Parsing JSON with SQL Server

With SQL Server JSON support, we can now store JSON documents in a table. This is very similar to how we have been able to store XML documents in an Xml Column. In this article, we take a look at parsing a complex JSON document with nested structures to return a flat structure similar to a database query.

Define the JSON Document

We start by defining our JSON structure. The data may not make much sense, but the intend here is to be able to flatten this structure into a table format.  We want to be able to return every player name with the team and associated score for that team. Teams and scores are mapped by the index number on the array.  It may not be clear, but there are some challenges here. 

DECLARE @json nvarchar(1000) = N'[     {               "player":{"name":"ozkary"},         "scores":[7,9],               "teams": [                      {"name":"team one"},{"name":"team two"}                ]             },     {               "player":{"name":"dani"},         "scores":[6,10] ,               "teams": [                      {"name":"team a"},{"name":"team b"}                    ]      }  ]

For example, how can we get the name properties for both player and team’s fields? Let’s look at those fields, and we can see that one is an object. The other is an array of objects.  In addition, the scores are stored in an array of values with no correlation to anything else.

We can try to parse this data to see how it looks once it is flatten.  We start by querying an object property. This can be done by providing the document path to that property. In this case, we can get the player name with this path:

'$.player.name'

The dollar sign ($) provides the root scope of the document. From there, we walk the properties down the different levels.

When it comes to arrays, we need to do a cross apply with the array and return the data as JSON. This enables us to join with that document section and parse it. We continue to repeat that process until we get to the object of each array element. This is shown in the example below, we return the teams property as JSON, and we then cross apply on that field to select the name.

teams nvarchar(max) '$.teams' AS JSON, scores nvarchar(max) '$.scores' AS JSON CROSS APPLY OPENJSON(teams) WITH (     team nvarchar(50) '$.name' )

Now that we understand the approach, we can implement a  solution that can help us get the information. The entire query should look as follows:

DECLARE @json nvarchar(2500) = N'[     {               "player":{"name":"ozkary"},         "scores":[7,9],               "teams": [                      {"name":"team one"},{"name":"team two"}                ]              },     {               "player":{"name":"dani"},         "scores":[6,10] ,               "teams": [                      {"name":"team a"},{"name":"team b"}                    ]      }  ]'SELECT     player.name, team, score FROM OPENJSON (@json) WITH(       name nvarchar(50) '$.player.name',     teams nvarchar(max) '$.teams' AS JSON,        scores nvarchar(max) '$.scores' AS JSON ) as player CROSS APPLY OPENJSON(teams) WITH (     team nvarchar(50) '$.name' )CROSS APPLY OPENJSON(scores) WITH (     score nvarchar(50) '$' )

We first open the entire JSON document with the OPENJSON function.  We select all the fields we need using the WITH directive. For every field that we return as JSON, we do a cross apply and open that JSON segment.  This enables us to select the object properties that we need. 

The resulting data should look as show below:

As we shown here, processing JSON documents with SQL Server is feasible, but we need to be mindful that JSON structures are defined to match an application model and attempting to query the data as a table structure may be a difficult task.

Thanks for reading.

Originally published by ozkary.com