Spark - 替换字符串中的第一次出现

Question

我想在 spark scala sql 中使用 replaceFirst() 函数。要么是否可以在 spark scala 数据帧中使用 replaceFirst() 函数？

不使用 UDF 是否可行？

我要做的功能是：

println("abcdefgbchijkl".replaceFirst("bc","**BC**"))
// a**BC**defgbchijkl

但是DataFrame的Column Type不能用Function来应用：

var test0 = Seq("abcdefgbchijkl").toDF("col0")

test0
.select(col("col0").replaceFirst("bc","**BC**"))
.show(false)
/*
<console>:230: error: value replaceFirst is not a member of org.apache.spark.sql.Column
       .select(col("col0").replaceFirst("bc","**BC**"))
*/

此外，我不知道如何以 SQL 形式使用它：

%sql
-- How to use replaceFirst()
select replaceFirst()

Answer 1

替换第一次出现并不是我看到的 Spark 开箱即用的支持，但可以通过组合几个函数来实现：

Spark >= 3.0.0

import org.apache.spark.sql.functions.{array_join, col, split}

val test0 = Seq("abcdefgbchijkl").toDF("col0") // replaced `var` with `val`

val stringToReplace = "bc"
val replacement = "**BC**"

test0
  // create a temporary column, splitting the string by the first occurrence of `bc`
  .withColumn("temp", split(col("col0"), stringToReplace, 2))
  // recombine the strings before and after `bc` with the desired replacement
  .withColumn("col0", array_join(col("temp"), replacement))
  // we no longer need this `temp` column
  .drop(col("temp"))
  .show(false)

给出：

+------------------+
|col0              |
+------------------+
|a**BC**defgbchijkl|
+------------------+

对于（火花）SQL：

-- recombine the strings before and after `bc` with the desired replacement
SELECT tempr[0] || "**BC**" || tempr[1] AS col0
FROM (
  -- create a temporary column, splitting the string by the first occurrence of `bc`
  SELECT split(col0, "bc", 2) AS tempr
  FROM (
    SELECT 'abcdefgbchijkl' AS col0
  )
)

Spark < 3.0.0（2020 年之前，使用 Spark 2.4.5 测试）

val test0 = Seq("abcdefgbchijkl").toDF("col0")

val stringToReplace = "bc"
val replacement = "**BC**"

val splitFirst = udf { (s: String) => s.split(stringToReplace, 2) }

spark.udf.register("splitFirst", splitFirst) // if you're using Spark SQL

test0
  // create a temporary column, splitting the string by the first occurrence of `bc`
  .withColumn("temp", splitFirst(col("col0")))
  // recombine the strings before and after `bc` with the desired replacement
  .withColumn("col0", array_join(col("temp"), replacement))
  // we no longer need this `temp` column
  .drop(col("temp"))
  .show(false)

给出：

+------------------+
|col0              |
+------------------+
|a**BC**defgbchijkl|
+------------------+

对于（火花）SQL：

-- recombine the strings before and after `bc` with the desired replacement
SELECT tempr[0] || "**BC**" || tempr[1] AS col0
FROM (
  -- create a temporary column, splitting the string by the first occurrence of `bc`
  SELECT splitFirst(col0) AS tempr -- `splitFirst` was registered above
  FROM (
    SELECT 'abcdefgbchijkl' AS col0
  )
)

Spark - 替换字符串中的第一次出现

Spark - Replace first occurrence in a string

scala

apache-spark

apache-spark-sql

Spark >= 3.0.0

Spark < 3.0.0（2020 年之前，使用 Spark 2.4.5 测试）