JSON_TABLE

JSON_TABLE is a table function that converts JSON data into a relational form.

Syntax

Description

JSON_TABLE can be used in contexts where a table reference can be used; in the FROM clause of a statement, and in multi-table / statements.

json_doc is the JSON document to extract data from. In the simplest case, it is a string literal containing JSON. In more complex cases it can be an arbitrary expression returning JSON. The expression may have references to columns of other tables. However, one can only refer to tables that precede this JSON_TABLE invocation. For RIGHT JOIN, it is assumed that its outer side precedes the inner. All tables in outer selects are also considered preceding.

context_path is a expression pointing to a collection of nodes in json_doc that will be used as the source of rows.

The COLUMNS clause declares the names and types of the columns that JSON_TABLE returns, as well as how the values of the columns are produced.

Column Definitions

The following types of columns are supported:

Path Columns

Locates the JSON node pointed to by path_str and returns its value. The path_str is evaluated using the current row source node as the context node.

The on_empty and on_error clauses specify the actions to be performed when the value was not found or there was an error condition. See the ON EMPTY and ON ERROR clauses section for details.

ORDINALITY Columns

Counts the rows, starting from 1.

Example:

EXISTS PATH Columns

Checks whether the node pointed to by value_path exists. The value_path is evaluated using the current row source node as the context node.

NESTED PATHs

NESTED PATH converts nested JSON structures into multiple rows.

It finds the sequence of JSON nodes pointed to by path and uses it to produce rows. For each found node, a row is generated with column values as specified by the NESTED PATH's COLUMNS clause. If path finds no nodes, only one row is generated with all columns having NULL values.

For example, consider a JSON document that contains an array of items, and each item, in turn, is expected to have an array of its available sizes:

NESTED PATH allows one to produce a separate row for each size each item has:

NESTED PATH clauses can be nested within one another. They can also be located next to each other. In that case, the nested path clauses will produce records one at a time. The ones that are not producing records will have all columns set to NULL.

Example:

ON EMPTY and ON ERROR Clauses

The ON EMPTY clause specifies what will be done when the element specified by the search path is missing in the JSON document.

When ON EMPTY clause is not present, NULL ON EMPTY is implied.

The ON ERROR clause specifies what should be done if a JSON structure error occurs when trying to extract the value pointed to by the path expression. A JSON structure error here occurs only when one attempts to convert a JSON non-scalar (array or object) into a scalar value. When the ON ERROR clause is not present, NULL ON ERROR is implied.

Note: A datatype conversion error (e.g. attempt to store a non-integer value into an field, or a column being truncated) is not considered a JSON error and so will not trigger the ON ERROR behavior. It will produce warnings, in the same way as would.

Replication

In the current code, evaluation of JSON_TABLE is deterministic, that is, for a given input string JSON_TABLE will always produce the same set of rows in the same order. However, one can think of JSON documents that one can consider identical which will produce different output. In order to be future-proof and withstand changes like

• sorting JSON object members by name (like MySQL does);

• changing the way duplicate object members are handled the function is marked as .

Extracting a Subdocument into a Column

See Also

• (video)

_{This page is licensed: CC BY-SA / Gnu FDL}