Spark - extracting numeric values from an alphanumeric string using regex

Question

I have an alphanumeric column named "Result" that I'd like to parse into 4 different columns: prefix, suffix, value, and pure_text.

I'd like to solve this using Spark SQL using RLIKE and REGEX, but also open to PySpark/Scala

pure_text: contains only alphabets (or) if there are numbers present, then they should either have a special character "-" or multiple decimals (i.e. 9.9.0) or number followed by an alphabet and then a number again (i.e. 3x4u)

prefix: anything that can't be categorized into "pure_text" will be taken into consideration. any character(s) before the 1st digit [0-9] needs to be extracted.

suffix: anything that can't be categorized into "pure_text" will be taken into consideration. any character(s) after the last digit [0-9] needs to be extracted.

value: anything that can't be categorized into "pure_text" will be taken into consideration. extract all numerical values including the decimal point.

Result 

11 H
111L
<.004
>= 0.78
val<=0.6
xyz 100 abc
1-9
aaa 100.3.4 
a1q1

Expected Output:

Result         Prefix     Suffix   Value    Pure_Text

11 H                           H      11
111L                           L     111       
.9                                   0.9
<.004               <              0.004
>= 0.78            >=               0.78
val<=0.6        val<=                0.6
xyz 100 abc      xyz         abc     100
1-9                                              1-9
aaa 100.3.4                              aaa 100.3.4 
a1q1                                            a1q1

Leo C · Accepted Answer · 2019-03-17 17:54:43Z

Here's one approach using a UDF that applies pattern matching to extract the string content into a case class. The pattern matching centers around the numeric value with Regex pattern [+-]?(?:\d*\.)?\d+ to extract the first occurrence of numbers like "1.23", ".99", "-100", etc. A subsequent check of numbers in the remaining substring captured in suffix determines whether the numeric substring in the original string is legitimate.

import org.apache.spark.sql.functions._
import spark.implicits._

case class RegexRes(prefix: String, suffix: String, value: Option[Double], pure_text: String)

val regexExtract = udf{ (s: String) =>
  val pattern = """(.*?)([+-]?(?:\d*\.)?\d+)(.*)""".r
  s match {
    case pattern(pfx, num, sfx) =>
      if (sfx.exists(_.isDigit))
        RegexRes("", "", None, s)
      else
        RegexRes(pfx, sfx, Some(num.toDouble), "")
    case _ =>
      RegexRes("", "", None, s)
  }
}

val df = Seq(
  "11 H", "111L", ".9", "<.004", ">= 0.78", "val<=0.6", "xyz 100 abc", "1-9", "aaa 100.3.4", "a1q1"
).toDF("result")

df.
  withColumn("regex_res", regexExtract($"result")).
  select($"result", $"regex_res.prefix", $"regex_res.suffix", $"regex_res.value", $"regex_res.pure_text").
  show
// +-----------+------+------+-----+-----------+
// |     result|prefix|suffix|value|  pure_text|
// +-----------+------+------+-----+-----------+
// |       11 H|      |     H| 11.0|           |
// |       111L|      |     L|111.0|           |
// |         .9|      |      |  0.9|           |
// |      <.004|     <|      |0.004|           |
// |    >= 0.78|   >= |      | 0.78|           |
// |   val<=0.6| val<=|      |  0.6|           |
// |xyz 100 abc|  xyz |   abc|100.0|           |
// |        1-9|      |      | null|        1-9|
// |aaa 100.3.4|      |      | null|aaa 100.3.4|
// |       a1q1|      |      | null|       a1q1|
// +-----------+------+------+-----+-----------+

Collectives™ on Stack Overflow

Spark - extracting numeric values from an alphanumeric string using regex

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related