jsql-neo 6.4.1 → 6.4.3

package/test/filter-compile.test.js

@@ -0,0 +1,199 @@
+/*
+ * 差分模糊测试：_compileMatcher（快路径）必须与 _matchRow（原实现）逐行一致。
+ *
+ * 背景：_matchRow 每行都做 Object.entries(query) / Object.keys(condition) /
+ * 一串 startswith 判断。实测这部分占单次全表扫描耗时的 **98%** ——
+ * 10 万行点查 13.9ms/次，其中 13.6ms 是反射开销；同样语义的手写循环只要 0.28ms。
+ * $like / $regex 更糟：正则每行每字段重编一次。
+ *
+ * _compileMatcher 把这些一次性挪到循环外。语义必须一字不差 ——
+ * 这个测试用随机数据 × 随机查询做交叉验证来保证。
+ *
+ * 它抓到过真 bug：$nin 的 target 非数组时，原实现返回 false（不匹配），
+ * 而编译版一度漏判、当成通过。
+ */
+const path = require('path');
+
+const ROOT = path.join(__dirname, '..');
+const Database = require(path.join(ROOT, 'lib/database'));
+
+let pass = 0, fail = 0;
+const failures = [];
+function ok(name, cond, extra) {
+  if (cond) { pass++; } else { fail++; failures.push(name + (extra !== undefined ? '  -> ' + JSON.stringify(extra).slice(0, 170) : '')); }
+}
+
+(async () => {
+  const db = new Database(':memory:', { autoSave: false });
+  await db.createTable('f', {
+    id: { type: 'integer', primaryKey: true },
+    num: { type: 'integer' },
+    str: { type: 'string' },
+    flt: { type: 'float' },
+    arr: { type: 'array' },
+    obj: { type: 'object' },
+  });
+  const STRS = ['a', 'ab', 'Ab', 'abc', 'b', '', 'a%b', 'a_b', 'A', 'zzz', 'aXb'];
+  const NUMS = [0, 1, 2, 3, 5, 10, -1, -5, 100, 25, 50];   // num 是 integer 列
+  const rows = [];
+  for (let i = 1; i <= 400; i++) {
+    rows.push({
+      id: i,
+      num: NUMS[i % NUMS.length],
+      str: STRS[i % STRS.length],
+      flt: (i % 17) / 2,
+      arr: i % 3 === 0 ? [1, 2, 3] : (i % 3 === 1 ? ['x', 'y'] : []),
+      obj: i % 4 === 0 ? { city: 'SH', n: i } : (i % 4 === 1 ? { city: 'BJ' } : { n: 1 }),
+    });
+  }
+  await db.insert('f', rows);
+  const t = db._tables.f;
+  const all = t._rows;
+
+  // ---- 可复现的伪随机（固定种子，失败可重放）----
+  let seed = 12345;
+  const rnd = () => { seed = (seed * 1103515245 + 12345) & 0x7fffffff; return seed / 0x7fffffff; };
+  const pick = (a) => a[Math.floor(rnd() * a.length)];
+
+  const FIELDS = ['num', 'str', 'flt', 'arr', 'obj', 'obj.city', 'id'];
+  const OPS = ['$eq', '$ne', '$gt', '$gte', '$lt', '$lte', '$in', '$nin', '$between',
+    '$exists', '$isNull', '$startsWith', '$endsWith', '$contains', '$type', '$size',
+    '$like', '$regex'];
+
+  const genValue = (op) => {
+    switch (op) {
+      case '$in': case '$nin':
+        return [pick(NUMS), pick(STRS), pick(NUMS)].slice(0, 1 + Math.floor(rnd() * 3));
+      case '$between': return [pick(NUMS), pick(NUMS)].sort((a, b) => a - b);
+      case '$exists': case '$isNull': return rnd() < 0.5;
+      case '$type': return pick(['string', 'number', 'object', 'undefined', 'boolean']);
+      case '$size': return pick([0, 1, 2, 3]);
+      case '$startsWith': case '$endsWith': return pick(['a', 'A', 'b', 'ab', '']);
+      case '$like': return pick(['a%', '%b', 'a_b', '%', 'a%b', 'A%', '_', 'abc']);
+      case '$regex': return pick(['^a', 'b$', '[0-9]+', 'a.*b', '.']);
+      case '$gt': case '$gte': case '$lt': case '$lte': case '$eq': case '$ne':
+        return rnd() < 0.25 ? null : pick(rnd() < 0.5 ? NUMS : STRS);
+      default: return pick(NUMS);
+    }
+  };
+  const genFieldCond = () => {
+    const f = pick(FIELDS);
+    if (rnd() < 0.35) return { [f]: genValue(pick(OPS)) };
+    const ops = {};
+    const n = 1 + Math.floor(rnd() * 2);
+    for (let i = 0; i < n; i++) ops[pick(OPS)] = genValue(pick(OPS));
+    return { [f]: ops };
+  };
+  const genQuery = () => {
+    const r = rnd();
+    if (r < 0.12) return {};                                     // 空查询
+    if (r < 0.30) {                                              // $or
+      const n = 1 + Math.floor(rnd() * 3);
+      return { $or: Array.from({ length: n }, () => genFieldCond()) };
+    }
+    if (r < 0.42) {                                              // $and
+      const n = 1 + Math.floor(rnd() * 3);
+      return { $and: Array.from({ length: n }, () => genFieldCond()) };
+    }
+    if (r < 0.50) return { $not: genFieldCond() };               // $not
+    const q = {};                                                // 普通多字段
+    const n = 1 + Math.floor(rnd() * 3);
+    for (let i = 0; i < n; i++) Object.assign(q, genFieldCond());
+    return q;
+  };
+
+  // ---- 差分比对 ----
+  const N = 4000;
+  let mismatches = 0, empty = 0;
+  for (let i = 0; i < N; i++) {
+    const q = genQuery();
+    if (Object.keys(q).length === 0) empty++;
+    let fast, slow, e1 = null, e2 = null;
+    try { fast = all.filter(r => t._compileMatcher(q)(r)); } catch (e) { e1 = e.message; }
+    try { slow = all.filter(r => t._matchRow(r, q)); } catch (e) { e2 = e.message; }
+    if (e1 || e2) {
+      // 抛错行为也必须一致
+      if (!!e1 !== !!e2) { mismatches++; if (failures.length < 10) failures.push('抛错不一致 ' + JSON.stringify(q)); }
+      continue;
+    }
+    if (fast.length !== slow.length) {
+      mismatches++;
+      if (failures.length < 10) failures.push(`行数不一致 ${fast.length} vs ${slow.length} : ${JSON.stringify(q)}`);
+      continue;
+    }
+    for (let j = 0; j < fast.length; j++) {
+      if (fast[j] !== slow[j]) { mismatches++; if (failures.length < 10) failures.push('行内容不一致 ' + JSON.stringify(q)); break; }
+    }
+  }
+  ok(`差分模糊测试 ${N} 个随机查询零偏差（含 ${empty} 个空查询）`, mismatches === 0, failures.slice(0, 4));
+  console.log(`  随机查询 ${N} 个，偏差 ${mismatches}`);
+
+  // ---- 定向验证：逐个操作符 + 边界 ----
+  const dir = [
+    [{ num: 25 }, '等值'],
+    [{ num: { $eq: 25 } }, '$eq'],
+    [{ num: { $ne: 25 } }, '$ne'],
+    [{ num: { $gt: 50 } }, '$gt'],
+    [{ num: { $gte: 100 } }, '$gte'],
+    [{ num: { $lt: 0 } }, '$lt'],
+    [{ num: { $lte: 0 } }, '$lte'],
+    [{ num: { $in: [1, 25] } }, '$in'],
+    [{ num: { $nin: [1, 25] } }, '$nin'],
+    [{ num: { $nin: true } }, '$nin 非数组 target（曾在此分歧）'],
+    [{ num: { $nin: 'Ab' } }, '$nin 字符串 target'],
+    [{ num: { $in: true } }, '$in 非数组 target'],
+    [{ num: { $between: [1, 50] } }, '$between'],
+    [{ num: { $between: true } }, '$between 非数组'],
+    [{ num: { $exists: true } }, '$exists'],
+    [{ num: { $isNull: true } }, '$isNull'],
+    [{ str: { $startsWith: 'a' } }, '$startsWith'],
+    [{ str: { $endsWith: 'b' } }, '$endsWith'],
+    [{ arr: { $contains: 1 } }, '$contains'],
+    [{ arr: { $size: 3 } }, '$size'],
+    [{ obj: { $type: 'object' } }, '$type'],
+    [{ str: { $like: 'a%' } }, '$like'],
+    [{ str: { $like: 'a_b' } }, '$like _'],
+    [{ str: { $like: 'a%b' } }, '$like 转义 %'],
+    [{ str: { $regex: '^a' } }, '$regex'],
+    [{ 'obj.city': 'SH' }, '点号路径'],
+    [{ obj: { city: 'SH' } }, '嵌套对象'],
+    [{ obj: { city: 'SH', n: 1 } }, '嵌套对象两字段'],
+    [{ $or: [{ num: 1 }, { str: 'b' }] }, '$or'],
+    [{ $and: [{ num: 1 }, { str: 'a' }] }, '$and'],
+    [{ $not: { num: 1 } }, '$not'],
+    [{ num: { $gt: 0 }, str: { $startsWith: 'a' } }, '多字段'],
+    [{}, '空查询'],
+    [{ $or: [{ num: { $nin: true } }] }, '$or 内 $nin 非数组'],
+  ];
+  for (const [q, label] of dir) {
+    const fast = all.filter(r => t._compileMatcher(q)(r));
+    const slow = all.filter(r => t._matchRow(r, q));
+    ok('定向: ' + label, fast.length === slow.length && fast.every((r, i) => r === slow[i]),
+      { fast: fast.length, slow: slow.length });
+  }
+
+  // ---- 软删除模式下也必须一致 ----
+  {
+    await db.query('DELETE FROM f WHERE id <= 40', { safety: false });
+    const q = { num: 25 };
+    const fast = t._applyFilterOptimized(t._getVisibleRows(), q);
+    const slow = t._rows.filter(r => !r._deleted).filter(r => t._matchRow(r, q));
+    ok('软删除模式下一致', fast.length === slow.length && fast.every((r, i) => r === slow[i]),
+      { fast: fast.length, slow: slow.length });
+  }
+
+  // ---- LIKE / REGEX 的正则只编译一次（性能回归护栏）----
+  {
+    const q = { str: { $like: 'a%' } };
+    const m = t._compileMatcher(q);
+    const t0 = Date.now();
+    for (let i = 0; i < 300; i++) all.filter(r => m(r));
+    const ms = Date.now() - t0;
+    ok(`编译版 LIKE 300 次全表扫描在 ${ms}ms 内（原实现每行重编正则）`, ms < 1500, ms);
+    console.log(`  编译版 LIKE ×300 全表扫描: ${ms}ms`);
+  }
+
+  console.log(`\n${pass} passed, ${fail} failed`);
+  if (failures.length) { console.log('\n失败项:'); failures.slice(0, 15).forEach(f => console.log('  - ' + f)); }
+  process.exit(fail === 0 ? 0 : 1);
+})().catch(e => { console.error('💥', e); process.exit(1); });

package/test/sql-semantics.test.js

@@ -0,0 +1,131 @@
+/*
+ * SQL 语义修正测试（IN 的三值逻辑 / 窗口默认帧 / JSON 路径操作符 / 补齐的标量函数）。
+ *
+ * 这一批全是「不报错、结果错」——静默失败最难发现，靠拿标准答案逐条对照才挖出来。
+ *
+ *   1. IN 列表里含 NULL 时整体失效：`v IN (10, NULL)` 连 v=10 的行都匹配不上。
+ *      正确语义是 `v=10 OR v=NULL` —— 有一个元素匹配就是 TRUE，
+ *      列表里别的位置有没有 NULL 都不影响；只有**没匹配上**时才变 UNKNOWN。
+ *   2. 窗口函数默认帧算错：`SUM(x) OVER (ORDER BY id)` 返回分区总和而不是累计值。
+ *      SQL 标准：有 ORDER BY 且未指定帧 → RANGE UNBOUNDED PRECEDING TO CURRENT ROW。
+ *   3. `d->>'$.name'`（文档里给的示例）永远返回 null —— 求值层把 '$' 开头的
+ *      JSONPath 当成字面键名去取了。纯键名写法 `d->>'name'` 一直是对的。
+ *   4. ASCII / CHAR / SPACE / FIELD / ELT / JSON_ARRAY_LENGTH 根本没实现，
+ *      调用时静默返回 null。
+ */
+const path = require('path');
+
+const ROOT = path.join(__dirname, '..');
+const { Database } = require(path.join(ROOT, 'index.js'));
+const { executeSQL } = require(path.join(ROOT, 'lib/sql.js'));
+
+const OPTS = { safety: false };
+let pass = 0, fail = 0;
+
+function ok(name, cond, extra) {
+  if (cond) { pass++; console.log('[OK]  ', name); }
+  else { fail++; console.log('[FAIL]', name, extra !== undefined ? '-> ' + JSON.stringify(extra).slice(0, 150) : ''); }
+}
+const eq = (a, b) => JSON.stringify(a) === JSON.stringify(b);
+
+(async () => {
+  const db = new Database(':memory:');
+  const q = (sql) => executeSQL(db, sql, OPTS);
+  const rows = async (sql) => (await q(sql)).rows;
+  const one = async (sql) => (await q(sql)).rows[0][0];
+
+  await q('CREATE TABLE t (id INT PRIMARY KEY, g TEXT, v INT, s TEXT)');
+  await q(`INSERT INTO t VALUES
+    (1,'a',10,'x'),(2,'a',20,'y'),(3,'b',10,NULL),(4,'b',30,'x'),(5,NULL,20,'y'),(6,NULL,NULL,'z')`);
+
+  console.log('\n--- IN 的三值逻辑 ---');
+  ok('v IN (10) 正常匹配', eq(await rows('SELECT COUNT(*) FROM t WHERE v IN (10)'), [[2]]));
+  ok('v IN (10, NULL) 仍能匹配 v=10（关键用例）',
+    eq(await rows('SELECT COUNT(*) FROM t WHERE v IN (10, NULL)'), [[2]]),
+    await rows('SELECT COUNT(*) FROM t WHERE v IN (10, NULL)'));
+  ok('NULL 在列表开头也一样', eq(await rows('SELECT COUNT(*) FROM t WHERE v IN (NULL, 10)'), [[2]]));
+  ok('多个非 NULL 元素', eq(await rows('SELECT COUNT(*) FROM t WHERE v IN (10,20,30)'), [[5]]));
+  ok('v IN (NULL) 恒为空集', eq(await rows('SELECT COUNT(*) FROM t WHERE v IN (NULL)'), [[0]]));
+  ok('v NOT IN (10)', eq(await rows('SELECT COUNT(*) FROM t WHERE v NOT IN (10)'), [[3]]));
+  ok('v NOT IN (10, NULL) 恒为空集（UNKNOWN 取反仍是 UNKNOWN）',
+    eq(await rows('SELECT COUNT(*) FROM t WHERE v NOT IN (10, NULL)'), [[0]]));
+  ok('与等价的 OR 写法结果一致',
+    eq(await rows('SELECT COUNT(*) FROM t WHERE v = 10 OR v = NULL'),
+      await rows('SELECT COUNT(*) FROM t WHERE v IN (10, NULL)')));
+  ok('字符串 IN 含 NULL', eq(await rows("SELECT COUNT(*) FROM t WHERE s IN ('x', NULL)"), [[2]]));
+  ok('字面量 "null" 不会被 NULL 误匹配',
+    eq(await rows("SELECT COUNT(*) FROM t WHERE s IN ('null')"), [[0]]));
+  ok('子查询 IN（结果含 NULL）',
+    eq(await rows('SELECT COUNT(*) FROM t WHERE v IN (SELECT v FROM t WHERE v = 10 OR v IS NULL)'), [[2]]));
+  ok('NOT IN 子查询含 NULL → 空集',
+    eq(await rows('SELECT COUNT(*) FROM t WHERE v NOT IN (SELECT v FROM t WHERE v = 10 OR v IS NULL)'), [[0]]));
+
+  console.log('\n--- 窗口函数默认帧 ---');
+  await q('CREATE TABLE e (id INT PRIMARY KEY, dept TEXT, sal INT)');
+  await q(`INSERT INTO e VALUES (1,'a',100),(2,'a',200),(3,'a',200),(4,'b',50),(5,'b',300)`);
+  ok('SUM OVER (ORDER BY) 是累计值（关键用例）',
+    eq(await rows('SELECT id, SUM(sal) OVER (PARTITION BY dept ORDER BY id) s FROM e ORDER BY id'),
+      [[1,100],[2,300],[3,500],[4,50],[5,350]]),
+    await rows('SELECT id, SUM(sal) OVER (PARTITION BY dept ORDER BY id) s FROM e ORDER BY id'));
+  ok('无 ORDER BY 时是整分区总和',
+    eq(await rows('SELECT id, SUM(sal) OVER (PARTITION BY dept) s FROM e ORDER BY id'),
+      [[1,500],[2,500],[3,500],[4,350],[5,350]]));
+  ok('无分区无排序 = 全表总和',
+    eq(await rows('SELECT id, SUM(sal) OVER () s FROM e ORDER BY id'),
+      [[1,850],[2,850],[3,850],[4,850],[5,850]]));
+  ok('COUNT OVER (ORDER BY) 累计',
+    eq(await rows('SELECT id, COUNT(*) OVER (ORDER BY id) c FROM e ORDER BY id'),
+      [[1,1],[2,2],[3,3],[4,4],[5,5]]));
+  ok('AVG OVER (ORDER BY) 累计',
+    eq(await rows("SELECT id, AVG(sal) OVER (PARTITION BY dept ORDER BY id) a FROM e WHERE dept='a' ORDER BY id"),
+      [[1,100],[2,150],[3,166.66666666666666]]));
+  ok('默认帧按 RANGE 语义：同值 peer 一并纳入',
+    eq(await rows("SELECT id, SUM(sal) OVER (PARTITION BY dept ORDER BY sal) s FROM e WHERE dept='a' ORDER BY id"),
+      [[1,100],[2,500],[3,500]]));
+  ok('显式 ROWS 帧不受 peer 影响',
+    eq(await rows("SELECT id, SUM(sal) OVER (PARTITION BY dept ORDER BY sal ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) s FROM e WHERE dept='a' ORDER BY id"),
+      [[1,100],[2,300],[3,500]]));
+  ok('ROW_NUMBER 不受默认帧影响',
+    eq(await rows('SELECT id, ROW_NUMBER() OVER (PARTITION BY dept ORDER BY sal DESC, id) r FROM e ORDER BY id'),
+      [[1,3],[2,1],[3,2],[4,2],[5,1]]));
+
+  console.log('\n--- JSON 路径操作符 ---');
+  await q('CREATE TABLE j (id INT PRIMARY KEY, d TEXT)');
+  await q(`INSERT INTO j VALUES (1,'{"name":"ann","n":5,"addr":{"city":"SH"},"arr":[10,20]}')`);
+  ok("d->>'$.name'（文档里的写法）", await one("SELECT d->>'$.name' FROM j") === 'ann');
+  ok("d->'$.name'", await one("SELECT d->'$.name' FROM j") === 'ann');
+  ok("d->>'$.n' 数字转文本", await one("SELECT d->>'$.n' FROM j") === '5');
+  ok("d->'$.n' 保留数字类型", await one("SELECT d->'$.n' FROM j") === 5);
+  ok("d->>'$.addr.city' 嵌套", await one("SELECT d->>'$.addr.city' FROM j") === 'SH');
+  ok("d->'$.addr' 返回对象", eq(await one("SELECT d->'$.addr' FROM j"), { city: 'SH' }));
+  ok("d->>'$.arr[1]' 数组下标", await one("SELECT d->>'$.arr[1]' FROM j") === '20');
+  ok("d->>'$.nope' 不存在 → null", await one("SELECT d->>'$.nope' FROM j") === null);
+  ok("纯键名写法不受影响 d->>'name'", await one("SELECT d->>'name' FROM j") === 'ann');
+  ok("链式 d->'addr'->>'city'", await one("SELECT d->'addr'->>'city' FROM j") === 'SH');
+  ok('JSON_EXTRACT 未受影响',
+    await one("SELECT JSON_EXTRACT(d,'$.addr.city') FROM j") === 'SH');
+  ok("JSON_EXTRACT '$' 取根", eq(await one("SELECT JSON_EXTRACT('{\"a\":1}','$')"), { a: 1 }));
+
+  console.log('\n--- 补齐的标量函数 ---');
+  ok("ASCII('A')", await one("SELECT ASCII('A')") === 65);
+  ok("ASCII('abc')", await one("SELECT ASCII('abc')") === 97);
+  ok("ASCII('') 为 0", await one("SELECT ASCII('')") === 0);
+  ok('ASCII(NULL) 为 NULL', await one('SELECT ASCII(NULL)') === null);
+  ok('CHAR(65)', await one('SELECT CHAR(65)') === 'A');
+  ok('CHAR 多参数', await one('SELECT CHAR(72,105)') === 'Hi');
+  ok('SPACE(3)', await one('SELECT LENGTH(SPACE(3))') === 3);
+  ok('SPACE(0)', await one('SELECT LENGTH(SPACE(0))') === 0);
+  ok("FIELD 命中", await one("SELECT FIELD('b','a','b','c')") === 2);
+  ok('FIELD 未命中为 0', await one("SELECT FIELD('z','a','b')") === 0);
+  ok('FIELD(NULL) 为 NULL', await one("SELECT FIELD(NULL,'a')") === null);
+  ok('ELT 命中', await one("SELECT ELT(2,'a','b','c')") === 'b');
+  ok('ELT 越界为 NULL', await one("SELECT ELT(9,'a','b')") === null);
+  ok('ELT(0) 为 NULL', await one("SELECT ELT(0,'a','b')") === null);
+  ok('JSON_ARRAY_LENGTH', await one("SELECT JSON_ARRAY_LENGTH('[1,2,3]')") === 3);
+  ok('JSON_ARRAY_LENGTH 非数组 → NULL',
+    await one("SELECT JSON_ARRAY_LENGTH('{\"a\":1}')") === null);
+  ok('JSON_ARRAY_LENGTH 空数组', await one("SELECT JSON_ARRAY_LENGTH('[]')") === 0);
+
+  console.log(`\n${pass} passed, ${fail} failed`);
+  process.exit(fail === 0 ? 0 : 1);
+})().catch(e => { console.error('💥', e); process.exit(1); });