------------------------------------------------------------------
    it('A27 DISTINCT 列别名与裸列结果一致', async () => {
      // 修复前：`SELECT DISTINCT g AS d` 返回 4 行 a,a,b,b
      //（DISTINCT 作用在投影前的 {id,g,n} 原始行上，四行互不相同）
      const aliased = rowsOf<Record<string, unknown>>(await db.query('SELECT DISTINCT g AS d FROM t ORDER BY d'));
      expect(aliased).toEqual([{ d: 'a' }, { d: 'b' }]);
      const bare = rowsOf<Record<string, unknown>>(await db.query('SELECT DISTINCT g FROM t ORDER BY g'));
      expect(bare).toEqual([{ g: 'a' }, { g: 'b' }]);
    });

    it('A27 DISTINCT 多列仍按输出列去重', async () => {
      const rows = rowsOf<Record<string, unknown>>(await db.query('SELECT DISTINCT g, n FROM t ORDER BY g, n'));
      expect(rows).toHaveLength(4);
    });

    it('A27 DISTINCT + ORDER BY 输出列（先去重再排序）', async () => {
      const rows = rowsOf<Record<string, unknown>>(await db.query('SELECT DISTINCT g FROM t ORDER BY g DESC'));
      expect(rows).toEqual([{ g: 'b' }, { g: 'a' }]);
    });

    // -------------------------------------------------------------------
    //  A30: INSERT 值个数与列个数不匹配 → 报错（此前静默丢弃多余值）
    // -------------------------------------------------------------------
    it('A30 显式列名时值多于列 → PARSE_ERROR（解析期即可判定）', async () => {
      // 修复前：`INSERT INTO t (id, g) VALUES ('9','z','LOST')` 报成功、'LOST' 消失。
      // 有了显式列名后 arity 无需 schema 即可判定，因此解析期就拦下
      //（v0.8.0 A16 的解析期校验；executor 侧对"未显式给列名"的情形兜底）。
      await expect(
        db.query("INSERT INTO t (id, g) VALUES ('9', 'z', 'LOST')"),
      ).rejects.toMatchObject({ code: 'PARSE_ERROR' });
      expect(rowsOf(await db.query("SELECT id FROM t WHERE id = '9'"))).toHaveLength(0);
    });

    it('A30 显式列名时值少于列 → PARSE_ERROR（缺列必须显式写出）', async () => {
      // `INSERT INTO t (id, g) VALUES ('9')` 是列/值个数不匹配的写法：
      // 用户想写的是 `INSERT INTO t (id) VALUES ('9')`。静默补 NULL 会让
      // 拼错列清单的语句"看起来成功"，因此同样报错。
      await expect(db.query("INSERT INTO t (id, g) VALUES ('9')")).rejects.toMatchObject({
        code: 'PARSE_ERROR',
      });
    });

    it('A30 未显式列名时值多于列 → VALIDATION_ERROR（需要 schema 才能判定）', async () => {
      // 不带列清单时个数要对照 schema 才能判断，由 executor 在拿到 schema 后校验
      await expect(db.query("INSERT INTO t VALUES ('9', 'z', 1, 'LOST')")).rejects.toMatchObject({
        code: 'VALIDATION_ERROR',
      });
      expect(rowsOf(await db.query("SELECT id FROM t WHERE id = '9'"))).toHaveLength(0);
    });

    it('A30 值少于列仍合法（缺列走 default / NULL）', async () => {
      await db.query("INSERT INTO t (id, g) VALUES ('9', 'z')");
      const rows = rowsOf<Record<string, unknown>>(await db.query("SELECT id, g, n FROM t WHERE id = '9'"));
      expect(rows).toEqual([{ id: '9', g: 'z', n: null }]);
    });

    // -------------------------------------------------------------------
    //  A36: 派生表别名引用
    // -------------------------------------------------------------------
    it('A36 派生表别名引用与裸列引用结果一致', async () => {
      // 修复前：`SELECT d.id FROM (SELECT ...) AS d` 返回 []（d.id 未剥离前缀），
      // 而同义的 `SELECT id FROM (...) AS d` 正确
      const aliased = rowsOf<Record<string, unknown>>(
        await db.query("SELECT d.id FROM (SELECT id, g FROM t) AS d WHERE d.g = 'a' ORDER BY d.id"),
      );
      expect(aliased).toEqual([{ id: '1' }, { id: '2' }]);
      const bare = rowsOf<Record<string, unknown>>(
        await db.query("SELECT id FROM (SELECT id, g FROM t) AS d WHERE g = 'a' ORDER BY id"),
      );
      expect(aliased).toEqual(bare);
    });

    it('A36 派生表别名用于聚合与排序', async () => {
      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT COUNT(d.id) AS c FROM (SELECT id FROM t) AS d'),
      );
      expect(rows).toEqual([{ c: 4 }]);
    });

    // -------------------------------------------------------------------
    //  A35: JOIN 的 NULL 键
    // -------------------------------------------------------------------
    it('A35 JOIN 的 NULL 键不成立，且与右表有无索引无关', async () => {
      await db.defineTable('l', { id: { type: 'string', primaryKey: true }, k: { type: 'string' } });
      await db.defineTable('ri', { id: { type: 'string', primaryKey: true }, k: { type: 'string', index: true } });
      await db.defineTable('rn', { id: { type: 'string', primaryKey: true }, k: { type: 'string' } });
      await db.query("INSERT INTO l VALUES ('l1','x'),('l2',NULL)");
      await db.query("INSERT INTO ri VALUES ('r1','x'),('r2',NULL)");
      await db.query("INSERT INTO rn VALUES ('n1','x'),('n2',NULL)");

      const indexed = rowsOf<Record<string, unknown>>(
        await db.query('SELECT l.id AS lid, ri.id AS rid FROM l JOIN ri ON l.k = ri.k'),
      );
      const unindexed = rowsOf<Record<string, unknown>>(
        await db.query('SELECT l.id AS lid, rn.id AS rid FROM l JOIN rn ON l.k = rn.k'),
      );
      // NULL = NULL 是 UNKNOWN → 两个 NULL 行都不得匹配
      expect(indexed).toEqual([{ lid: 'l1', rid: 'r1' }]);
      expect(unindexed).toEqual([{ lid: 'l1', rid: 'n1' }]);
      // 关键不变量：结果与"右表该列有没有索引"无关
      expect(indexed.map((r) => r.lid)).toEqual(unindexed.map((r) => r.lid));
    });

    it('A35 LEFT JOIN 保留未匹配行（NULL 键行保留、右表列补 NULL）', async () => {
      await db.defineTable('l', { id: { type: 'string', primaryKey: true }, k: { type: 'string' } });
      await db.defineTable('r', { id: { type: 'string', primaryKey: true }, k: { type: 'string' } });
      await db.query("INSERT INTO l VALUES ('l1','x'),('l2',NULL)");
      await db.query("INSERT INTO r VALUES ('r1','x')");

      const rows = rowsOf<Record<string, unknown>>(
        await db.query('SELECT l.id AS lid, r.id AS rid FROM l LEFT JOIN r ON l.k = r.k ORDER BY lid'),
      );
      expect(rows).toEqual([{ lid: 'l1', rid: 'r1' }, { lid: 'l2', rid: null }]);
    });
  });
});

// ---------------------------------------------------------------------------
//  maxRowsPerQuery 与 NaN 落盘（与引擎无关，用 memory 验证即可）
// ---------------------------------------------------------------------------

describe('[v0.8.0] A29 写路径的行数上限保护', () => {
  it('INSERT ... SELECT 超过 maxRowsPerQuery → 显式报错（不静默截断）', async () => {
    const db = await MetonaSqlark.create({ name: 'a29-limit', mode: 'memory', maxRowsPerQuery: 2 });
    await db.defineTable('src', { id: { type: 'string', primaryKey: true } });
    await db.defineTable('dst', { id: { type: 'string', primaryKey: true } });
    // 注意：每条 INSERT 都受上限约束，因此分 4 条语句各写 1 行来堆积 4 行源数据
    for (const id of ['1', '2', '3', '4']) {
      await db.query(`INSERT INTO src VALUES ('${id}')`);
    }
    expect(rowsOf<{ c: number }>(await db.query('SELECT COUNT(*) AS c FROM src'))).toEqual([{ c: 4 }]);

    // 修复前：行源被 maxRowsPerQuery 静默截断为 2 行 → 只写入 2 行并报成功。
    // 现在：行源不截断，写路径显式报错，dst 保持空。
    await expect(db.query('INSERT INTO dst SELECT id FROM src')).rejects.toMatchObject({
      code: 'QUERY_ERROR',
    });
    expect(rowsOf(await db.query('SELECT id FROM dst'))).toHaveLength(0);
    await db.close();
  });

  it('INSERT ... SELECT 未超上限时正常写入全部行源', async () => {
    const db = await MetonaSqlark.create({ name: 'a29-within', mode: 'memory', maxRowsPerQuery: 2 });
    await db.defineTable('src', { id: { type: 'string', primaryKey: true } });
    await db.defineTable('dst', { id: { type: 'string', primaryKey: true } });
    await db.query("INSERT INTO src VALUES ('1'),('2')");
    await db.query('INSERT INTO dst SELECT id FROM src');
    expect(rowsOf(await db.query('SELECT id FROM dst ORDER BY id'))).toEqual([{ id: '1' }, { id: '2' }]);
    await db.close();
  });

  it('INSERT ... VALUES 同样受上限保护', async () => {
    const db = await MetonaSqlark.create({ name: 'a29-limit-values', mode: 'memory', maxRowsPerQuery: 2 });
    await db.defineTable('t', { id: { type: 'string', primaryKey: true } });
    await expect(db.query("INSERT INTO t VALUES ('1'),('2'),('3')")).rejects.toMatchObject({
      code: 'QUERY_ERROR',
    });
    expect(rowsOf(await db.query('SELECT id FROM t'))).toHaveLength(0);
    await db.close();
  });
});

describe('[v0.8.0] NaN / Infinity 拒绝落盘（B-1 规范化契约）', () => {
  const ENGINES2 = ENGINES;

  it.each(ENGINES2)('%s: NaN 写入被拒绝（否则重启后变 null）', async (label, mode, extra) => {
    const db = await MetonaSqlark.create({ name: `nan-${label}`, mode, ...extra });
    await db.defineTable('t', {
      id: { type: 'string', primaryKey: true },
      v: { type: 'number' },
    });
    await expect(db.table('t').insert({ id: '1', v: Number.NaN } as never)).rejects.toMatchObject({
      code: 'VALIDATION_ERROR',
    });
    await expect(db.table('t').insert({ id: '2', v: Number.POSITIVE_INFINITY } as never)).rejects.toMatchObject({
      code: 'VALIDATION_ERROR',
    });
    // 正常数值不受影响
    await db.table('t').insert({ id: '3', v: 1.5 } as never);
    expect(rowsOf<{ v: number }>(await db.query('SELECT v FROM t'))).toEqual([{ v: 1.5 }]);
    await db.close();
  });
});