.JSON File Format – Syntax

Grammar

The whole grammar, in the ABNF of RFC 8259, with the strings and numbers on their own pages:

JSON-text = ws value ws

value  = false / null / true / object / array / number / string

object = begin-object [ member *( value-separator member ) ] end-object
member = string name-separator value

array  = begin-array [ value *( value-separator value ) ] end-array

false  = %x66.61.6c.73.65   ; false
null   = %x6e.75.6c.6c      ; null
true   = %x74.72.75.65      ; true

Any value may stand at the top level. That has been so since RFC 7159; RFC 4627 allowed only an object or an array, as the versions page explains.

White space

Four characters count as white space, and they may appear before or after any of the structural characters, and before and after the value at the top level:

No other character is white space: not a form feed, a vertical tab, a no-break space or a byte order mark.

Structural characters

A comma only ever separates: there is no comma after the last member or element. JSON has no comments.

Literal names

The three literal names are true, false and null. They must be lower case, and there are no others: True, NULL, undefined, NaN and Infinity are not JSON.

Objects

An object is zero or more members between braces. A member is a string, called its name, then a colon, then a value. The names need not look like identifiers: any string will do, including the empty string.

{ "x": 1, "y": [true, null], "": "an empty name is allowed" }

Names should be unique

RFC 8259 says the names within an object should be unique. It does not forbid repeats, so {"a":1,"a":2} is grammatically valid, but it warns that the behaviour of software receiving such an object is unpredictable: some report only the last pair, some report an error, and others report all of them. An object whose names are all unique is the only kind every implementation agrees on. Whether two names are the same is decided after escapes are decoded, as described under comparing strings.

Member order

RFC 8259 calls an object an unordered collection. Parsers differ on whether they show the order of the members to the program using them, so software should not depend on it.

Arrays

An array is zero or more values between square brackets, separated by commas. The values need not be of the same kind, so [1, "two", [3], {"four": 4}, null] is one valid array. Unlike the members of an object, the elements of an array are ordered.