JSON

JSON Type

The JSON data type provides the capability to store native JSON in a single column and apply JSON-specific functions to it.


Create Table

SQL 
1

CREATE TABLE example.js ( j JSON )
Python 
1
2
3
4
5
6

table_name = 'example.js'
columns = [
    ['j', 'string', 'json']
]

gpudb.GPUdbTable(columns, table_name, db = kinetica)

Insert Data

SQL 
1
2
3
4

INSERT INTO example.js
VALUES
    ( '[1, 2, "C"]' ),
    ( '{"A": "B", "C": {"D": 5, "F": 7}}' )
Python (as strings) 
1
2
3
4
5
6
7
8

data = [
    [ '[1, 2, "C"]' ],
    [ '{"A": "B", "C": {"D": 5, "F": 7}}' ]
]

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica)

t.insert_records(data)
Python (as native types) 
1
2
3
4
5
6
7
8

data = [
    [ '[1, 2, "C"]' ],
    [ {"A": "B", "C": {"D": 5, "F": 7}} ]
]

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica)

t.insert_records(data)

Retrieve Data

To render a JSON value as a single-line string value:

SQL 
1
2

SELECT j
FROM example.js
Python 
1
2
3
4
5
6

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica)

records = t.get_records_by_column(["j"])

for record in records['j']:
    print(f'{record}')
Python (w/ native type conversions) 
1
2
3
4
5
6

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica, convert_special_types_on_retrieval = True)

records = t.get_records_by_column(["j"], encoding = 'json')

for record in records['j']:
    print(f'{record}')

JSON array and JSON map output:

+-------------------------------+
| j                             |
+-------------------------------+
| [1,2,"C"]                     |
| {"A":"B","C":{"D":5,"F":7}}   |
+-------------------------------+

To render a JSON value as a formatted structure:

SQL 
1
2

SELECT JSON_PRETTY(j) AS j
FROM example.js
Python (via SQL) 
1
2
3
4
5
6

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica)

records = t.get_records_by_column(["JSON_PRETTY(j) AS j"])

for record in records['j']:
    print(f'{record}')
Python (via native library) 
1
2
3
4
5
6

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica)

records = t.get_records_by_column(["j"])

for record in records['j']:
    print(json.dumps(json.loads(record), indent=4))

JSON array and JSON map output:

+---------------------+
| j                   |
+---------------------+
| [
    1, 
    2, 
    "C"
] |
| {
    "A": "B", 
    "C": {
        "D": 5, 
        "F": 7
    }
} |
+---------------------+

JSON Operators

JSON operators can be used to determine whether a given value meets (or does NOT meet) the criteria for being of a certain JSON type category.

Note that the following syntax differs between SQL and the native APIs (like Python) and that the category is specified differently between the two.

SQL 
1

<value> IS [NOT] JSON [<category name>]
Native API 
1
2
3

IS_JSON(<value> [,<category code>])

NOT(IS_JSON(<value> [,<category code>]))

The given category code (APIs) or category name (SQL) will check the given value against the following JSON types:

Category CodeCategory NameJSON Types
0N/AJSON array or object (API default)
1VALUEAny JSON type (SQL default)
2ARRAYJSON array
3OBJECTJSON object (map)
4SCALAR
  • BOOLEAN
  • LONG
  • DOUBLE
  • STRING

Example

To pretty-print only JSON objects (maps) in a given JSON column:

SQL 
1
2
3

SELECT JSON_PRETTY(j)
FROM example.js
WHERE j IS JSON OBJECT
Python 
1
2
3
4
5
6

t = gpudb.GPUdbTable(name = 'example.js', db = kinetica)

records = t.get_records_by_column(
    ["JSON_PRETTY(j) AS j"],
    options = {"expression": "IS_JSON(j, 3)"}
)

JSON Functions

Scalar Functions

These functions can be applied to individual json column values, as well as array & string columns. However, for some functions to work as expected, the array or string value may need to be cast to JSON first.

FunctionDescription
JSON(expr)

Casts the given expr to JSON data type

Examples:

Function CallResult
JSON('[1, 2, "C"]')[1,2,"C"]
JSON('{"A": "B", "C": {"D": 5, "F": 7}}'){"A":"B","C":{"D":5,"F":7}}
JSON(null)null

JSON_ARRAY

(value[,...] <null_hand>)

Creates a JSON array out of the given list of values.

An optional <null_hand> parameter can be added to the end of the list to specify handling of null values, though, at present, this only exists for compatibility with other databases' syntax--the option will be ignored and all null values will be added to the array as null. Accepted null handling options:

  • NULL ON NULL

Note

This function is only available in SQL or in the native API via /execute/sql.

Examples:

Function CallResult
JSON_ARRAY(1, 2, 'C')[1,2,"C"]
JSON_ARRAY(JSON('{"A": "B"}'), JSON('{"C": 4}'))[{"A":"B"},{"C":4}]
JSON_ARRAY(1, null, 'C' NULL ON NULL)[1,null,"C"]
JSON_ARRAY_APPEND(arr, value)

Appends the given value to the given array arr of type string, array, or JSON

Examples:

Function CallResult
JSON_ARRAY_APPEND('[1, 2]', 'C')[1,2,"C"]
JSON_ARRAY_APPEND('{"A": "B"}', JSON('{"C": 4}'))[{"A":"B"},{"C":4}]

JSON_ARRAY_CONTAINS

(json, path, val)

Returns whether the given JSON json contains the primitive value val at the JSON query path, path

Examples:

Function CallResult
JSON_ARRAY_CONTAINS('[1, 2]', '$', 2)1
JSON_ARRAY_CONTAINS('[1, 2]', '$', 'C')0
JSON_ARRAY_CONTAINS('{"A": [2, 3]}', '$.A', 2)1
JSON_ARRAY_CONTAINS('{"A": {"B": [3, 4]}}', '$["A"]["B"]', 4)1
JSON_CARDINALITY(json)Alias for JSON_LENGTH
JSON_EXISTS(json, path)

Returns whether the given JSON json contains the JSON query path, path

Examples:

Function CallResult
JSON_EXISTS('[1, 2]', '$')true
JSON_EXISTS('[1, 2]', '$.A')false
JSON_EXISTS('{"A": [2, 3]}', '$.A')true
JSON_EXISTS('{"A": {"B": 3, "D": 5}}', '$["A"]["D"]')true
JSON_KEYS(json, path)

Returns the set of keys of the JSON object at the given path in json; null, if the path doesn't exist or contains a non-object (non-map) value

Examples:

Function CallResult
JSON_KEYS('{}', '$')[]
JSON_KEYS('[1, 2]', '$.A')null
JSON_KEYS('{"A": [2, 3]}', '$')["A"]
JSON_KEYS('{"A": {"B": 3, "D": 5}}', '$["A"]')["B","D"]
JSON_KEYS('{"A": {"B": 3, "D": 5}}', '$["A"]["D"]')null
JSON_KEYS('{"A": {"B": 3, "D": {}}}', '$["A"]["D"]')[]
JSON_LENGTH(json[, path])

Returns the number of items in the top-level object of the given json, or at the level of the optional JSON path

Examples:

Function CallResult
JSON_LENGTH('{}')0
JSON_LENGTH('[]')0
JSON_LENGTH('[1, 2]')2
JSON_LENGTH('{"A": [2, 3]}')1
JSON_LENGTH('{"A": {"B": 3}, "D": "E"}')2
JSON_LENGTH('{"A": [2, 3]}', '$.A')2
JSON_LENGTH('{"A": {"B": 3}, "D": "E"}', '$.A')1
JSON_MAKE_ARRAY(value)

Creates a JSON array out of the given value

Examples:

Function CallResult
JSON_MAKE_ARRAY(1)[1]
JSON_MAKE_ARRAY(JSON_ARRAY(1,2,3))[[1,2,3]]
JSON_MAKE_ARRAY(JSON('{"A": "B"}'))[{"A":"B"}]

JSON_OBJECT

(value[,...] <null_hand>)

Creates a JSON map out of the given value set

An optional <null_hand> parameter can be added to the end of the list to specify handling of null values, though, at present, this only exists for compatibility with other databases' syntax--the option will be ignored and all null values will be added to the array as null. Accepted null handling options:

  • NULL ON NULL

Note

This function is only available in SQL or in the native API via /execute/sql.

Examples:

Function CallResult
JSON_OBJECT(){}
JSON_OBJECT('A': 2){"A":2}
JSON_OBJECT('A' VALUE 2){"A":2}
JSON_OBJECT(KEY 'A' VALUE 2){"A":2}
JSON_OBJECT('A': 'B', 'C': null NULL ON NULL){"A":"B","C":null}
JSON_PRETTY(json)

Returns a formatted version of the given json object

Examples:

Function CallJSON_PRETTY('[1,2,"C"]')
Return
[
    1,
    2,
    "C"
]
Function CallJSON_PRETTY('{"A": "B", "C": {"D": 5, "F": 7}}')
Return
{
    "A": "B",
    "C": {
        "D": 5,
        "F": 7
    }
]

JSON_QUERY

(json, path [<return>])

Returns the JSON object at the given path in json; null, if the path doesn't exist or contains a primitive (non-object) value

An optional return clause can be added to the end of the path to specify the return type of the object found, as well as the error handling for the object lookup. The format of the clause is:

[<wrapper> WRAPPER] [<empty_hand> ON EMPTY] [<error hand> ON ERROR]

Note

Return clause is only available in SQL or in the native API via /execute/sql.

ParametersDescription
<wrapper>

Whether the returned object(s) should be wrapped in an array before being returned

Valid wrapping schemes include:

  • WITHOUT [ARRAY] - no wrapper is applied (default)
  • WITH [UNCONDITIONAL] [ARRAY] - return the object(s) in an array
  • WITH CONDITIONAL [ARRAY] - if returning a single object, don't wrap it; if returning multiple objects, wrap them in an array
<empty hand>

The scheme to use for handling empty returned objects

Valid schemes include:

  • NULL - return null (default)
  • ERROR - throw an error
  • EMPTY ARRAY - return an empty array
  • EMPTY OBJECT - return an empty object
<error hand>

The scheme to use for handling errors in looking up the object

Valid schemes include:

  • NULL - return null (default)
  • ERROR - throw an error
  • EMPTY ARRAY - return an empty array
  • EMPTY OBJECT - return an empty object

Examples:

Function CallResult
JSON_QUERY('[1, 2]', '$')[1,2]
JSON_QUERY('[1, 2]', '$.A')null
JSON_QUERY('{"A": [2, 3]}', '$.A')[2,3]
JSON_QUERY('{"A": {"B": 3, "D": 5}}', '$["A"]'){"B":3,"D":5}
JSON_QUERY('{"A": {"B": 3, "D": 5}}', '$["A"]["D"]')null
JSON_QUERY('{"A": {"B": 3, "D": [5, 6]}}', '$["A"]["D"]')[5,6]
JSON_QUERY('{"A": []}', '$.B'  EMPTY ARRAY ON EMPTY)[]
JSON_QUERY('{"A": 0}', '$.A'  EMPTY OBJECT ON ERROR){}
JSON_QUERY('{"A": 2, "C": 4}', '$.*')null
JSON_QUERY('{"A": 2, "C": 4}', '$.*' WITH WRAPPER)[2,4]
JSON_REPLACE(json, path, repl)

Returns the given json with the value at the JSON path path replaced with the replacement value repl

Examples:

Function CallResult
JSON_REPLACE('[1, 2]', '$', '[3, 4]')[3,4]
JSON_REPLACE('[1, 2]', '$.A', '"B"')[1,2]
JSON_REPLACE('{"A": [2, 3]}', '$.A', '{"D": 5}'){"A":{"D":5}}
JSON_REPLACE('{"A": {"B": 3}}', '$["A"]', '"D"'){"A":"D"}
JSON_REPLACE('{"A": {"B": [3, 4]}}', '$["A"]["B"]', 5){"A":{"B":5}}
JSON_TYPE(expr)

Returns the JSON type of the given expr

Expression TypeJSON Type
BOOLEANBOOLEAN
  • TINYINT
  • SMALLINT
  • INTEGER
  • BIGINT
  • UNSIGNED BIGINT
LONG
  • REAL
  • DOUBLE
DOUBLE
  • JSON
  • VARCHAR
  • BOOLEAN - if true or false
  • LONG - if an integer value
  • DOUBLE - if a floating-point value
  • STRING - if a double-quoted literal
  • OBJECT - if a JSON map
  • ARRAY - if a JSON array
ARRAY (any)ARRAY

JSON_VALUE

(json, path [return])

Returns the JSON value at the given path in json; null, if the path doesn't exist or contains an object (non-primitive) value

An optional return clause can be added to the end of the path to specify the return type of the value found, as well as the error handling for the value lookup. The format of the clause is:

[RETURNING <type>] [<empty_hand> ON EMPTY] [<error hand> ON ERROR]

Note

Return clause is only available in SQL or in the native API via /execute/sql.

ParametersDescription
<type>

The JSON type of the value to return; if the value is not already of that type, it will be cast to that type, if possible

Valid types include:

  • BOOLEAN
  • LONG
  • DOUBLE
  • STRING
<empty hand>

The scheme to use for handling empty returned values

Valid schemes include:

  • NULL - return null (default)
  • ERROR - throw an error
  • DEFAULT <expr> - return the given expression expr
<error hand>

The scheme to use for handling errors in looking up the value

Valid schemes include:

  • NULL - return null (default)
  • ERROR - throw an error
  • DEFAULT <expr> - return the given expression expr

Examples:

Function CallResult
JSON_VALUE('1', '$')1
JSON_VALUE('[1, 2]', '$.A')null
JSON_VALUE('["A", "B"]', '$.1')B
JSON_VALUE('{"A": [2, 3]}', '$["A"]["1"]')3
JSON_VALUE('{"A": {"B": 3, "D": 5}}', '$["A"]["D"]')5
JSON_VALUE('{"A": {"B": 3, "D": [5, 6]}}', '$.A.D.1')6
JSON_VALUE('1', '$' RETURNING DOUBLE)1.0
JSON_VALUE('{"A": []}', '$.B' DEFAULT '<null>' ON EMPTY)<null>
JSON_VALUE('{"A": []}', '$.A' ERROR ON ERROR)error returned

Aggregation Functions

These functions can be used across multiple primitive column values within grouping queries.

Note

These functions are only available in SQL or in the native API via /execute/sql.

FunctionDescription
JSON_ARRAYAGG(expr)

Combines column values within a group into a single JSON array of those values

See examples.

JSON_ARRAYAGG_DISTINCT(expr)

Combines column values within a group into a single JSON array of those values, removing any duplicates

See examples.


Examples

JSON_ARRAYAGG
Input 
+------+---------+---------------------------------+
|   id | name    | j                               |
+------+---------+---------------------------------+
|    1 | Joe     | {}                              |
|    2 | Jane    | {"A":"B"}                       |
|    3 | Jeff    | [1,2,3]                         |
|    4 | Joyce   | {"A":2,"C":{"D":5,"F":[7,8]}}   |
+------+---------+---------------------------------+

JSON Array Aggregation Query 
1
2
3
4
5

SELECT
    JSON_ARRAYAGG(id) AS id_list,
    JSON_ARRAYAGG(name) AS name_list,
    JSON_ARRAYAGG(j) AS json_list
FROM example.js_arr_agg

Output 
+-------------+---------------------------------+--------------------------------------------------------+
| id_list     | name_list                       | json_list                                              |
+-------------+---------------------------------+--------------------------------------------------------+
| [4,1,2,3]   | ["Joyce","Joe","Jane","Jeff"]   | [{"A":2,"C":{"D":5,"F":[7,8]}},{},{"A":"B"},[1,2,3]]   |
+-------------+---------------------------------+--------------------------------------------------------+
JSON_ARRAYAGG_DISTINCT
Input 
+------+---------+---------------------------------+
|   id | name    | j                               |
+------+---------+---------------------------------+
|    1 | Joe     | {}                              |
|    1 | Joyce   | [1,2,3]                         |
|    2 | Jane    | {"A":"B"}                       |
|    2 | Jeff    | {"A":2,"C":{"D":5,"F":[7,8]}}   |
|    3 | Jane    | {}                              |
|    3 | Jeff    | [1,2,3]                         |
|    4 | Joe     | {"A":"B"}                       |
|    4 | Joyce   | {"A":2,"C":{"D":5,"F":[7,8]}}   |
+------+---------+---------------------------------+

Distinct JSON Array Aggregation Query 
1
2
3
4
5

SELECT
    JSON_ARRAYAGG_DISTINCT(id) AS id_list,
    JSON_ARRAYAGG_DISTINCT(name) AS name_list,
    JSON_ARRAYAGG_DISTINCT(j) AS json_list
FROM example.js_arr_agg

Output 
+-------------+---------------------------------+--------------------------------------------------------+
| id_list     | name_list                       | json_list                                              |
+-------------+---------------------------------+--------------------------------------------------------+
| [1,2,3,4]   | ["Jane","Jeff","Joe","Joyce"]   | [[1,2,3],{},{"A":"B"},{"A":2,"C":{"D":5,"F":[7,8]}}]   |
+-------------+---------------------------------+--------------------------------------------------------+

UNNEST_JSON_ARRAY

The UNNEST_JSON_ARRAY function transposes array values into columnar data.

The two forms of the UNNEST_JSON_ARRAY function follow, one for cases where only one column needs to be unnested and one for cases where more than one column needs to be unnested. See examples.

UNNEST_JSON_ARRAY for 1 Column Syntax 
1
2

SELECT <column alias> [, ordinality]
FROM UNNEST_JSON_ARRAY(SELECT <column name> FROM <table name>) [WITH ORDINALITY] [<column alias>]
ParametersDescription
<column alias>Column alias to use for the resulting unnested data column; if no alias specified, use * to select the column
<table name>Name of the table containing the JSON array column
<column name>Name of the JSON array column to unnest
WITH ORDINALITYAlso return the 0-based index in the source JSON arrays for each transposed array value; the index value will be returned as the column alias ordinality

UNNEST_JSON_ARRAY for Multiple Columns Syntax 
1
2

SELECT <column alias list> [, <ordinality alias>]
FROM <table name> <table alias>, UNNEST_JSON_ARRAY(<column list>) [WITH ORDINALITY] <table/columns aliases>
ParametersDescription
<column alias list>Column aliases given to the resulting unnested data columns, specified in <table/columns aliases>
<ordinality alias>Column alias given to the resulting ordinality column in <table/columns aliases>, if WITH ORDINALITY is specified
<table name>Name of the table containing the JSON array column(s)
<table alias>Alias to use for the table containing the JSON array column(s)
<column list>Comma-separated list of JSON array columns to unnest
<table/columns aliases>

Alias to use for the virtual table and columns being used to return the unnested data (including any ordinality alias, if specified), using the form:

<table alias>(<column alias list> [,<ordinality alias>])

For example, to return two columns of unnested JSON array data as uai & uav, respectively, and the corresponding ordinality index as index, all within the virtual table unn use:

unn(uai, uav, index)
WITH ORDINALITYAlso return the 0-based index in the source JSON arrays for each transposed array value; the index value will be returned in an alias specified in <table/columns aliases>

Examples

The following examples use this data set as a baseline:

+-----------+-----------------+
| ai        | av              |
+-----------+-----------------+
| [1,2,3]   | ["A","B","C"]   |
| [4,5,6]   | ["D","E","F"]   |
| [7,8,9]   | ["D","E","F"]   |
| [1,2,3]   | ["G","H","I"]   |
+-----------+-----------------+

To turn the string arrays in the av JSON column in the example.js_unn_arr table into a column of values aliased as uav:

1
2

SELECT uav
FROM UNNEST_JSON_ARRAY(SELECT av FROM example.js_unn_arr) uav

+-------+
| uav   |
+-------+
| A     |
| B     |
| C     |
| D     |
| D     |
| E     |
| E     |
| F     |
| F     |
| G     |
| H     |
| I     |
+-------+

To turn the integer arrays in the ai JSON column in the example.js_unn_arr table into a column of values aliased as uai and the corresponding index in the original json for each as a column of indices named ordinality:

1
2

SELECT uai, ordinality
FROM UNNEST_JSON_ARRAY(SELECT ai FROM example.js_unn_arr) WITH ORDINALITY uai

+-------+--------------+
|   uai |   ordinality |
+-------+--------------+
|     1 |            0 |
|     1 |            0 |
|     2 |            1 |
|     2 |            1 |
|     3 |            2 |
|     3 |            2 |
|     4 |            0 |
|     5 |            1 |
|     6 |            2 |
|     7 |            0 |
|     8 |            1 |
|     9 |            2 |
+-------+--------------+

To turn the integer & string arrays in the ai & av JSON columns in the example.js_unn_arr table into two columns of values aliased as uai & uav in a virtual table with alias t, respectively:

1
2

SELECT uai, uav
FROM example.js_unn_arr jua, UNNEST_JSON_ARRAY(jua.ai, jua.av) t(uai, uav)

+-------+-------+
|   uai | uav   |
+-------+-------+
|     1 | A     |
|     1 | G     |
|     2 | B     |
|     2 | H     |
|     3 | C     |
|     3 | I     |
|     4 | D     |
|     5 | E     |
|     6 | F     |
|     7 | D     |
|     8 | E     |
|     9 | F     |
+-------+-------+

To turn the integer & string arrays in the ai & av JSON columns in the example.js_unn_arr table into two columns of values aliased as uai & uav in a virtual table with alias t, respectively, and adding the corresponding index in the original array for each as a column of indices aliased as index:

1
2

SELECT uai, uav, index
FROM example.js_unn_arr jua, UNNEST_JSON_ARRAY(jua.ai, jua.av) WITH ORDINALITY t(uai, uav, index)

+-------+-------+---------+
|   uai | uav   |   index |
+-------+-------+---------+
|     1 | A     |       0 |
|     1 | G     |       0 |
|     2 | B     |       1 |
|     2 | H     |       1 |
|     3 | C     |       2 |
|     3 | I     |       2 |
|     4 | D     |       0 |
|     5 | E     |       1 |
|     6 | F     |       2 |
|     7 | D     |       0 |
|     8 | E     |       1 |
|     9 | F     |       2 |
+-------+-------+---------+

To turn the integer arrays in the ai JSON column in the example.js_unn_arr table into a column of values aliased as uai, removing any duplicates across all arrays:

1
2

SELECT DISTINCT *
FROM UNNEST_JSON_ARRAY(SELECT ai FROM example.js_unn_arr)

+-------+
|   uai |
+-------+
|     1 |
|     2 |
|     3 |
|     4 |
|     5 |
|     6 |
|     7 |
|     8 |
|     9 |
+-------+

UNNEST_JSON_OBJECT

The UNNEST_JSON_OBJECT function transposes JSON objects (maps) into columnar data. Each source object is split out into two columns: a string column containing the top-level object keys, and a JSON column containing the value objects for those keys.

The two forms of the UNNEST_JSON_OBJECT function follow, one for cases where only one column needs to be unnested and one for cases where more than one column needs to be unnested. See examples.

UNNEST_JSON_OBJECT for 1 Column Syntax 
1
2

SELECT <column alias list> [, ordinality]
FROM UNNEST_JSON_OBJECT(SELECT <column name> FROM <table name>) [WITH ORDINALITY] [<table/columns aliases>]
ParametersDescription
<column alias list>Column aliases given to the resulting unnested key & value data columns specified in <table/columns aliases>; if no aliases are specified, use * to select the columns
<table name>Name of the table containing the JSON object (map) column
<column name>Name of the JSON object (map) column to unnest
<table/columns aliases>

Alias to use for the virtual table and key/value columns being used to return the unnested data (including any ordinality alias, if specified), using the form:

<table alias>(<key alias>, <value alias> [,<ordinality alias>])

For example, to return a column as unnested JSON map data, referenced as the key/value pairs j_k & j_v, respectively, with the corresponding ordinality index as index, all within the virtual table unn, use:

unn(j_k, j_v, index)

If no aliases are given, the two key/value columns will be returned as KEY_0 & VALUE_0, respectively.

WITH ORDINALITYAlso return the 0-based index in the source maps for each transposed map value; the index value will be returned as the column alias ordinality unless overridden in <table/columns aliases>

UNNEST_JSON_OBJECT for Multiple Columns Syntax 
1
2

SELECT <column alias list> [, <ordinality alias>]
FROM <table name> <table alias>, UNNEST_JSON_OBJECT(<column list>) [WITH ORDINALITY] <table/columns aliases>
ParametersDescription
<column alias list>Column aliases given to the resulting unnested key & value data columns specified in <table/columns aliases>
<ordinality alias>Column alias given to the resulting ordinality column in <table/columns aliases>, if WITH ORDINALITY is specified
<table name>Name of the table containing the JSON object (map) column(s)
<table alias>Alias to use for the table containing the JSON object (map) column(s)
<column list>Comma-separated list of JSON object (map) columns to unnest
<table/columns aliases>

Alias to use for the virtual table and columns being used to return the unnested data (including any ordinality alias, if specified), using the form:

<table alias>(<key alias>, <value alias>[,...] [,<ordinality alias>])

For example, to return two JSON map data columns as unnested, referenced as the key/value pairs j1_k & j1_v and j2_k & j2_v, respectively, with the corresponding ordinality index as index, all within the virtual table unn, use:

unn(j1_k, j1_v, j2_k, j2_v, index)
WITH ORDINALITYAlso return the 0-based index in the source maps for each transposed map value; the index value will be returned in an alias specified in <table/columns aliases>

Examples

The following examples use this data set as a baseline:

+-----------------------------+------------------------------------------+
| j1                          | j2                                       |
+-----------------------------+------------------------------------------+
| {"A":"B","C":4,"E":5}       | {"A":{"B":"C"},"D":5,"F":[7,8,9]}        |
| {"A":2,"C":"D","E":"F"}     | {"A":[2,3],"D":"E","F":{"G":"H"}}        |
| {"G":"H","I":10,"K":"12"}   | {"I":{"J":"K"},"L":[13,14,15],"P":"Q"}   |
| {"A":"B","C":4,"E":5}       | {"A":{"B":"C"},"D":5,"F":[7,8,9]}        |
+-----------------------------+------------------------------------------+

To turn the JSON maps in the j1 JSON column in the example.js_unn_obj table into key/value pair columns of values aliased as j1_k & j1_v, respectively:

1
2

SELECT j1_k, j1_v
FROM UNNEST_JSON_OBJECT(SELECT j1 FROM example.js_unn_obj) t(j1_k, j1_v)

+--------+--------+
| j1_k   | j1_v   |
+--------+--------+
| A      | "B"    |
| A      | "B"    |
| A      | 2      |
| C      | "D"    |
| C      | 4      |
| C      | 4      |
| E      | "F"    |
| E      | 5      |
| E      | 5      |
| G      | "H"    |
| I      | 10     |
| K      | "12"   |
+--------+--------+

To turn the JSON maps in the j2 JSON column in the example.js_unn_obj table into key/value pair columns of values aliased as j2_k & j2_v, and the corresponding index in the original map for each as a column of indices aliased as index:

1
2

SELECT j2_k, j2_v, index
FROM UNNEST_JSON_OBJECT(SELECT j2 FROM example.js_unn_obj) WITH ORDINALITY t(j2_k, j2_v, index)

+--------+--------------+---------+
| j2_k   | j2_v         |   index |
+--------+--------------+---------+
| A      | [2,3]        |       0 |
| A      | {"B":"C"}    |       0 |
| A      | {"B":"C"}    |       0 |
| D      | "E"          |       1 |
| D      | 5            |       1 |
| D      | 5            |       1 |
| F      | [7,8,9]      |       2 |
| F      | [7,8,9]      |       2 |
| F      | {"G":"H"}    |       2 |
| I      | {"J":"K"}    |       0 |
| L      | [13,14,15]   |       1 |
| P      | "Q"          |       2 |
+--------+--------------+---------+

To turn the JSON maps in the j1 & j2 JSON columns in the example.js_unn_obj table into two sets of key/value pair columns of values aliased as j1_k & j1_v and j2_k & j2_v, respectively, in a virtual table with alias t:

1
2

SELECT j1_k, j1_v, j2_k, j2_v
FROM example.js_unn_obj juo, UNNEST_JSON_OBJECT(juo.j1, juo.j2) t(j1_k, j1_v, j2_k, j2_v)

+--------+--------+--------+--------------+
| j1_k   | j1_v   | j2_k   | j2_v         |
+--------+--------+--------+--------------+
| A      | "B"    | A      | {"B":"C"}    |
| A      | "B"    | A      | {"B":"C"}    |
| A      | 2      | A      | [2,3]        |
| C      | "D"    | D      | "E"          |
| C      | 4      | D      | 5            |
| C      | 4      | D      | 5            |
| E      | "F"    | F      | {"G":"H"}    |
| E      | 5      | F      | [7,8,9]      |
| E      | 5      | F      | [7,8,9]      |
| G      | "H"    | I      | {"J":"K"}    |
| I      | 10     | L      | [13,14,15]   |
| K      | "12"   | P      | "Q"          |
+--------+--------+--------+--------------+

To turn the JSON maps in the j1 & j2 JSON columns in the example.js_unn_obj table into two sets of key/value pair columns of values aliased as j1_k & j1_v and j2_k & j2_v, respectively, in a virtual table with alias t, and adding the corresponding index in the original map for each as a column of indices aliased as index:

1
2

SELECT j1_k, j1_v, j2_k, j2_v, index
FROM example.js_unn_obj juo, UNNEST_JSON_OBJECT(juo.j1, juo.j2) WITH ORDINALITY t(j1_k, j1_v, j2_k, j2_v, index)

+--------+--------+--------+--------------+---------+
| j1_k   | j1_v   | j2_k   | j2_v         |   index |
+--------+--------+--------+--------------+---------+
| A      | "B"    | A      | {"B":"C"}    |       0 |
| A      | "B"    | A      | {"B":"C"}    |       0 |
| A      | 2      | A      | [2,3]        |       0 |
| C      | "D"    | D      | "E"          |       1 |
| C      | 4      | D      | 5            |       1 |
| C      | 4      | D      | 5            |       1 |
| E      | "F"    | F      | {"G":"H"}    |       2 |
| E      | 5      | F      | [7,8,9]      |       2 |
| E      | 5      | F      | [7,8,9]      |       2 |
| G      | "H"    | I      | {"J":"K"}    |       0 |
| I      | 10     | L      | [13,14,15]   |       1 |
| K      | "12"   | P      | "Q"          |       2 |
+--------+--------+--------+--------------+---------+