Spark 2.0将json读入数据帧，并在键中带有引号-与Spark 1.6的行为不同……错误？

Question

We are in the unfortunate situation of having to deal with messy incoming json data, and have found a difference in the way that Spark 2.0 (pyspark) handles quotes within a json key.

If we use the following as a sample file (sample.json):

{"event":"abc"}
{"event":"xyz","otherdata[\"this.is.ugly\"]":"value1"}

In Spark 1.6.2, we can run the following and get results:

from pyspark import  SparkConf
from pyspark.sql import SQLContext
conf = SparkConf()
conf.setAppName('temp_quotes')

sc = SparkContext(conf=conf)
sqlContext = SQLContext(sc)

data =     sqlContext.read.json("sample.json")
data.printSchema()

Results are:

root
 |-- event: string (nullable = true)
 |-- otherdata["this.is.ugly"]: string (nullable = true)

And we can see data when we do a show:

data.show(2)

+-----+-------------------------+
|event|otherdata["this.is.ugly"]|
+-----+-------------------------+
|  abc|                     null|
|  xyz|                   value1|
+-----+-------------------------+

However, running the same code in Spark 2.0 shows the same schema:

from pyspark import  SparkConf
from pyspark.sql import SQLContext
conf = SparkConf()
conf.setAppName('temp_quotes')

sc = SparkContext(conf=conf)
sqlContext = SQLContext(sc)

data =     sqlContext.read.json("sample.json")
data.printSchema()

root
 |-- event: string (nullable = true)
 |-- otherdata["this.is.ugly"]: string (nullable = true)

But the show fails:

data.show(2)

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/spark/python/pyspark/sql/dataframe.py", line 287, in show
    print(self._jdf.showString(n, truncate))
  File "/usr/lib/spark/python/lib/py4j-0.10.1-src.zip/py4j/java_gateway.py", line 933, in __call__
  File "/usr/lib/spark/python/pyspark/sql/utils.py", line 69, in deco
    raise AnalysisException(s.split(': ', 1)[1], stackTrace)
pyspark.sql.utils.AnalysisException: u'Unable to resolve otherdata["this.is.ugly"] given [event, otherdata["this.is.ugly"]];'

Is this a bug or is there a parameter in Spark 2.0 that I'm missing?

Answer 1

I believe this is addressed in https://issues.apache.org/jira/browse/SPARK-16698 (dot in JSON keys). The fix is scheduled to be released in 2.0.1

(I don't have enough reputation to comment)