CODEWITHSUNDEEP
pythonjsondataframeapache-sparkpyspark
Sreedhar
Sreedhar

Reputation: 30045

Convert JSON Data in Spark DataFrame column into tabular format

I got spark dataframe which is loaded from a multiline JSON file.

One of the column (data) schema is as below:

 root
 |-- data: array (nullable = true)
 |    |-- element: struct (containsNull = true)
 |    |    |-- f: struct (nullable = true)
 |    |    |    |-- 0: struct (nullable = true)
 |    |    |    |    |-- v: double (nullable = true)
 |    |    |-- ts: string (nullable = true)

And sample Data:

array
0: {"f": {"0": {"v": 25.08}}, "ts": "2021-01-11T05:59:00.170Z"}
1: {"f": {"0": {"v": 25.92}}, "ts": "2021-03-22T03:29:00.170Z"}
2: {"f": {"0": {"v": 25.94}}, "ts": "2021-03-22T03:39:00.173Z"}
3: {"f": {"0": {"v": 25.95}}, "ts": "2021-03-22T03:49:00.170Z"}
4: {"f": {"0": {"v": 25.99}}, "ts": "2021-03-22T04:00:00.173Z"}

I want to just extract ts and v.

Example result

1

Upvotes: 0

Views: 859

Answers (1)

mck
mck

Reputation: 42422

You can explode the array of structs into multiple rows, and select the required struct elements:

import pyspark.sql.functions as F

df2 = df.select(F.explode('data').alias('data')).select('data.ts', 'data.f.0.v')

df2.show(truncate=False)
+------------------------+-----+
|ts                      |v    |
+------------------------+-----+
|2021-01-11T05:59:00.170Z|25.08|
|2021-03-22T03:29:00.170Z|25.92|
|2021-03-22T03:39:00.173Z|25.94|
|2021-03-22T03:49:00.170Z|25.95|
|2021-03-22T04:00:00.173Z|25.99|
+------------------------+-----+

Upvotes: 2

Related Questions