Parsing JSON with the  Extract Nested Data component within Matillion Data Productivity Cloud connected to Snowflake simplifies the parsing for many semi-structured data patterns.   The JSON format has become a more popular format for semi-structured data, primarily because it is more consistent containing all key:value pairs. JSON handles repeating elements by containing them in an array as a value of a key:value pair.   For this article, I am using the same example data set that was used in part one on XML only this sample data is represented as JSON. I also walk you through how to convert the XML to JSON to simplify parsing XML.  Extract Nested Data  We start by using the Extract Nested Data component, which simplifies parsing semi-structured data. In this example, we’re using several of them to traverse the nested elements.   First, the JSON file is loaded into a table called donut_json, which contains a single column defined as a variant “data_value.”  Next, configure the Columns property of the Extract Nested component. I used “Autofill”’ and let the component identify the structure of the JSON. I have deselected all the columns and chosen to pass through the Item attributes and element values. In the example, I also passed through the Filling element, keeping it a variant for further processing downstream.  Since the topping elements are repeating at the first level, the component has flattened toppings into separate rows automatically, so I was able to select the element value level for toppings.   Another property to call out is the Outer join property on the Configuration tab. Since all of the elements do not exist for every item, I needed to set Outer Join = “Yes.”  This will retain all the rows for all items, even though only two items have Fillings.   Flatten Variant  The Flatten Variant component is used to flatten arrays. Although the Extract Nested Data component can sometimes be used, the Flatten Variant lets you explicitly break a column into more rows than the original extract nested data if you are seeking further granularity from the extract nested component.  The batter element in this example has two formats, so I have to treat the Batter array differently by using a Flatten Variant component to parse the array of batters into separate rows. The initial Extract Nested Data component created a new row for each item and each topping. From there, we want a new row for each item, topping and batter. I tested the batter element to determine if it’s an array, by using the IS_ARRAY() function in a Calculator component.   IS_ARRAY("items_item-element_batters_batter") After that,  Flatten the array into separate rows per batter element before extracting the attributes.   Set the Column Flatten property to read the batter array column In the column mappings, use the flatten alias to map to an output variant column   Finally, we bring all the rows back together, remove unwanted columns, and write to a new table. The Unite component unions all the rows back together The Rename component allows us to remove any unwanted fields, like the arrays, and rename and reorder the fields  The Rewrite component writes to a new table   The resulting final pipeline is much simpler than the previous XML one.   Convert XML to JSON Our example pipeline started with a file that was already in a JSON format.  However, if you have an XML file that needs to be converted and you would like to convert the XML to JSON inside a pipeline, you’ll use the code below.  Create an Orchestration Pipeline First, I created a separate Orchestration pipeline that contains a SQL Script component to create a Snowflake UDF using the code below. This code calls a Snowflake Snowpark package called “xmltodict.”  Our example XML_to_JSON Python code follows. Parse With the Calculator Component Next in my Transformation pipeline, I called the procedure in a Calculator component. The parse_json function formats the JSON so it’s readable.  Normalizing Semi-Structured Data  Semi-structured files typically contain data