Postgres json_to_record() function

You can use the json_to_record function to convert a top-level JSON object into a row, with the type specified by the AS clause.

This function is useful when you need to parse JSON data received from external sources, such as APIs or file uploads, and store it in a structured format. By using json_to_record, you can easily extract values from JSON and map them to the corresponding columns in your database table.

Function signature

json_to_record(json JSON) AS (column_name column_type [, ...])

The function's definition includes a column definition list, where you specify the name and data type of each column in the resulting record.

Example usage

Consider a scenario in which you have JSON data representing employee information, and you want to ingest it for easier processing later. The JSON data looks like this:

{
  "id": "123",
  "name": "John Doe",
  "department": "Engineering",
  "salary": "75000"
}

The table you want to insert data into is defined as follows:

CREATE TABLE employees (
    id INT,
    name TEXT,
    department TEXT,
    salary NUMERIC
);

Using json_to_record, you can insert the input data into the employees table as shown:

INSERT INTO employees
SELECT *
FROM json_to_record('{"id": "123", "name": "John Doe", "department": "Engineering", "salary": "75000"}') AS x(id INT, name TEXT, department TEXT, salary NUMERIC);

To verify the data was inserted, you can run the following query:

SELECT * FROM employees;

This query returns the following result:

| id | name     | department   | salary |
|----|----------|--------------|--------|
| 123| John Doe | Engineering  | 75000  |

Advanced examples

This section provides advanced json_to_record examples.

Handling partial data with json_to_record

For datapoints where the JSON objects have missing keys, json_to_record can still cast them into records, producing NULL values for the unmatched columns. For example:

INSERT INTO employees
SELECT *
FROM json_to_record('{
  "id": "124",
  "name": "Jane Smith"
}') AS x(id INT, name TEXT, department TEXT, salary NUMERIC)
RETURNING *;

This query returns the following result:

| id | name       | department   | salary |
|----|------------|--------------|--------|
| 124| Jane Smith |              |        |

Handling nested data with json_to_record

json_to_record can also be used to handle nested JSON input data (i.e., keys with values that are JSON objects themselves). You need to first define a custom Postgres type. The newly created type can then be used in the column definition list along with the other columns.

In the following example, we handle the address field by creating an ADDRESS_TYPE type first.

CREATE TYPE ADDRESS_TYPE AS (
  street TEXT,
  city TEXT
);

SELECT *
FROM json_to_record('{
  "id": "125",
  "name": "Emily Clark",
  "department": "Marketing",
  "salary": "68000",
  "address": {
    "street": "123 Elm St",
    "city": "Springfield"
  }
}') AS x(id INT, name TEXT, department TEXT, salary NUMERIC, address ADDRESS_TYPE);

This query returns the following result:

| id | name        | department | salary | address                     |
|----|-------------|------------|--------|-----------------------------|
| 1  | Emily Clark | Marketing  | 68000  | ("123 Elm St", Springfield) |

Alternative functions

• json_populate_record: This function can also be used to create records using values from a JSON object. The difference is that json_populate_record requires the record type to be defined beforehand, while json_to_record needs the type definition inline.

• json_to_recordset: This function can be used similarly to parse JSON, the difference being that it returns a set of records instead of a single record. For example, if you have an array of JSON objects, you can use json_to_recordset to convert each object into a new row.

• jsonb_to_record: This function provides the same functionality as json_to_record, but accepts JSONB input instead of JSON. In cases where the input payload type isn't exactly specified, either of the two functions can be used. For example, take this json_to_record query:

  SELECT *
  FROM json_to_record('{"id": "123", "name": "John Doe", "department": "Engineering"}') 
  AS x(id INT, name TEXT, department TEXT);

  It works just as well as this JSONB variant (below) since Postgres casts the literal JSON object to JSON or JSONB depending on the context.

  SELECT *
  FROM jsonb_to_record('{"id": "123", "name": "Sally", "department": "Engineering"}')
  AS x(id INT, name TEXT, department TEXT);

Resources

• PostgreSQL documentation: JSON functions