jq is a very high-level lexically scoped functional programming language in which every JavaScript Object Notation (JSON) value is a constant. jq supports backtracking and managing indefinitely long streams of JSON data. It is related to two programming languages Icon and Haskell. The language supports a namespace-based module system and has some support for closures. functions and functional expressions can be used as parameters of other functions.
The original implementation of jq was in Haskell[3] before being quickly ported to C.
jq was created by Stephen Dolan, and released in October 2012.[4][5] It was described as being "like sed for JSON data".[6] Support for regular expressions was added in jq version 1.5.
A "wrapper" program for jq named yq adds support for YAML, XML and TOML. It was first released in 2017.[7]
The Go implementation, gojq, was initially released in 2019.[8] gojq notably extends jq to include support for YAML.
The Rust implementation, jaq, has as its project goals a faster and more correct implementation of jq, while preserving compatibility with jq in most cases. Explicitly excluded from the project goals as of March 2024 are certain advanced features of jq such as modules, SQL-style operators, and a streaming parser for very large JSON documents.[9]
The jq implementation, jqjq, was initially released in 2022. jqjq notably can run itself, has a REPL and supports eval.
jq is typically used at the command line and can be used with other command-line utilities, such as curl. Here is an example showing how the output of a curl command can be piped to a jq filter to determine the category names associated with this Wikipedia page:
curl
$ curl -s 'https://en.wikipedia.org/w/api.php?action=parse&page=jq_(programming_language)&format=json' | jq '.parse.categories[]."*"'
The output produced by this pipeline consists of a stream of JSON strings, the first few of which are:
"Articles_with_short_description" "Short_description_matches_Wikidata" "Dynamically_typed_programming_languages" "Functional_languages" "Programming_languages" "Programming_languages_created_in_2012" "Query_languages" "2012_software"
The curl command above uses the MediaWiki API for this page to produce a JSON response. The pipe | allows the output of curl to be accessed by jq, a standard Unix shell mechanism.[10]
|
The jq filter shown is an abbreviation for the jq pipeline:
.["parse"] | .["categories"] | .[] | .["*"]
This corresponds to the nested JSON structure produced by the call to curl. Notice that the jq pipeline is constructed in the same manner using the | character as the Unix-style pipeline.
Both the C and the Go implementations provide libraries so that jq functions can be embedded in other applications and programming environments.
For example, gojq has been integrated with SQLite so that a jq function is available in SQL statements.[11] This function is marked as "deterministic" and can therefore be used in "CREATE INDEX" commands.[12]
jq
jq by default acts as a "stream editor" for JSON inputs, much like the sed utility can be thought of as a "stream editor" for lines of text. However jq has several other modes of operation:
The streaming parser is very useful when one of more of the JSON inputs is too large to fit in memory, since its memory needs are usually quite small. For example, for an arbitrarily large array of JSON objects, the peak memory need is little more than needed to handle the largest top-level object.
These modes of operation can, within certain limitations, be combined.
Every JSON value is also a value in jq, which accordingly has the data types shown in the table below.[13] The gojq and jaq implementations distinguish between integers and non-integer numbers. The gojq implementation supports unbounded-precision integer arithmetic, as did the original implementation of jq in Haskell.
3
3.2
1e6
nan
infinite
"Hello"
"đ"
true
false
[1, "2", {"mixed": "type"}, [3,4]]
{"one": 1, "two": "2", "three": [3]}
null
null is a value, just like any other JSON scalar; it is not a pointer or a "null-pointer". nan (corresponding to NaN) and infinite (see IEEE 754) are the only two jq scalars that are not also JSON values.
Special syntactic forms exist for function creation, conditionals, stream reduction, and the module system.
Here is an example of defining a named, parameterized filter for formatting an integer in any base from 2 to 36 inclusive. The implementation illustrates tacit (or point-free) programming:
# Use gojq for infinite precision integer arithmetic def tobase($b): def digit: "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"[.:.+1]; def mod: . % $b; def div: ((. - mod) / $b); def digits: recurse( select(. >= $b) | div) | mod ; select(2 <= $b and $b <= 36) | [digits | digit] | reverse | add;
The next example demonstrates the use of generators in the classic "SEND MORE MONEY" verbal arithmetic game:
def send_more_money: def choose(m;n;used): ([range(m;n+1)] - used)[]; def num(a;b;c;d): 1000*a + 100*b + 10*c + d; def num(a;b;c;d;e): 10*num(a;b;c;d) + e; first( 1 as $m | 0 as $o | choose(8;9;[]) as $s | choose(2;9;[$s]) as $e | choose(2;9;[$s,$e]) as $n | choose(2;9;[$s,$e,$n]) as $d | choose(2;9;[$s,$e,$n,$d]) as $r | choose(2;9;[$s,$e,$n,$d,$r]) as $y | select(num($s;$e;$n;$d) + num($m;$o;$r;$e) == num($m;$o;$n;$e;$y)) | [$s,$e,$n,$d,$m,$o,$r,$e,$m,$o,$n,$e,$y] );
There is a very close relationship between jq and the parsing expression grammar (PEG) formalism. [14] The relationship stems from the equivalence of the seven basic PEG operations and the jq constructs shown in the following table.
e1 e2
e1 | e2
e1 / e2
e1 // e2
e*
def star(E): (E | star(E)) // . ;
e+
def plus(E): E | (plus(E) // . );
e?
def optional(E): E // .;
&e
def amp(E): . as $in | E | $in;
!e
def neg(E): select( [E] == [] );
gojq is a pure Go implementation. There is also a Rust implementation of a dialect of jq named jaq[9] for which a denotational semantics has been specified.[15]