Adding Custom Schema to Spark Dataframe

In the last blog, we have loaded our data to Spark Dataframe. We have also used “inferschema” option to let spark figure out the schema of the Dataframe on its own. But in many cases, you would like to specify a schema for Dataframe. This will give you much better control over column names and especially data types. Let us see how we can add our custom schema while reading data in Spark.

Adding Custom Schema

In spark, schema is array StructField of type StructType. Each StructType has 4 parameters.

  • Column Name
  • Data type of that column
  • Boolean value indication if values in this column can be null or not
  • Metadata column – this is optional column which can be used to add additional information about column

Let us write our custom schema for our flights data.

Now our data frame has column names and data types which we have specified. If you want to print schema for any dataframe you can use below function.

Using Metadata With Custom Schema

We can add extra information about columns using the metadata filed. This filed takes key-value pairs and we can choose any number of keys and values depending on our needs.

We can check our data frame and its schema now.

custom schema with metadata
Custom schema with Metadata

If you want to check schema with its metadata then we need to use following code. We can read all of schema with this function or also read schema for one column as well.

This is how we can add a custom schema to our dataframes. I hope this helps. See you in the next blog.

Similar Posts

Leave a Reply

Your email address will not be published.


  1. Hi mahesh

    I have txt file say temp.txt with following data
    b,(1,2)“csv”) \
    .option(“header”,False) \
    Now my question is how to define schema for such type of dataset.

    1. Hello Vicky,
      Thanks for reaching out.
      To solve this issue you need to such data enclosed in quotes

      Then you can use option(“escapeQuotes”, “true”) while reading data.

      If this is not possible, then it is slightly difficult to get it working.
      You can read entire row as single column. And then format data according to your needs.

      df =\
      .option(“header”, “true”)\
      .option(“inferSchema”, “true”)\
      .option(“sep”, “\n”)\

      | a,(1,2)|
      | b,(1,2)|

      You will need to run some transformations to get that data in required format.
      Please let me know if you need my help in this.

  2. Hi Mahesh

    I see you have defined schema in the code itself. How about importing a schema from some json file while reading dataframe.

    from pyspark.sql.types import StructField, StructType, StringType,LongType
    custom_schema =
    df ="csv") \
    .schema(custom_schema) \
    .option("header", True) \