Hive：错误计算 SUM 然后 MAX 分组项目

Question

我想运行一个查询来计算每张信用卡每个月的最大支出金额。对于每张信用卡，我需要计算每个月花费的金额。我有一张包含信用卡交易的表格credit_transact：

processdate timestamp   ""
cardno_hash string  ""
amount  int ""
year    int ""
month   int ""

虚构的样本数据：

card    year    month    amount
a123    2016    12       23160
a123    2016    10       287
c123    2016    11       5503
c123    2016    11       4206

我想要：

card    year    month    amount
a123    2016    12       23160
c123    2016    11       9709

一个重要的事情是年和月是分区列。

我尝试过如下子查询：

USE credit_card_db;
SELECT sum_amount_transact.cardno_hash, sum_amount_transact.year, sum_amount_transact.month, MAX(sum_amount_transact.sum_amount)
FROM
(
  SELECT cardno_hash, year, month, SUM(amount) AS sum_amount FROM credit_transact
  GROUP BY cardno_hash, year, month
) AS sum_amount_transact
GROUP BY sum_amount_transact.cardno_hash, sum_amount_transact.year;

但是，显示以下错误：

java.lang.Exception: org.apache.hive.service.cli.HiveSQLException: Error while compiling statement: FAILED: SemanticException Line 0:-1 Invalid column reference 'month'

以下子查询运行良好并按预期返回结果：

SELECT cardno_hash, year, month, SUM(amount) AS sum_amount FROM credit_transact
  GROUP BY cardno_hash, year, month

结果是：

card    year    month    amount
a123    2016    12       23160
a123    2016    10       287
c123    2016    11       9709

如果有人能帮助解决这个问题，我们将不胜感激。

Answer 1

我不太清楚你真正想要什么，但我很确定你想要row_number()。我想你想要每年最多的月份：

SELECT ct.*
FROM (SELECT cardno_hash, year, month, SUM(amount) AS sum_amount,
             ROW_NUMBER() OVER (PARTITION BY cardno_hash, year ORDER BY SUM(amount) DESC) as seqnum
      FROM credit_transact
      GROUP BY cardno_hash, year, month
     ) ct
WHERE seqnum = 1;