Spark.sql 按 MAX 过滤行

Question

下面是源文件的一部分，您可以想象它更大：

date,code1,postcode,cityname,total
2020-03-27,2011,X700,Curepipe,44
2020-03-29,2011,X700,Curepipe,44
2020-03-26,2011,X700,Curepipe,22
2020-03-27,2035,X920,vacoas,3
2020-03-25,2011,X920,vacoas,1
2020-03-24,2122,X760,souillac,22
2020-03-23,2122,X760,souillac,11
2020-03-22,2257,X760,souillac,10
2020-03-27,2480,X510,rosehill,21
2020-03-22,2035,X510,rosehill,7
2020-03-20,2035,X510,rosehill,3

以下代码后：

#Load data
from pyspark.sql import SparkSession
spark = SparkSession.builder.master("local").appName("source").getOrCreate()
dfcases = spark.read.format("csv").option("header", "true").load("sourcefile.csv")

dfcases.createOrReplaceTempView("tablecases")
spark.sql(XXXXXXXXXXXXX).show() #Mysql code to insert

我想得到这个结果：

Curepipe,X700,2020-03-27,44
Curepipe,X700,2020-03-29,44
souillac,X760,2020-03-24,22
rosehill,X510,2020-03-27,21
vacoas,X920,2020-03-27,3

目的是：

Select 每个城市名称的总数达到 MAX 的日期（注意，如果一个城市在 2 个不同日期的总数达到 MAX，则可以出现两次），
按总降序排序，然后按日期升序排序，然后按城市名称升序排序。

谢谢！

Answer 1

以下查询生成您想要的输出

SELECT cityname, postcode, date, COUNT(*) AS total 
FROM tablecases 
GROUP BY cityname, postcode, date 
HAVING COUNT(*) > 1
ORDER BY COUNT(*) DESC, date, cityname

演示在 db<>fiddle

Answer 2

您可以在请求中使用 SQL window 来获得结果，如下所示：

SELECT
  cityname, 
  postcode, 
  date, 
  total
FROM
 (SELECT 
    cityname, 
    postcode, 
    date, 
    total, 
    MAX(total) OVER (PARTITION BY cityname ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) AS max_total 
  FROM tablecases)
WHERE max_total = total
ORDER BY max_total DESC, date, cityname

Spark.sql 按 MAX 过滤行

Spark.sql Filter rows by MAX

sql

apache-spark

pyspark

apache-spark-sql