转换字符串并与DF列值Spark Scala进行比较
我有一个字符串类型转换字符串并与DF列值Spark Scala进行比较,scala,apache-spark,apache-spark-sql,Scala,Apache Spark,Apache Spark Sql,我有一个字符串类型val current_dates=“a:2021-04-02,B:2021-04-02,C:2021-04-01,D:2021-04-01”。这里A、B、C、D是id字段及其对应的日期 现在,我有了一个输入数据框,其中有多条记录的id为&date列 val input_df = sc.parallelize(Seq(("A","2021-04-01"),("A","2021-04-02"),(&q
val current_dates=“a:2021-04-02,B:2021-04-02,C:2021-04-01,D:2021-04-01”
。这里A、B、C、D
是id字段及其对应的日期
现在,我有了一个输入数据框,其中有多条记录的id为&date列
val input_df = sc.parallelize(Seq(("A","2021-04-01"),("A","2021-04-02"),("B","2021-04-01"),("B","2021-04-02"),("C","2021-04-01"),("C","2021-04-02"),("D","2021-04-01"),("D","2021-04-02"))).toDF("id","create_date")
input_df.show()
+---+-----------+
| id|create_date|
+---+-----------+
| A| 2021-04-01|
| A| 2021-04-02|
| B| 2021-04-01|
| B| 2021-04-02|
| C| 2021-04-01|
| C| 2021-04-02|
| D| 2021-04-01|
| D| 2021-04-02|
+---+-----------+
现在,我想将每个记录的日期值与字符串中每个id的对应日期进行比较,并在dataFrame中导出新的日期列
expected_df.select((input_df.columns ++ Array("new_dt")).head, (input_df.columns ++ Array("new_dt")).tail: _*).orderBy("id").show()
+---+-----------+----------+
| id|create_date| new_dt|
+---+-----------+----------+
| A| 2021-04-01|2021-04-02|
| A| 2021-04-02|2021-04-02|
| B| 2021-04-02|2021-04-02|
| B| 2021-04-01|2021-04-02|
| C| 2021-04-02|2021-04-02|
| C| 2021-04-01|2021-04-01|
| D| 2021-04-01|2021-04-01|
| D| 2021-04-02|2021-04-02|
+---+-----------+----------+
目前,我正在将字符串转换为另一个数据帧,并将其与输入数据帧连接起来,然后以下面的方式导出新列
val current_dates_df = sc.parallelize(current_dates.split(",").map(_.split(":")).map{ case Array(a,b) => (a, b) }).toDF("previous_run_id", "previous_run_date")
current_dates_df.show()
+---------------+-----------------+
|previous_run_id|previous_run_date|
+---------------+-----------------+
| A| 2021-04-02|
| B| 2021-04-02|
| C| 2021-04-01|
| D| 2021-04-01|
+---------------+-----------------+
val deriveNewDt: UserDefinedFunction = udf[String, String, String]((create_date: String, previous_run_date: String) => {
val date_format: String = "yyyy-MM-dd"
val new_dt = {
if (new SimpleDateFormat(date_format).parse(create_date).after(new SimpleDateFormat(date_format).parse(previous_run_date))) create_date
else previous_run_date
}
new_dt
})
val joined_df = input_df.join(current_dates_df, input_df("id") === current_dates_df("previous_run_id"), "left_outer")
val expected_df = joined_df.withColumn("new_dt", deriveNewDt($"create_date", $"previous_run_date"))
expected_df.select((input_df.columns ++ Array("new_dt")).head, (input_df.columns ++ Array("new_dt")).tail: _*).show()
是否有更好的方法来处理字符串并在不将字符串转换为数据帧的情况下实现相同的功能。您可以使用来获取给定id
的日期,并使用magest
来获取两者之间的较晚日期:
val current_dates = "A:2021-04-02,B:2021-04-02,C:2021-04-01,D:2021-04-01"
val result = input_df.withColumn(
"new_dt",
expr(s"greatest(str_to_map('$current_dates, ',', ':')[id], create_date)")
)
result.show
+---+-----------+----------+
| id|create_date| new_dt|
+---+-----------+----------+
| A| 2021-04-01|2021-04-02|
| A| 2021-04-02|2021-04-02|
| B| 2021-04-01|2021-04-02|
| B| 2021-04-02|2021-04-02|
| C| 2021-04-01|2021-04-01|
| C| 2021-04-02|2021-04-02|
| D| 2021-04-01|2021-04-01|
| D| 2021-04-02|2021-04-02|
+---+-----------+----------+