CWR&L

今天终于把窗口函数的执行流程搞清楚了。

核心理解：窗口函数不会像 GROUP BY 那样折叠行，而是在每一行上计算一个"窗口"范围内的聚合值，行数保持不变。

SELECT
  name,
  department,
  salary,
  RANK() OVER (PARTITION BY department ORDER BY salary DESC) AS dept_rank,
  AVG(salary) OVER (PARTITION BY department) AS dept_avg
FROM employees;

关键点：

PARTITION BY 相当于分组，但不折叠
ORDER BY 决定窗口内的排序和帧范围
执行顺序：WHERE → GROUP BY → HAVING → 窗口函数 → ORDER BY → LIMIT