CODEWITHSUNDEEP
apache-sparkpyspark
Hardik Gupta
Hardik Gupta

Reputation: 4790

PySpark: Convert String to Array of String for a column

I have a dataframe like this

data = [(('ID1', "[apples, mangos, eggs, milk, oranges]")),
   (('ID1', "[eggs, milk, cereals, mangos, apples]"))]
df = spark.createDataFrame(data, ['ID', "colval"])
df.show(truncate=False)
df.printSchema()

+---+-------------------------------------+
|ID |colval                               |
+---+-------------------------------------+
|ID1|[apples, mangos, eggs, milk, oranges]|
|ID1|[eggs, milk, cereals, mangos, apples]|
+---+-------------------------------------+

root
 |-- ID: string (nullable = true)
 |-- colval: string (nullable = true)

I want to convert colval to type Array of String

And when I take the first element after split, it returns me the same result as first. Any help?

root
 |-- ID: string (nullable = true)
 |-- colval: array (nullable = true)
 |    |-- element: string (containsNull = true)

I tried using split, however end up getting this result

df = df.withColumn('colval', split('colval', "', ?'"))
df.show(truncate = False)
df.printSchema()

+---+---------------------------------------+
|ID |colval                                 |
+---+---------------------------------------+
|ID1|[[apples, mangos, eggs, milk, oranges]]|
|ID1|[[eggs, milk, cereals, mangos, apples]]|
+---+---------------------------------------+

root
 |-- ID: string (nullable = true)
 |-- colval: array (nullable = true)
 |    |-- element: string (containsNull = true)

Upvotes: 1

Views: 275

Answers (1)

anky
anky

Reputation: 75080

You can replace the [ and ] and then split:

df.withColumn("colval",F.split(F.regexp_replace("colval",r"\[|\]",""),",")).show()

+---+-----------------------------------------+
|ID |colval                                   |
+---+-----------------------------------------+
|ID1|[apples,  mangos,  eggs,  milk,  oranges]|
|ID1|[eggs,  milk,  cereals,  mangos,  apples]|
+---+-----------------------------------------+


root
 |-- ID: string (nullable = true)
 |-- colval: array (nullable = true)
 |    |-- element: string (containsNull = true)

Incase you want to trim after splitting, you can use higher order functions after splitting :

(df.withColumn("colval",F.split(F.regexp_replace("colval",r"\[|\]",""),","))
.withColumn("colval",F.expr("transform(colval,x-> trim(x))")))

verification and difference between approach 1 and 2 (Note extra spaces)

df.withColumn("colval",F.split(F.regexp_replace("colval",r"\[|\]",""),",")).collect()
[Row(ID='ID1', colval=['apples', ' mangos', ' eggs', ' milk', ' oranges']),
 Row(ID='ID1', colval=['eggs', ' milk', ' cereals', ' mangos', ' apples'])]


(df.withColumn("colval",F.split(F.regexp_replace("colval",r"\[|\]",""),","))
 .withColumn("colval",F.expr("transform(colval,x-> trim(x))"))).collect()

[Row(ID='ID1', colval=['apples', 'mangos', 'eggs', 'milk', 'oranges']),
 Row(ID='ID1', colval=['eggs', 'milk', 'cereals', 'mangos', 'apples'])]

Upvotes: 2

Related Questions