@clickzetta/cz-cli-darwin-x64 0.3.92 → 0.3.94

package/bin/skills/clickzetta-sql-syntax-guide/references/vs-spark.md

@@ -1,229 +0,0 @@
-# ClickZetta Lakehouse vs Spark SQL 差异
-
-> 来源：产品文档 + Spark Connector 文档
-
-## 数据类型映射
-
-| ClickZetta | Spark SQL | 说明 |
-|---|---|---|
-| `BOOLEAN` | `BooleanType` | 相同 |
-| `TINYINT` | `ByteType` | 1字节 |
-| `SMALLINT` | `ShortType` | 2字节 |
-| `INT` | `IntegerType` | 4字节 |
-| `BIGINT` | `LongType` | 8字节 |
-| `FLOAT` | `FloatType` | 4字节浮点 |
-| `DOUBLE` | `DoubleType` | 8字节浮点 |
-| `DECIMAL(p,s)` | `DecimalType(p,s)` | 精确数值 |
-| `STRING` / `VARCHAR` | `StringType` | 字符串 |
-| `BINARY` | `BinaryType` | 二进制 |
-| `DATE` | `DateType` | 日期 |
-| `TIMESTAMP` | `TimestampType` | 带时区时间戳 |
-| `TIMESTAMP_NTZ` | `TimestampNTZType` | 无时区时间戳 |
-| `ARRAY<T>` | `ArrayType` | 数组 |
-| `MAP<K,V>` | `MapType` | 键值对 |
-| `STRUCT<f:T>` | `StructType` | 结构体 |
-
----
-
-## 建表语法差异
-
-### 分区
-
-```sql
--- Spark SQL：PARTITIONED BY
-CREATE TABLE orders (id INT, amount DECIMAL, dt STRING)
-USING PARQUET
-PARTITIONED BY (dt);
-
--- ClickZetta：相同语法，但不需要 USING 子句
-CREATE TABLE orders (id INT, amount DECIMAL, dt STRING)
-PARTITIONED BY (dt);
-```
-
-### Bucket（分桶）
-
-```sql
--- Spark SQL
-CREATE TABLE orders (id INT, amount DECIMAL)
-CLUSTERED BY (id) INTO 8 BUCKETS;
-
--- ClickZetta：相同语法
-CREATE TABLE orders (id INT, amount DECIMAL)
-CLUSTERED BY (id) INTO 8 BUCKETS;
-```
-
-### 表属性
-
-```sql
--- Spark SQL：TBLPROPERTIES
-CREATE TABLE orders (id INT)
-TBLPROPERTIES ('delta.enableChangeDataFeed' = 'true');
-
--- ClickZetta：PROPERTIES
-CREATE TABLE orders (id INT)
-PROPERTIES ('data_lifecycle' = '30');  -- 数据保留天数
-```
-
----
-
-## 查询语法差异
-
-### LATERAL VIEW（展开数组）
-
-```sql
--- 两者语法相同（ClickZetta 兼容 Hive/Spark 风格）
-SELECT id, skill
-FROM employees
-LATERAL VIEW EXPLODE(skills) t AS skill;
-
--- POSEXPLODE（带位置索引）
-SELECT id, pos, skill
-FROM employees
-LATERAL VIEW POSEXPLODE(skills) t AS pos, skill;
-```
-
-### 窗口函数
-
-```sql
--- 两者基本相同
-SELECT id, amount,
-    ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY created_at DESC) AS rn,
-    SUM(amount) OVER (PARTITION BY customer_id
-                      ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS running_total
-FROM orders;
-```
-
-### CTE（公用表表达式）
-
-```sql
--- 两者语法相同
-WITH
-    monthly_sales AS (
-        SELECT DATE_TRUNC('month', order_date) AS month, SUM(amount) AS total
-        FROM orders GROUP BY 1
-    ),
-    ranked AS (
-        SELECT *, RANK() OVER (ORDER BY total DESC) AS rnk FROM monthly_sales
-    )
-SELECT * FROM ranked WHERE rnk <= 3;
-```
-
-### STRUCT / ARRAY 操作
-
-```sql
--- Spark SQL
-SELECT address.city FROM users;                    -- STRUCT 字段访问
-SELECT skills[0] FROM employees;                   -- ARRAY 索引
-SELECT EXPLODE(skills) FROM employees;             -- 展开数组
-SELECT TRANSFORM(skills, x -> UPPER(x)) FROM emp; -- 数组变换
-
--- ClickZetta（相同语法）
-SELECT address.city FROM users;
-SELECT skills[0] FROM employees;
-SELECT EXPLODE(skills) FROM employees;
-SELECT TRANSFORM(skills, x -> UPPER(x)) FROM emp;
-```
-
----
-
-## 函数差异
-
-### 日期函数
-
-```sql
--- 两者基本兼容
-DATE_ADD(date, days)
-DATE_SUB(date, days)
-DATEDIFF(end_date, start_date)   -- 注意：ClickZetta 参数顺序与 Snowflake 相反
-DATE_TRUNC('month', date)
-DATE_FORMAT(date, 'yyyy-MM-dd')
-FROM_UNIXTIME(unix_ts)
-UNIX_TIMESTAMP(date_str)
-```
-
-### 字符串函数
-
-```sql
--- 两者基本兼容
-CONCAT(s1, s2, ...)
-CONCAT_WS(',', s1, s2, ...)
-SPLIT(str, ',')
-REGEXP_EXTRACT(str, pattern, group)
-REGEXP_REPLACE(str, pattern, replacement)
-INSTR(str, substr)
-SUBSTR(str, pos, len)
-TRIM(str) / LTRIM(str) / RTRIM(str)
-```
-
-### 聚合函数
-
-```sql
--- 两者基本兼容
-COUNT(*) / COUNT(DISTINCT col)
-SUM / AVG / MAX / MIN
-COLLECT_LIST(col)    -- Spark：返回数组（含重复）
-COLLECT_SET(col)     -- Spark：返回去重数组
-ARRAY_AGG(col)       -- ClickZetta：等价于 COLLECT_LIST
-```
-
----
-
-## ClickZetta 特有功能（Spark 无对应）
-
-```sql
--- 1. VCLUSTER（计算集群管理）
-CREATE VCLUSTER my_vc VCLUSTER_TYPE = ANALYTICS VCLUSTER_SIZE = 4;
-USE VCLUSTER my_vc;
-
--- 2. DYNAMIC TABLE（增量计算）
-CREATE DYNAMIC TABLE sales_summary
-    REFRESH INTERVAL 5 MINUTE VCLUSTER default_ap
-AS SELECT customer_id, SUM(amount) FROM orders GROUP BY 1;
-
--- 3. TABLE STREAM（CDC 变更捕获）
-CREATE TABLE STREAM orders_stream ON TABLE orders
-    WITH PROPERTIES ('TABLE_STREAM_MODE' = 'STANDARD');
-
--- 4. PIPE（持续导入）
-CREATE PIPE my_pipe
-    AS COPY INTO orders FROM VOLUME my_volume USING CSV;
-
--- 5. VECTOR 类型（向量检索）
-CREATE TABLE embeddings (id INT, vec VECTOR(FLOAT, 1024));
-SELECT id, cosine_distance(vec, vector(0.1, 0.2, ...)) AS dist
-FROM embeddings ORDER BY dist LIMIT 10;
-
--- 6. Time Travel
-SELECT * FROM orders TIMESTAMP AS OF '2024-01-01 00:00:00';
-RESTORE TABLE orders TO TIMESTAMP AS OF '2024-01-01 00:00:00';
-UNDROP TABLE orders;
-
--- 7. SHARE（跨实例数据共享）
-CREATE SHARE my_share;
-GRANT SELECT, READ METADATA ON TABLE public.orders TO SHARE my_share;
-```
-
----
-
-## Spark SQL 特有功能（ClickZetta 无对应或语法不同）
-
-```sql
--- 1. Delta Lake 特有语法（ClickZetta 无对应）
-OPTIMIZE table_name ZORDER BY (col);   -- ClickZetta 有 OPTIMIZE 但无 ZORDER
-VACUUM table_name RETAIN 168 HOURS;   -- ClickZetta 自动管理，无需手动 VACUUM
-
--- 2. SHOW TABLES EXTENDED（ClickZetta 无对应）
-SHOW TABLES EXTENDED IN schema LIKE 'orders*';
-
--- 3. DESCRIBE HISTORY（Delta）→ ClickZetta 用 DESC HISTORY
--- Spark/Delta:
-DESCRIBE HISTORY orders;
--- ClickZetta:
-DESC HISTORY orders;
-
--- 4. 生成列（语法不同）
--- Spark:
-CREATE TABLE orders (id INT, year INT GENERATED ALWAYS AS (YEAR(order_date)));
--- ClickZetta（相同语法，也支持）:
-CREATE TABLE orders (id INT, year INT GENERATED ALWAYS AS (YEAR(order_date)));
-```