sql-queries

SQL 查询技能

在所有主流数据仓库方言中编写正确、高性能且易读的 SQL。

方言专属参考

PostgreSQL（包括 Aurora、RDS、Supabase、Neon）

日期/时间:

-- Current date/time
CURRENT_DATE, CURRENT_TIMESTAMP, NOW()

-- Date arithmetic
date_column + INTERVAL '7 days'
date_column - INTERVAL '1 month'

-- Truncate to period
DATE_TRUNC('month', created_at)

-- Extract parts
EXTRACT(YEAR FROM created_at)
EXTRACT(DOW FROM created_at)  -- 0=Sunday

-- Format
TO_CHAR(created_at, 'YYYY-MM-DD')

字符串函数:

-- Concatenation
first_name || ' ' || last_name
CONCAT(first_name, ' ', last_name)

-- Pattern matching
column ILIKE '%pattern%'  -- case-insensitive
column ~ '^regex_pattern$'  -- regex

-- String manipulation
LEFT(str, n), RIGHT(str, n)
SPLIT_PART(str, delimiter, position)
REGEXP_REPLACE(str, pattern, replacement)

数组与 JSON:

-- JSON access
data->>'key'  -- text
data->'nested'->'key'  -- json
data#>>'{path,to,key}'  -- nested text

-- Array operations
ARRAY_AGG(column)
ANY(array_column)
array_column @> ARRAY['value']

性能建议:

• 使用 EXPLAIN ANALYZE 分析查询
• 在经常用于过滤或连接的列上创建索引
• 对相关子查询优先使用 EXISTS 而非 IN
• 为常见过滤条件使用部分索引
• 并发访问时使用连接池

---

Snowflake

日期/时间:

-- Current date/time
CURRENT_DATE(), CURRENT_TIMESTAMP(), SYSDATE()

-- Date arithmetic
DATEADD(day, 7, date_column)
DATEDIFF(day, start_date, end_date)

-- Truncate to period
DATE_TRUNC('month', created_at)

-- Extract parts
YEAR(created_at), MONTH(created_at), DAY(created_at)
DAYOFWEEK(created_at)

-- Format
TO_CHAR(created_at, 'YYYY-MM-DD')

字符串函数:

-- Case-insensitive by default (depends on collation)
column ILIKE '%pattern%'
REGEXP_LIKE(column, 'pattern')

-- Parse JSON
column:key::string  -- dot notation for VARIANT
PARSE_JSON('{"key": "value"}')
GET_PATH(variant_col, 'path.to.key')

-- Flatten arrays/objects
SELECT f.value FROM table, LATERAL FLATTEN(input => array_col) f

半结构化数据:

-- VARIANT type access
data:customer:name::STRING
data:items[0]:price::NUMBER

-- Flatten nested structures
SELECT
    t.id,
    item.value:name::STRING as item_name,
    item.value:qty::NUMBER as quantity
FROM my_table t,
LATERAL FLATTEN(input => t.data:items) item

性能建议:

• 在大表上使用聚簇键（而非传统索引）
• 过滤聚簇键列以实现分区裁剪
• 根据查询复杂度设置合适的仓库大小
• 使用 RESULT_SCAN(LAST_QUERY_ID()) 避免重复运行高成本查询
• 对暂存/临时数据使用临时表（transient table）

---

BigQuery（Google Cloud）

日期/时间:

-- Current date/time
CURRENT_DATE(), CURRENT_TIMESTAMP()

-- Date arithmetic
DATE_ADD(date_column, INTERVAL 7 DAY)
DATE_SUB(date_column, INTERVAL 1 MONTH)
DATE_DIFF(end_date, start_date, DAY)
TIMESTAMP_DIFF(end_ts, start_ts, HOUR)

-- Truncate to period
DATE_TRUNC(created_at, MONTH)
TIMESTAMP_TRUNC(created_at, HOUR)

-- Extract parts
EXTRACT(YEAR FROM created_at)
EXTRACT(DAYOFWEEK FROM created_at)  -- 1=Sunday

-- Format
FORMAT_DATE('%Y-%m-%d', date_column)
FORMAT_TIMESTAMP('%Y-%m-%d %H:%M:%S', ts_column)

字符串函数:

-- No ILIKE, use LOWER()
LOWER(column) LIKE '%pattern%'
REGEXP_CONTAINS(column, r'pattern')
REGEXP_EXTRACT(column, r'pattern')

-- String manipulation
SPLIT(str, delimiter)  -- returns ARRAY
ARRAY_TO_STRING(array, delimiter)

数组与结构体:

-- Array operations
ARRAY_AGG(column)
UNNEST(array_column)
ARRAY_LENGTH(array_column)
value IN UNNEST(array_column)

-- Struct access
struct_column.field_name

性能建议:

• 始终过滤分区列（通常为日期）以减少扫描字节数
• 对分区内经常过滤的列使用聚簇
• 对大规模基数估算使用 APPROX_COUNT_DISTINCT()
• 避免使用 SELECT * -- 计费按扫描字节数计算
• 在参数化脚本中使用 DECLARE 和 SET
• 执行大型查询前先通过试运行（dry run）预估查询成本

---

Redshift（Amazon）

日期/时间:

-- Current date/time
CURRENT_DATE, GETDATE(), SYSDATE

-- Date arithmetic
DATEADD(day, 7, date_column)
DATEDIFF(day, start_date, end_date)

-- Truncate to period
DATE_TRUNC('month', created_at)

-- Extract parts
EXTRACT(YEAR FROM created_at)
DATE_PART('dow', created_at)

字符串函数:

-- Case-insensitive
column ILIKE '%pattern%'
REGEXP_INSTR(column, 'pattern') > 0

-- String manipulation
SPLIT_PART(str, delimiter, position)
LISTAGG(column, ', ') WITHIN GROUP (ORDER BY column)

性能建议:

• 设计分布键（DISTKEY）以实现共置连接
• 对经常过滤的列使用排序键（SORTKEY）
• 使用 EXPLAIN 检查查询计划
• 避免跨节点数据移动（注意 DS_BCAST 和 DS_DIST）
• 定期执行 ANALYZE 和 VACUUM
• 使用延迟绑定视图以获得更灵活的模式管理

---

Databricks SQL

日期/时间:

-- Current date/time
CURRENT_DATE(), CURRENT_TIMESTAMP()

-- Date arithmetic
DATE_ADD(date_column, 7)
DATEDIFF(end_date, start_date)
ADD_MONTHS(date_column, 1)

-- Truncate to period
DATE_TRUNC('MONTH', created_at)
TRUNC(date_column, 'MM')

-- Extract parts
YEAR(created_at), MONTH(created_at)
DAYOFWEEK(created_at)

Delta Lake 功能:

-- Time travel
SELECT * FROM my_table TIMESTAMP AS OF '2024-01-15'
SELECT * FROM my_table VERSION AS OF 42

-- Describe history
DESCRIBE HISTORY my_table

-- Merge (upsert)
MERGE INTO target USING source
ON target.id = source.id
WHEN MATCHED THEN UPDATE SET *
WHEN NOT MATCHED THEN INSERT *

性能建议:

• 使用 Delta Lake 的 OPTIMIZE 和 ZORDER 提升查询性能
• 对计算密集型查询利用 Photon 引擎
• 对频繁访问的数据集使用 CACHE TABLE
• 按低基数日期列进行分区

---

常见 SQL 模式

窗口函数

-- Ranking
ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY created_at DESC)
RANK() OVER (PARTITION BY category ORDER BY revenue DESC)
DENSE_RANK() OVER (ORDER BY score DESC)

-- Running totals / moving averages
SUM(revenue) OVER (ORDER BY date_col ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) as running_total
AVG(revenue) OVER (ORDER BY date_col ROWS BETWEEN 6 PRECEDING AND CURRENT ROW) as moving_avg_7d

-- Lag / Lead
LAG(value, 1) OVER (PARTITION BY entity ORDER BY date_col) as prev_value
LEAD(value, 1) OVER (PARTITION BY entity ORDER BY date_col) as next_value

-- First / Last value
FIRST_VALUE(status) OVER (PARTITION BY user_id ORDER BY created_at ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING)
LAST_VALUE(status) OVER (PARTITION BY user_id ORDER BY created_at ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING)

-- Percent of total
revenue / SUM(revenue) OVER () as pct_of_total
revenue / SUM(revenue) OVER (PARTITION BY category) as pct_of_category

使用 CTE 提升可读性

WITH
-- Step 1: Define the base population
base_users AS (
    SELECT user_id, created_at, plan_type
    FROM users
    WHERE created_at >= DATE '2024-01-01'
      AND status = 'active'
),

-- Step 2: Calculate user-level metrics
user_metrics AS (
    SELECT
        u.user_id,
        u.plan_type,
        COUNT(DISTINCT e.session_id) as session_count,
        SUM(e.revenue) as total_revenue
    FROM base_users u
    LEFT JOIN events e ON u.user_id = e.user_id
    GROUP BY u.user_id, u.plan_type
),

-- Step 3: Aggregate to summary level
summary AS (
    SELECT
        plan_type,
        COUNT(*) as user_count,
        AVG(session_count) as avg_sessions,
        SUM(total_revenue) as total_revenue
    FROM user_metrics
    GROUP BY plan_type
)

SELECT * FROM summary ORDER BY total_revenue DESC;

同期群留存

WITH cohorts AS (
    SELECT
        user_id,
        DATE_TRUNC('month', first_activity_date) as cohort_month
    FROM users
),
activity AS (
    SELECT
        user_id,
        DATE_TRUNC('month', activity_date) as activity_month
    FROM user_activity
)
SELECT
    c.cohort_month,
    COUNT(DISTINCT c.user_id) as cohort_size,
    COUNT(DISTINCT CASE
        WHEN a.activity_month = c.cohort_month THEN a.user_id
    END) as month_0,
    COUNT(DISTINCT CASE
        WHEN a.activity_month = c.cohort_month + INTERVAL '1 month' THEN a.user_id
    END) as month_1,
    COUNT(DISTINCT CASE
        WHEN a.activity_month = c.cohort_month + INTERVAL '3 months' THEN a.user_id
    END) as month_3
FROM cohorts c
LEFT JOIN activity a ON c.user_id = a.user_id
GROUP BY c.cohort_month
ORDER BY c.cohort_month;

漏斗分析

WITH funnel AS (
    SELECT
        user_id,
        MAX(CASE WHEN event = 'page_view' THEN 1 ELSE 0 END) as step_1_view,
        MAX(CASE WHEN event = 'signup_start' THEN 1 ELSE 0 END) as step_2_start,
        MAX(CASE WHEN event = 'signup_complete' THEN 1 ELSE 0 END) as step_3_complete,
        MAX(CASE WHEN event = 'first_purchase' THEN 1 ELSE 0 END) as step_4_purchase
    FROM events
    WHERE event_date >= CURRENT_DATE - INTERVAL '30 days'
    GROUP BY user_id
)
SELECT
    COUNT(*) as total_users,
    SUM(step_1_view) as viewed,
    SUM(step_2_start) as started_signup,
    SUM(step_3_complete) as completed_signup,
    SUM(step_4_purchase) as purchased,
    ROUND(100.0 * SUM(step_2_start) / NULLIF(SUM(step_1_view), 0), 1) as view_to_start_pct,
    ROUND(100.0 * SUM(step_3_complete) / NULLIF(SUM(step_2_start), 0), 1) as start_to_complete_pct,
    ROUND(100.0 * SUM(step_4_purchase) / NULLIF(SUM(step_3_complete), 0), 1) as complete_to_purchase_pct
FROM funnel;

去重

-- Keep the most recent record per key
WITH ranked AS (
    SELECT
        *,
        ROW_NUMBER() OVER (
            PARTITION BY entity_id
            ORDER BY updated_at DESC
        ) as rn
    FROM source_table
)
SELECT * FROM ranked WHERE rn = 1;

错误处理与调试

查询失败时：

1. 语法错误：检查是否使用了方言专属语法（例如 BigQuery 中不支持 ILIKE，SAFE_DIVIDE 仅 BigQuery 可用）
2. 找不到列：对照模式核实列名 -- 检查拼写错误和大小写敏感问题（PostgreSQL 中带引号的标识符区分大小写）
3. 类型不匹配：比较不同类型时显式转换（CAST(col AS DATE)、col::DATE）
4. 除零错误：使用 NULLIF(denominator, 0) 或方言专属的安全除法函数
5. 列名歧义：在 JOIN 中始终使用表别名限定列名
6. GROUP BY 错误：所有非聚合列都必须出现在 GROUP BY 中（BigQuery 允许按别名分组，属例外）