Create DataFrame from list of tuples using pyspark

Question

I am working with data extracted from SFDC using simple-salesforce package. I am using Python3 for scripting and Spark 1.5.2.

I created an rdd containing the following data:

[('Id', 'a0w1a0000003xB1A'), ('PackSize', 1.0), ('Name', 'A')]
[('Id', 'a0w1a0000003xAAI'), ('PackSize', 1.0), ('Name', 'B')]
[('Id', 'a0w1a00000xB3AAI'), ('PackSize', 30.0), ('Name', 'C')]
...

This data is in RDD called v_rdd

My schema looks like this:

StructType(List(StructField(Id,StringType,true),StructField(PackSize,StringType,true),StructField(Name,StringType,true)))

I am trying to create DataFrame out of this RDD:

sqlDataFrame = sqlContext.createDataFrame(v_rdd, schema)

I print my DataFrame:

sqlDataFrame.printSchema()

And get the following:

+--------------------+--------------------+--------------------+
|                  Id|  PackSize|                          Name|
+--------------------+--------------------+--------------------+
|[Ljava.lang.Objec...|[Ljava.lang.Objec...|[Ljava.lang.Objec...|
|[Ljava.lang.Objec...|[Ljava.lang.Objec...|[Ljava.lang.Objec...|
|[Ljava.lang.Objec...|[Ljava.lang.Objec...|[Ljava.lang.Objec...|

I am expecting to see actual data, like this:

+------------------+------------------+--------------------+
|                Id|PackSize|                          Name|
+------------------+------------------+--------------------+
|a0w1a0000003xB1A  |               1.0|       A            |
|a0w1a0000003xAAI  |               1.0|       B            |
|a0w1a00000xB3AAI  |              30.0|       C            |

Can you please help me identify what I am doing wrong here.

My Python script is long, I am not sure it would be convenient for people to sift through it, so I posted only parts I am having issue with.

Thank a ton in advance!

Answer 1

Hey could you next time provide a working example. That would be easier.

The way how your RDD is presented is basically weird to create a DataFrame. This is how you create a DF according to Spark Documentation.

>>> l = [('Alice', 1)]
>>> sqlContext.createDataFrame(l).collect()
[Row(_1=u'Alice', _2=1)]
>>> sqlContext.createDataFrame(l, ['name', 'age']).collect()
[Row(name=u'Alice', age=1)]

So concerning your example you can create your desired output like this way:

# Your data at the moment
data = sc.parallelize([ 
[('Id', 'a0w1a0000003xB1A'), ('PackSize', 1.0), ('Name', 'A')],
[('Id', 'a0w1a0000003xAAI'), ('PackSize', 1.0), ('Name', 'B')],
[('Id', 'a0w1a00000xB3AAI'), ('PackSize', 30.0), ('Name', 'C')]
    ])
# Convert to tuple
data_converted = data.map(lambda x: (x[0][1], x[1][1], x[2][1]))

# Define schema
schema = StructType([
    StructField("Id", StringType(), True),
    StructField("Packsize", StringType(), True),
    StructField("Name", StringType(), True)
])

# Create dataframe
DF = sqlContext.createDataFrame(data_converted, schema)

# Output
DF.show()
+----------------+--------+----+
|              Id|Packsize|Name|
+----------------+--------+----+
|a0w1a0000003xB1A|     1.0|   A|
|a0w1a0000003xAAI|     1.0|   B|
|a0w1a00000xB3AAI|    30.0|   C|
+----------------+--------+----+

Hope this helps