@clickzetta/cz-cli-darwin-x64 0.3.1

package/bin/skills/lakehouse-doc/references/groupby.md

@@ -0,0 +1,1260 @@
+# GROUP BY
+
+## 概述
+
+`GROUP BY` 子句是 SQL 查询中用于数据分组和聚合分析的核心组件。通过 `GROUP BY`，可以将数据集按照指定的列或表达式进行分组，并对每个分组应用聚合函数（如 `SUM()`、`COUNT()`、`AVG()`、`MAX()`、`MIN()` 等）计算汇总值。
+
+云器 Lakehouse 完整支持标准 SQL GROUP BY 语法，并提供高级聚合能力：
+
+| 功能                | 说明         | 应用场景           |
+| ----------------- | ---------- | -------------- |
+| **标准 GROUP BY**   | 基础分组聚合     | 日常统计、基础报表      |
+| **GROUPING SETS** | 自定义多个分组集合  | 多维度业务分析        |
+| **ROLLUP**        | 层级汇总（从细到总） | 财务报表、销售层级分析    |
+| **CUBE**          | 全维度交叉分析    | BI 多维数据透视、交叉分析 |
+
+***
+
+## 语法结构
+
+```sql
+SELECT 
+    column1,
+    column2,
+    aggregate_function(column3) AS alias
+FROM table_name
+WHERE condition
+GROUP BY 
+    [ group_expression [, group_expression, ...] |
+      GROUPING SETS (grouping_set [, grouping_set, ...]) |
+      ROLLUP(expression [, expression, ...]) |
+      CUBE(expression [, expression, ...]) ]
+HAVING aggregate_condition
+ORDER BY column
+LIMIT n;
+```
+
+### 关键要点
+
+1. **SELECT 子句约束**：SELECT 列表中的非聚合列必须出现在 GROUP BY 子句中
+2. **执行顺序**：FROM → WHERE → GROUP BY → HAVING → SELECT → ORDER BY → LIMIT
+3. **HAVING 过滤**：用于过滤聚合结果，而 WHERE 过滤原始行
+4. **NULL 值处理**：NULL 被视为一个独立的分组
+
+***
+
+## 准备测试数据
+
+本文档使用汽车经销商销售数据作为示例。创建测试视图：
+
+```sql
+CREATE OR REPLACE VIEW dealer (id, city, car_model, quantity) AS
+VALUES 
+    (100, 'Fremont', 'Honda Civic', 10),
+    (100, 'Fremont', 'Honda Accord', 15),
+    (100, 'Fremont', 'Honda CRV', 7),
+    (200, 'Dublin', 'Honda Civic', 20),
+    (200, 'Dublin', 'Honda Accord', 10),
+    (200, 'Dublin', 'Honda CRV', 3),
+    (300, 'San Jose', 'Honda Civic', 5),
+    (300, 'San Jose', 'Honda Accord', 8);
+```
+
+查看原始数据：
+
+```sql
+SELECT * FROM dealer ORDER BY id, car_model LIMIT 50;
+```
+
+**查询结果**：
+
+| id  | city     | car\_model   | quantity |
+| --- | -------- | ------------ | -------- |
+| 100 | Fremont  | Honda Accord | 15       |
+| 100 | Fremont  | Honda CRV    | 7        |
+| 100 | Fremont  | Honda Civic  | 10       |
+| 200 | Dublin   | Honda Accord | 10       |
+| 200 | Dublin   | Honda CRV    | 3        |
+| 200 | Dublin   | Honda Civic  | 20       |
+| 300 | San Jose | Honda Accord | 8        |
+| 300 | San Jose | Honda Civic  | 5        |
+
+***
+
+## 一、标准 GROUP BY
+
+### 1.1 单列分组
+
+按经销商 ID 分组，统计每个经销商的销量：
+
+```sql
+SELECT 
+    id, 
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY id
+ORDER BY id
+LIMIT 50;
+```
+
+**查询结果**：
+
+| id  | total\_quantity |
+| --- | --------------- |
+| 100 | 32              |
+| 200 | 33              |
+| 300 | 13              |
+
+**业务解读**：
+
+*   经销商 100（Fremont）销售了 32 辆车
+*   经销商 200（Dublin）销售了 33 辆车
+*   经销商 300（San Jose）销售了 13 辆车
+
+***
+
+### 1.2 多列分组
+
+按城市和车型的组合进行分组统计：
+
+```sql
+SELECT 
+    id, 
+    city, 
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY id, city
+ORDER BY total_quantity DESC
+LIMIT 50;
+```
+
+**查询结果**：
+
+| id  | city     | total\_quantity |
+| --- | -------- | --------------- |
+| 200 | Dublin   | 33              |
+| 100 | Fremont  | 32              |
+| 300 | San Jose | 13              |
+
+在本示例中，每个经销商 ID 对应唯一的城市，因此结果与单列分组相同。
+
+***
+
+### 1.3 使用表达式分组
+
+使用 `CASE` 表达式对销量进行分段统计：
+
+```sql
+SELECT 
+    CASE 
+        WHEN quantity < 10 THEN 'Low (< 10)'
+        WHEN quantity < 15 THEN 'Medium (10-14)'
+        ELSE 'High (>= 15)'
+    END AS sales_level,
+    COUNT(*) AS record_count,
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY 
+    CASE 
+        WHEN quantity < 10 THEN 'Low (< 10)'
+        WHEN quantity < 15 THEN 'Medium (10-14)'
+        ELSE 'High (>= 15)'
+    END
+ORDER BY sales_level
+LIMIT 50;
+```
+
+**查询结果**：
+
+| sales\_level   | record\_count | total\_quantity |
+| -------------- | ------------- | --------------- |
+| High (>= 15)   | 2             | 35              |
+| Low (< 10)     | 4             | 23              |
+| Medium (10-14) | 2             | 20              |
+
+**业务解读**：
+
+* 高销量记录（≥15 辆）：2 条记录，共 35 辆
+* 低销量记录（<10 辆）：4 条记录，共 23 辆
+* 中等销量记录（10-14 辆）：2 条记录，共 20 辆
+
+***
+
+### 1.4 全局聚合（无分组）
+
+不使用 GROUP BY 时，对整个表进行全局聚合：
+
+```sql
+SELECT SUM(quantity) AS total_quantity 
+FROM dealer
+LIMIT 50;
+```
+
+**查询结果**：
+
+| total\_quantity |
+| --------------- |
+| 78              |
+
+全部 8 条记录的总销量为 78 辆。
+
+***
+
+## 二、多聚合函数组合
+
+### 2.1 常用聚合函数汇总
+
+在单次查询中使用多个聚合函数：
+
+```sql
+SELECT 
+    city,
+    COUNT(*) AS model_count,
+    SUM(quantity) AS total_quantity,
+    AVG(quantity) AS avg_quantity,
+    MAX(quantity) AS max_quantity,
+    MIN(quantity) AS min_quantity
+FROM dealer
+GROUP BY city
+ORDER BY total_quantity DESC
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | model\_count | total\_quantity | avg\_quantity      | max\_quantity | min\_quantity |
+| -------- | ------------ | --------------- | ------------------ | ------------- | ------------- |
+| Dublin   | 3            | 33              | 11.0               | 20            | 3             |
+| Fremont  | 3            | 32              | 10.666666666666666 | 15            | 7             |
+| San Jose | 2            | 13              | 6.5                | 8             | 5             |
+
+**业务洞察**：
+
+* **Dublin**：3 个车型，平均每个车型销售 11 辆，最高单品 20 辆（Honda Civic）
+* **Fremont**：3 个车型，平均销量 10.67 辆，销量较均衡
+* **San Jose**：仅 2 个车型，平均销量 6.5 辆，市场规模最小
+
+***
+
+### 2.2 按车型统计覆盖城市数
+
+统计每个车型在多少个城市有销售：
+
+```sql
+SELECT 
+    car_model,
+    SUM(quantity) AS total_sales,
+    COUNT(DISTINCT city) AS city_count,
+    ROUND(AVG(quantity), 2) AS avg_per_dealer
+FROM dealer
+GROUP BY car_model
+ORDER BY total_sales DESC
+LIMIT 50;
+```
+
+**查询结果**：
+
+| car\_model   | total\_sales | city\_count | avg\_per\_dealer |
+| ------------ | ------------ | ----------- | ---------------- |
+| Honda Civic  | 35           | 3           | 11.67            |
+| Honda Accord | 33           | 3           | 11.00            |
+| Honda CRV    | 10           | 2           | 5.00             |
+
+**业务洞察**：
+
+* **Honda Civic**：最畅销车型，总销量 35 辆，覆盖 3 个城市，平均每地 11.67 辆
+* **Honda Accord**：总销量 33 辆，覆盖 3 个城市
+* **Honda CRV**：总销量最低，仅在 2 个城市有销售（未进入 San Jose 市场）
+
+***
+
+## 三、GROUPING SETS（自定义分组集合）
+
+### 3.1 基本概念
+
+`GROUPING SETS` 允许在一次查询中同时生成多个不同维度的分组结果。它本质上是多个 `GROUP BY` 查询通过 `UNION ALL` 组合的简写形式。
+
+**语法**：
+
+```sql
+GROUP BY GROUPING SETS (
+    (column1, column2),  -- 分组集1：按 column1 和 column2 分组
+    (column1),           -- 分组集2：仅按 column1 分组
+    ()                   -- 分组集3：全局聚合（无分组）
+)
+```
+
+***
+
+### 3.2 多维度聚合分析
+
+在一次查询中生成四个维度的统计结果：
+
+```sql
+SELECT 
+    city, 
+    car_model, 
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY GROUPING SETS (
+    (city, car_model),  -- 城市+车型明细
+    (city),             -- 按城市汇总
+    (car_model),        -- 按车型汇总
+    ()                  -- 总计
+)
+ORDER BY city, car_model
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | car\_model   | total\_quantity |
+| -------- | ------------ | --------------- |
+| NULL     | NULL         | 78              |
+| NULL     | Honda Accord | 33              |
+| NULL     | Honda CRV    | 10              |
+| NULL     | Honda Civic  | 35              |
+| Dublin   | NULL         | 33              |
+| Dublin   | Honda Accord | 10              |
+| Dublin   | Honda CRV    | 3               |
+| Dublin   | Honda Civic  | 20              |
+| Fremont  | NULL         | 32              |
+| Fremont  | Honda Accord | 15              |
+| Fremont  | Honda CRV    | 7               |
+| Fremont  | Honda Civic  | 10              |
+| San Jose | NULL         | 13              |
+| San Jose | Honda Accord | 8               |
+| San Jose | Honda Civic  | 5               |
+
+**结果解读**：
+
+* **总计行**（city=NULL, car\_model=NULL）：全部销量 78 辆
+* **车型汇总行**（city=NULL, car\_model有值）：各车型跨城市总销量
+* **城市汇总行**（city有值, car\_model=NULL）：各城市跨车型总销量
+* **明细行**（city和car\_model都有值）：城市+车型组合的具体销量
+
+**等价查询**：
+
+```sql
+-- GROUPING SETS 等价于以下 UNION ALL 查询
+SELECT city, car_model, SUM(quantity) FROM dealer GROUP BY city, car_model
+UNION ALL
+SELECT city, NULL, SUM(quantity) FROM dealer GROUP BY city
+UNION ALL
+SELECT NULL, car_model, SUM(quantity) FROM dealer GROUP BY car_model
+UNION ALL
+SELECT NULL, NULL, SUM(quantity) FROM dealer;
+```
+
+**性能优势**：`GROUPING SETS` 在内部只扫描一次数据，比 `UNION ALL` 更高效。
+
+***
+
+### 3.3 识别汇总层级
+
+使用 `COALESCE` 函数替换 NULL 值，使汇总行更易读：
+
+```sql
+SELECT 
+    COALESCE(city, 'All Cities') AS city,
+    COALESCE(car_model, 'All Models') AS car_model,
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY ROLLUP(city, car_model)
+ORDER BY city, car_model
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city       | car\_model   | total\_quantity |
+| ---------- | ------------ | --------------- |
+| All Cities | All Models   | 78              |
+| Dublin     | All Models   | 33              |
+| Dublin     | Honda Accord | 10              |
+| Dublin     | Honda CRV    | 3               |
+| Dublin     | Honda Civic  | 20              |
+| Fremont    | All Models   | 32              |
+| Fremont    | Honda Accord | 15              |
+| Fremont    | Honda CRV    | 7               |
+| Fremont    | Honda Civic  | 10              |
+| San Jose   | All Models   | 13              |
+| San Jose   | Honda Accord | 8               |
+| San Jose   | Honda Civic  | 5               |
+
+***
+
+## 四、ROLLUP（层级汇总）
+
+### 4.1 基本概念
+
+`ROLLUP` 用于生成从细粒度到全局的层级汇总。对于 `ROLLUP(a, b, c)`，将生成以下分组集：
+
+* `(a, b, c)` - 最细粒度
+* `(a, b)` - 中间层级
+* `(a)` - 高层级
+* `()` - 全局汇总
+
+**等价关系**：
+
+```sql
+ROLLUP(city, car_model) 
+-- 等价于
+GROUPING SETS (
+    (city, car_model),
+    (city),
+    ()
+)
+```
+
+***
+
+### 4.2 销售层级汇总
+
+生成"城市→车型"两级汇总：
+
+```sql
+SELECT 
+    city, 
+    car_model, 
+    SUM(quantity) AS total_quantity 
+FROM dealer
+GROUP BY ROLLUP(city, car_model)
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | car\_model   | total\_quantity |
+| -------- | ------------ | --------------- |
+| Fremont  | Honda Civic  | 10              |
+| Fremont  | Honda Accord | 15              |
+| Fremont  | Honda CRV    | 7               |
+| Dublin   | Honda Civic  | 20              |
+| Dublin   | Honda Accord | 10              |
+| Dublin   | Honda CRV    | 3               |
+| San Jose | Honda Civic  | 5               |
+| San Jose | Honda Accord | 8               |
+| Fremont  | NULL         | 32              |
+| Dublin   | NULL         | 33              |
+| San Jose | NULL         | 13              |
+| NULL     | NULL         | 78              |
+
+**数据层级**：
+
+1. **明细层**（前8行）：每个城市的每种车型的销量
+2. **城市小计**（car\_model=NULL）：每个城市的总销量
+3. **总计**（city=NULL, car\_model=NULL）：所有城市的总销量
+
+***
+
+### 4.3 使用 GROUPING() 函数标识层级
+
+`GROUPING()` 函数返回 0 或 1，用于区分真实 NULL 和聚合标记：
+
+* **返回 0**：该列参与分组，值为实际数据
+* **返回 1**：该列未参与分组，NULL 表示汇总
+
+```sql
+SELECT 
+    city,
+    car_model,
+    SUM(quantity) AS total_quantity,
+    GROUPING(city) AS city_grouping,
+    GROUPING(car_model) AS model_grouping
+FROM dealer
+GROUP BY ROLLUP(city, car_model)
+ORDER BY city_grouping, city, model_grouping, car_model
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | car\_model   | total\_quantity | city\_grouping | model\_grouping |
+| -------- | ------------ | --------------- | -------------- | --------------- |
+| Dublin   | Honda Accord | 10              | 0              | 0               |
+| Dublin   | Honda CRV    | 3               | 0              | 0               |
+| Dublin   | Honda Civic  | 20              | 0              | 0               |
+| Dublin   | NULL         | 33              | 0              | 1               |
+| Fremont  | Honda Accord | 15              | 0              | 0               |
+| Fremont  | Honda CRV    | 7               | 0              | 0               |
+| Fremont  | Honda Civic  | 10              | 0              | 0               |
+| Fremont  | NULL         | 32              | 0              | 1               |
+| San Jose | Honda Accord | 8               | 0              | 0               |
+| San Jose | Honda Civic  | 5               | 0              | 0               |
+| San Jose | NULL         | 13              | 0              | 1               |
+| NULL     | NULL         | 78              | 1              | 1               |
+
+**GROUPING() 值解读**：
+
+* (**0, 0**)：明细数据，两列都参与分组
+* (**0, 1**)：城市小计，仅 city 参与分组
+* (**1, 1**)：全局总计，两列都不参与分组
+
+***
+
+## 五、CUBE（全维度交叉分析）
+
+### 5.1 基本概念
+
+`CUBE` 生成指定列的所有可能组合的分组集。对于 `CUBE(a, b)`，将生成 2² = 4 个分组集：
+
+* `(a, b)` - 明细
+* `(a)` - 按 a 汇总
+* `(b)` - 按 b 汇总
+* `()` - 全局汇总
+
+**等价关系**：
+
+```sql
+CUBE(city, car_model)
+-- 等价于
+GROUPING SETS (
+    (city, car_model),
+    (city),
+    (car_model),
+    ()
+)
+```
+
+***
+
+### 5.2 多维交叉分析
+
+生成城市和车型的所有维度组合：
+
+```sql
+SELECT 
+    city, 
+    car_model, 
+    SUM(quantity) AS total_quantity 
+FROM dealer
+GROUP BY city, car_model WITH CUBE
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | car\_model   | total\_quantity |
+| -------- | ------------ | --------------- |
+| Fremont  | Honda Civic  | 10              |
+| Fremont  | Honda Accord | 15              |
+| Fremont  | Honda CRV    | 7               |
+| Dublin   | Honda Civic  | 20              |
+| Dublin   | Honda Accord | 10              |
+| Dublin   | Honda CRV    | 3               |
+| San Jose | Honda Civic  | 5               |
+| San Jose | Honda Accord | 8               |
+| NULL     | Honda CRV    | 10              |
+| NULL     | Honda Civic  | 35              |
+| NULL     | Honda Accord | 33              |
+| Fremont  | NULL         | 32              |
+| Dublin   | NULL         | 33              |
+| San Jose | NULL         | 13              |
+| NULL     | NULL         | 78              |
+
+**结果包含**：
+
+1. **明细行**（前8行）：城市+车型组合
+2. **车型汇总行**（city=NULL）：跨城市的车型销量
+3. **城市汇总行**（car\_model=NULL）：跨车型的城市销量
+4. **总计行**（city=NULL, car\_model=NULL）：全局总计
+
+***
+
+### 5.3 标识聚合层级
+
+使用 `CASE` 语句结合 `GROUPING()` 函数标识每行的聚合层级：
+
+```sql
+SELECT 
+    city,
+    car_model,
+    SUM(quantity) AS total_quantity,
+    CASE 
+        WHEN GROUPING(city) = 1 AND GROUPING(car_model) = 1 THEN 'Total'
+        WHEN GROUPING(city) = 0 AND GROUPING(car_model) = 1 THEN 'City Subtotal'
+        WHEN GROUPING(city) = 1 AND GROUPING(car_model) = 0 THEN 'Model Subtotal'
+        ELSE 'Detail'
+    END AS aggregation_level
+FROM dealer
+GROUP BY CUBE(city, car_model)
+ORDER BY 
+    GROUPING(city),
+    GROUPING(car_model),
+    city,
+    car_model
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | car\_model   | total\_quantity | aggregation\_level |
+| -------- | ------------ | --------------- | ------------------ |
+| Dublin   | Honda Accord | 10              | Detail             |
+| Dublin   | Honda CRV    | 3               | Detail             |
+| Dublin   | Honda Civic  | 20              | Detail             |
+| Fremont  | Honda Accord | 15              | Detail             |
+| Fremont  | Honda CRV    | 7               | Detail             |
+| Fremont  | Honda Civic  | 10              | Detail             |
+| San Jose | Honda Accord | 8               | Detail             |
+| San Jose | Honda Civic  | 5               | Detail             |
+| Dublin   | NULL         | 33              | City Subtotal      |
+| Fremont  | NULL         | 32              | City Subtotal      |
+| San Jose | NULL         | 13              | City Subtotal      |
+| NULL     | Honda Accord | 33              | Model Subtotal     |
+| NULL     | Honda CRV    | 10              | Model Subtotal     |
+| NULL     | Honda Civic  | 35              | Model Subtotal     |
+| NULL     | NULL         | 78              | Total              |
+
+此查询清晰地标识了每行数据所属的聚合层级。
+
+***
+
+### 5.4 ROLLUP vs CUBE
+
+| 维度              | ROLLUP         | CUBE                |
+| --------------- | -------------- | ------------------- |
+| **生成分组集**       | 层级子集（从左到右）     | 所有可能组合              |
+| **ROLLUP(a,b**) | (a,b), (a), () | (a,b), (a), (b), () |
+| **分组数量**        | n+1（n为列数）      | 2ⁿ                  |
+| **使用场景**        | 层级报表、自上而下分析    | 多维交叉分析、BI 透视表       |
+| **性能**          | 生成分组数少，性能更优    | 生成分组数多，性能开销较大       |
+
+**选择建议**：
+
+* 需要**层级汇总**（如 地区→城市→门店）时使用 `ROLLUP`
+* 需要**多维交叉分析**（如 地区×产品×时间）时使用 `CUBE`
+* 需要**自定义组合**时使用 `GROUPING SETS`
+
+***
+
+## 六、HAVING 子句：过滤聚合结果
+
+### 6.1 WHERE vs HAVING
+
+| 子句         | 执行时机 | 作用对象  | 可用函数            |
+| ---------- | ---- | ----- | --------------- |
+| **WHERE**  | 分组前  | 原始行数据 | 标量函数、列比较        |
+| **HAVING** | 分组后  | 聚合结果  | 聚合函数、GROUP BY 列 |
+
+**执行顺序**：FROM → WHERE → GROUP BY → HAVING → SELECT → ORDER BY → LIMIT
+
+***
+
+### 6.2 使用 HAVING 过滤聚合结果
+
+筛选总销量超过 15 辆的城市：
+
+```sql
+SELECT 
+    city,
+    COUNT(*) AS record_count,
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY city
+HAVING SUM(quantity) > 15
+ORDER BY total_quantity DESC
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city    | record\_count | total\_quantity |
+| ------- | ------------- | --------------- |
+| Dublin  | 3             | 33              |
+| Fremont | 3             | 32              |
+
+San Jose（总销量 13 辆）被过滤掉，因为不满足 `HAVING` 条件。
+
+***
+
+### 6.3 WHERE 与 HAVING 组合
+
+先用 `WHERE` 过滤原始数据，再用 `HAVING` 过滤聚合结果：
+
+```sql
+SELECT 
+    city,
+    COUNT(*) AS record_count,
+    SUM(quantity) AS total_quantity
+FROM dealer
+WHERE quantity >= 10  -- 先过滤：只统计销量≥10的记录
+GROUP BY city
+HAVING COUNT(*) >= 2  -- 再过滤：只显示记录数≥2的城市
+ORDER BY total_quantity DESC
+LIMIT 50;
+```
+
+**执行逻辑**：
+
+1. **WHERE 过滤**：排除 quantity < 10 的记录（过滤掉 4 条记录）
+2. **GROUP BY 分组**：对剩余记录按 city 分组
+3. **HAVING 过滤**：只保留记录数 ≥ 2 的分组
+
+***
+
+## 七、高级应用场景
+
+### 7.1 行转列：透视分析
+
+将车型从行转换为列，实现透视表效果：
+
+```sql
+SELECT 
+    city,
+    SUM(CASE WHEN car_model = 'Honda Civic' THEN quantity ELSE 0 END) AS civic_sales,
+    SUM(CASE WHEN car_model = 'Honda Accord' THEN quantity ELSE 0 END) AS accord_sales,
+    SUM(CASE WHEN car_model = 'Honda CRV' THEN quantity ELSE 0 END) AS crv_sales,
+    SUM(quantity) AS total_sales
+FROM dealer
+GROUP BY city
+ORDER BY total_sales DESC
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | civic\_sales | accord\_sales | crv\_sales | total\_sales |
+| -------- | ------------ | ------------- | ---------- | ------------ |
+| Dublin   | 20           | 10            | 3          | 33           |
+| Fremont  | 10           | 15            | 7          | 32           |
+| San Jose | 5            | 8             | 0          | 13           |
+
+**业务价值**：
+
+* 清晰展示每个城市各车型的销量分布
+* San Jose 没有 CRV 销售记录（显示为 0）
+* 适合用于 Excel 报表、BI 仪表板
+
+***
+
+### 7.2 占比分析：窗口函数与 GROUP BY
+
+计算每个车型在其所在城市的销量占比：
+
+```sql
+SELECT 
+    city,
+    car_model,
+    quantity,
+    SUM(quantity) OVER (PARTITION BY city) AS city_total,
+    ROUND(quantity * 100.0 / SUM(quantity) OVER (PARTITION BY city), 2) AS percentage
+FROM dealer
+ORDER BY city, car_model
+LIMIT 50;
+```
+
+**查询结果**：
+
+| city     | car\_model   | quantity | city\_total | percentage |
+| -------- | ------------ | -------- | ----------- | ---------- |
+| Dublin   | Honda Accord | 10       | 33          | 30.30      |
+| Dublin   | Honda CRV    | 3        | 33          | 9.09       |
+| Dublin   | Honda Civic  | 20       | 33          | 60.61      |
+| Fremont  | Honda Accord | 15       | 32          | 46.88      |
+| Fremont  | Honda CRV    | 7        | 32          | 21.88      |
+| Fremont  | Honda Civic  | 10       | 32          | 31.25      |
+| San Jose | Honda Accord | 8        | 13          | 61.54      |
+| San Jose | Honda Civic  | 5        | 13          | 38.46      |
+
+**业务洞察**：
+
+* **Dublin**：Honda Civic 占比 60.61%，是绝对主力车型
+* **Fremont**：销量分布较均衡，Accord 占比最高（46.88%）
+* **San Jose**：Accord 占比 61.54%，市场偏好明显
+
+**技术说明**：窗口函数 `SUM() OVER (PARTITION BY city)` 计算每个城市的总销量，但不改变行数。
+
+***
+
+### 7.3 同比环比分析模式
+
+虽然示例数据不包含时间维度，但提供标准的同比/环比分析 SQL 模式：
+
+```sql
+-- 示例：按月统计销量，计算环比增长
+SELECT 
+    DATE_FORMAT(sale_date, 'yyyy-MM') AS month,
+    SUM(quantity) AS monthly_sales,
+    LAG(SUM(quantity)) OVER (ORDER BY DATE_FORMAT(sale_date, 'yyyy-MM')) AS prev_month_sales,
+    ROUND(
+        (SUM(quantity) - LAG(SUM(quantity)) OVER (ORDER BY DATE_FORMAT(sale_date, 'yyyy-MM'))) * 100.0 / 
+        LAG(SUM(quantity)) OVER (ORDER BY DATE_FORMAT(sale_date, 'yyyy-MM')), 
+        2
+    ) AS mom_growth_rate
+FROM sales_table
+GROUP BY DATE_FORMAT(sale_date, 'yyyy-MM')
+ORDER BY month;
+```
+
+**关键技术**：
+
+* `LAG()` 窗口函数获取上一行的值
+* 环比增长率 = (当前值 - 上期值) / 上期值 × 100%
+
+***
+
+### 7.4 TopN 分析：每组取前 N
+
+使用窗口函数结合 GROUP BY 实现分组 TopN：
+
+```sql
+-- 示例：每个城市销量前 2 的车型
+SELECT *
+FROM (
+    SELECT 
+        city,
+        car_model,
+        quantity,
+        ROW_NUMBER() OVER (PARTITION BY city ORDER BY quantity DESC) AS rank
+    FROM dealer
+) ranked
+WHERE rank <= 2
+ORDER BY city, rank;
+```
+
+**预期逻辑**（基于当前数据）：
+
+* Dublin：Honda Civic (20), Honda Accord (10)
+* Fremont：Honda Accord (15), Honda Civic (10)
+* San Jose：Honda Accord (8), Honda Civic (5)
+
+***
+
+## 八、性能优化与最佳实践
+
+### 8.1 查询性能优化
+
+#### 1. 合理使用 LIMIT
+
+对于大数据集，始终添加 `LIMIT` 限制返回行数：
+
+```sql
+-- 推荐
+SELECT city, SUM(quantity) 
+FROM large_table 
+GROUP BY city
+LIMIT 100;
+
+-- 避免：可能返回数百万行
+SELECT city, SUM(quantity) 
+FROM large_table 
+GROUP BY city;
+```
+
+#### 2. 选择合适的聚合工具
+
+| 需求     | 推荐方案            | 生成分组数 | 性能    |
+| ------ | --------------- | ----- | ----- |
+| 简单层级汇总 | `ROLLUP`        | n+1   | ⭐⭐⭐⭐⭐ |
+| 多维交叉分析 | `CUBE`          | 2ⁿ    | ⭐⭐⭐   |
+| 自定义分组集 | `GROUPING SETS` | 自定义   | ⭐⭐⭐⭐  |
+
+**示例**：对于 3 列分组
+
+* `ROLLUP(a,b,c)` 生成 4 个分组集
+* `CUBE(a,b,c)` 生成 8 个分组集
+* `GROUPING SETS` 可自定义任意组合
+
+#### 3. 降低分组列基数
+
+对高基数列（如 ID、timestamp）使用函数降低粒度：
+
+```sql
+-- 避免：按秒级 timestamp 分组
+GROUP BY timestamp  -- 可能产生数百万个分组
+
+-- 推荐：按小时或天分组
+GROUP BY DATE_FORMAT(timestamp, 'yyyy-MM-dd')
+GROUP BY DATE_TRUNC('hour', timestamp)
+```
+
+#### 4. WHERE 过滤优于 HAVING 过滤
+
+尽量在 `WHERE` 中过滤数据，减少 `GROUP BY` 处理的数据量：
+
+```sql
+-- 推荐：先过滤再分组
+SELECT city, SUM(quantity)
+FROM dealer
+WHERE car_model = 'Honda Civic'  -- 先过滤，减少数据量
+GROUP BY city;
+
+-- 避免：先分组再过滤（处理了不必要的数据）
+SELECT city, car_model, SUM(quantity)
+FROM dealer
+GROUP BY city, car_model
+HAVING car_model = 'Honda Civic';
+```
+
+***
+
+### 8.2 常见错误与解决方案
+
+#### 错误 1：SELECT 列未在 GROUP BY 中
+
+```sql
+-- ❌ 错误示例
+SELECT city, car_model, quantity
+FROM dealer
+GROUP BY city;
+-- 错误：car_model 和 quantity 未分组或聚合
+
+-- ✅ 正确方案1：添加到 GROUP BY
+SELECT city, car_model, SUM(quantity)
+FROM dealer
+GROUP BY city, car_model;
+
+-- ✅ 正确方案2：使用聚合函数
+SELECT city, MAX(car_model), SUM(quantity)
+FROM dealer
+GROUP BY city;
+```
+
+#### 错误 2：WHERE 中使用聚合函数
+
+```sql
+-- ❌ 错误示例
+SELECT city, SUM(quantity)
+FROM dealer
+WHERE SUM(quantity) > 20  -- WHERE 不能使用聚合函数
+GROUP BY city;
+
+-- ✅ 正确方案：使用 HAVING
+SELECT city, SUM(quantity)
+FROM dealer
+GROUP BY city
+HAVING SUM(quantity) > 20;
+```
+
+#### 错误 3：混淆 ROLLUP/CUBE 语法
+
+```sql
+-- ❌ 错误示例
+GROUP BY city, ROLLUP(car_model)  -- 不支持混合语法
+
+-- ✅ 正确方案
+GROUP BY ROLLUP(city, car_model)
+```
+
+#### 错误 4：忽略 NULL 值影响
+
+```sql
+-- 在 ROLLUP/CUBE 中，NULL 既可能是真实 NULL，也可能是汇总标记
+-- 使用 GROUPING() 函数区分
+
+SELECT 
+    CASE WHEN GROUPING(city) = 1 THEN 'Total' ELSE city END AS city,
+    SUM(quantity)
+FROM dealer
+GROUP BY ROLLUP(city);
+```
+
+***
+
+### 8.3 数据质量检查
+
+#### 检查分组完整性
+
+验证分组汇总是否等于总计：
+
+```sql
+-- 方法1：使用 ROLLUP 自动生成总计行
+SELECT 
+    COALESCE(city, 'TOTAL') AS city,
+    SUM(quantity) AS total
+FROM dealer
+GROUP BY ROLLUP(city);
+
+-- 方法2：手动验证
+WITH group_totals AS (
+    SELECT SUM(quantity) AS grouped_sum
+    FROM dealer
+    GROUP BY city
+),
+grand_total AS (
+    SELECT SUM(quantity) AS total_sum
+    FROM dealer
+)
+SELECT 
+    (SELECT SUM(grouped_sum) FROM group_totals) AS sum_of_groups,
+    (SELECT total_sum FROM grand_total) AS grand_total,
+    CASE 
+        WHEN (SELECT SUM(grouped_sum) FROM group_totals) = (SELECT total_sum FROM grand_total)
+        THEN 'PASS'
+        ELSE 'FAIL'
+    END AS validation_result;
+```
+
+***
+
+## 九、实战案例汇总
+
+### 9.1 销售分析报表
+
+**需求**：生成包含明细、小计、总计的销售报表
+
+```sql
+SELECT 
+    COALESCE(city, 'All Cities') AS city,
+    COALESCE(car_model, 'All Models') AS car_model,
+    SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY ROLLUP(city, car_model)
+ORDER BY 
+    GROUPING(city),
+    city,
+    GROUPING(car_model),
+    car_model
+LIMIT 50;
+```
+
+***
+
+### 9.2 多维交叉分析
+
+**需求**：BI 透视表，需要城市和车型的所有维度组合
+
+```sql
+SELECT 
+    city,
+    car_model,
+    SUM(quantity) AS total_quantity,
+    CASE 
+        WHEN GROUPING(city) = 1 AND GROUPING(car_model) = 1 THEN 'Grand Total'
+        WHEN GROUPING(city) = 1 THEN 'Model Total'
+        WHEN GROUPING(car_model) = 1 THEN 'City Total'
+        ELSE 'Detail'
+    END AS level
+FROM dealer
+GROUP BY CUBE(city, car_model)
+ORDER BY 
+    GROUPING(city),
+    GROUPING(car_model),
+    city,
+    car_model
+LIMIT 50;
+```
+
+***
+
+### 9.3 Top/Bottom N 分析
+
+**需求**：找出销量最高和最低的城市
+
+```sql
+-- Top 2 城市
+SELECT city, SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY city
+ORDER BY total_quantity DESC
+LIMIT 2;
+
+-- Bottom 2 城市
+SELECT city, SUM(quantity) AS total_quantity
+FROM dealer
+GROUP BY city
+ORDER BY total_quantity ASC
+LIMIT 2;
+```
+
+***
+
+### 9.4 占比与排名
+
+**需求**：计算每个车型的销量占比和排名
+
+```sql
+SELECT 
+    car_model,
+    SUM(quantity) AS total_sales,
+    ROUND(SUM(quantity) * 100.0 / (SELECT SUM(quantity) FROM dealer), 2) AS percentage,
+    RANK() OVER (ORDER BY SUM(quantity) DESC) AS rank
+FROM dealer
+GROUP BY car_model
+ORDER BY total_sales DESC
+LIMIT 50;
+```
+
+***
+
+## 十、功能对比与选择指南
+
+### 10.1 四种分组方式对比
+
+| 特性        | GROUP BY | GROUPING SETS | ROLLUP | CUBE     |
+| --------- | -------- | ------------- | ------ | -------- |
+| **语法复杂度** | 简单       | 中等            | 简单     | 简单       |
+| **生成分组数** | 1        | 自定义           | n+1    | 2ⁿ       |
+| **灵活性**   | 低        | 高             | 中      | 中        |
+| **性能**    | 最优       | 优             | 优      | 较差（列数多时） |
+| **适用场景**  | 基础统计     | 自定义多维分析       | 层级报表   | 全维交叉分析   |
+
+***
+
+### 10.2 选择决策树
+
+```
+是否需要多维度汇总？
+├─ 否 → 使用标准 GROUP BY
+└─ 是 → 需要哪些维度组合？
+    ├─ 所有可能组合 → 使用 CUBE
+    ├─ 层级汇总（从细到总）→ 使用 ROLLUP
+    └─ 自定义特定组合 → 使用 GROUPING SETS
+```
+
+***
+
+### 10.3 推荐使用场景
+
+| 业务场景     | 推荐方案                 | 示例           |
+| -------- | -------------------- | ------------ |
+| 日常统计报表   | `GROUP BY`           | 按部门统计销售额     |
+| 财务分层汇总   | `ROLLUP`             | 公司→部门→团队层级汇总 |
+| BI 数据透视表 | `CUBE`               | 产品×地区×时间三维分析 |
+| 自定义业务报表  | `GROUPING SETS`      | 同时输出日报、周报、月报 |
+| 占比分析     | `GROUP BY` + 窗口函数    | 各产品销量占比      |
+| Top N 分析 | `GROUP BY` + `LIMIT` | 销量前10的产品     |
+
+***
+
+## 十一、进阶技巧
+
+### 11.1 动态分组
+
+使用变量或子查询实现动态分组条件（示例逻辑）：
+
+```sql
+-- 根据销量范围动态分组
+SELECT 
+    CASE 
+        WHEN SUM(quantity) >= 30 THEN 'High Volume'
+        WHEN SUM(quantity) >= 15 THEN 'Medium Volume'
+        ELSE 'Low Volume'
+    END AS volume_tier,
+    COUNT(DISTINCT city) AS city_count
+FROM (
+    SELECT city, SUM(quantity) AS city_total
+    FROM dealer
+    GROUP BY city
+) city_summary
+GROUP BY 
+    CASE 
+        WHEN city_total >= 30 THEN 'High Volume'
+        WHEN city_total >= 15 THEN 'Medium Volume'
+        ELSE 'Low Volume'
+    END;
+```
+
+***
+
+### 11.2 递归汇总（模拟）
+
+虽然云器 Lakehouse 目前还不直接支持递归 CTE，但可通过 UNION ALL 模拟多层级汇总：
+
+```sql
+-- 三层汇总：明细、城市小计、总计
+SELECT city, car_model, quantity, 'Detail' AS level
+FROM dealer
+UNION ALL
+SELECT city, NULL, SUM(quantity), 'City Subtotal'
+FROM dealer
+GROUP BY city
+UNION ALL
+SELECT NULL, NULL, SUM(quantity), 'Grand Total'
+FROM dealer
+ORDER BY 
+    CASE level 
+        WHEN 'Detail' THEN 1
+        WHEN 'City Subtotal' THEN 2
+        WHEN 'Grand Total' THEN 3
+    END,
+    city,
+    car_model;
+```
+
+***
+
+### 11.3 NULL 值处理技巧
+
+#### 技巧 1：区分真实 NULL 和聚合 NULL
+
+```sql
+SELECT 
+    city,
+    car_model,
+    SUM(quantity) AS total,
+    CASE 
+        WHEN GROUPING(city) = 1 THEN '[Subtotal]'
+        WHEN city IS NULL THEN '[Real NULL]'
+        ELSE city
+    END AS city_display
+FROM dealer
+GROUP BY ROLLUP(city, car_model);
+```
+
+#### 技巧 2：使用 COALESCE 美化输出
+
+```sql
+SELECT 
+    COALESCE(city, '📊 All Cities') AS city,
+    COALESCE(car_model, '🚗 All Models') AS car_model,
+    SUM(quantity) AS total
+FROM dealer
+GROUP BY CUBE(city, car_model);
+```
+
+***
+
+## 十二、总结
+
+### 核心要点
+
+1. **GROUP BY** 是数据聚合分析的基础，掌握其用法是 SQL 分析的必备技能
+2. **GROUPING SETS / ROLLUP / CUBE** 提供了强大的多维分析能力，一次查询生成多层级汇总
+3. **HAVING** 用于过滤聚合结果，与 WHERE 配合实现灵活的数据筛选
+4. **窗口函数** 与 GROUP BY 结合，可实现复杂的占比、排名、同环比分析
+5. **性能优化** 需要权衡灵活性与效率，合理选择分组方式
+
+***
+
+### 学习路径建议
+
+1. **基础阶段**：掌握标准 GROUP BY 和常用聚合函数
+2. **进阶阶段**：学习 GROUPING SETS、ROLLUP、CUBE 的使用场景
+3. **高级阶段**：结合窗口函数、CTE 实现复杂业务分析
+4. **实战阶段**：应用于实际业务场景，优化查询性能
+
+***
+
+### 快速参考
+
+```sql
+-- 基础分组
+SELECT column, AGG_FUNC(column) FROM table GROUP BY column;
+
+-- 多维分组
+SELECT col1, col2, AGG_FUNC(col3) 
+FROM table 
+GROUP BY GROUPING SETS ((col1, col2), (col1), ());
+
+-- 层级汇总
+SELECT col1, col2, AGG_FUNC(col3) 
+FROM table 
+GROUP BY ROLLUP(col1, col2);
+
+-- 交叉分析
+SELECT col1, col2, AGG_FUNC(col3) 
+FROM table 
+GROUP BY col1, col2 WITH CUBE;
+
+-- 条件过滤
+SELECT col1, AGG_FUNC(col2) 
+FROM table 
+WHERE condition
+GROUP BY col1 
+HAVING AGG_FUNC(col2) > value;
+```
+
+^