编辑“︁Apache Drill Parquet文件查询”︁（章节）

== 实际案例 ==

=== 日志分析场景 ===
假设有按日期分区的web日志Parquet文件：

<syntaxhighlight lang="sql">
-- 分析错误请求
SELECT 
    request_url,
    COUNT(*) as error_count,
    AVG(response_time) as avg_time
FROM dfs.`/logs/web/year=2023/month=*/day=*`
WHERE status_code >= 400
GROUP BY request_url
ORDER BY error_count DESC
LIMIT 10;
</syntaxhighlight>

=== 电商数据分析 ===
分析用户购买行为：

<syntaxhighlight lang="sql">
-- 用户购买频率分析
WITH user_purchases AS (
    SELECT 
        user_id,
        COUNT(DISTINCT order_id) as purchase_count,
        SUM(amount) as total_spent
    FROM dfs.`/data/ecommerce/transactions.parquet`
    WHERE transaction_date BETWEEN '2023-01-01' AND '2023-12-31'
    GROUP BY user_id
)
SELECT 
    CASE 
        WHEN purchase_count <= 3 THEN '低频'
        WHEN purchase_count <= 10 THEN '中频'
        ELSE '高频'
    END as frequency,
    COUNT(*) as user_count,
    AVG(total_spent) as avg_spent
FROM user_purchases
GROUP BY frequency
ORDER BY avg_spent DESC;
</syntaxhighlight>