在hadoop中运行cte join时，结果是0行。

WITH 

cte1 as 
(
    SELECT 
        product_segment,
        'a' as AT
    FROM t1
    WHERE data_type = 'v1'
),

cte2 as 
(
    SELECT 'a' as AT
    FROM t1
    WHERE data_type = 'v2'
)

SELECT a.product_segment
FROM cte1 a 
INNER JOIN cte2 b ON (a.AT = b.AT)

我知道t1有一堆数据，CTE根据一个维度将表拆分为2。有很多重叠。即使我使用左连接，它也返回零行。
你能加入hadoop中的ctes吗？