(see also licenses for dev. deps.)
JSONPath Plus
Analyse, transform, and selectively extract data from JSON
documents (and JavaScript objects).jsonpath-plus
expands on the original specification to add some
additional operators and makes explicit some behaviors the original
did not spell out.Try the browser demo or Runkit (Node).
Please note: This project is not currently being actively maintained. We may accept well-documented PRs or some simple updates, but are not looking to make fixes or add new features ourselves.
Features
- Compliant with the original jsonpath spec
- Convenient additions or elaborations not provided in the original spec:
* `^` for grabbing the **parent** of a matching item
* `~` for grabbing **property names** of matching items (as array)
* **Type selectors** for obtaining:
* Basic JSON types: `@null()`, `@boolean()`, `@number()`, `@string()`, `@array()`, `@object()`
* `@integer()`
* The compound type `@scalar()` (which also accepts `undefined` and
non-finite numbers when querying JavaScript objects as well as all of the basic non-object/non-function types)
* `@other()` usable in conjunction with a user-defined `otherTypeCallback`
* Non-JSON types that can nevertheless be used when querying
non-JSON JavaScript objects (`@undefined()`, `@function()`, `@nonFinite()`)
* `@path`/`@parent`/`@property`/`@parentProperty`/`@root` **shorthand selectors** within filters
* **Escaping**
* `` ` `` for escaping remaining sequence
* `@['...']`/`?@['...']` syntax for escaping special characters within
property names in filters
* Documents `$..` (**getting all parent components**)
- ESM and UMD export formats
- In addition to queried values, can return various meta-information
including paths or pointers to the value, as well as the parent
object and parent property name (to allow for modification).
- Utilities for converting between paths, arrays, and pointers
- Option to prevent evaluations permitted in the original spec or supply
a **sandbox** for evaluated values.
- Option for callback to handle results as they are obtained.
Benchmarking
jsonpath-plus
is consistently performant with both large and small datasets compared to other json querying libraries per json-querying-performance-testing. You can verify these findings by running the project yourself and adding more perf cases.Install
npm install jsonpath-plus
Setup
Node.js
const {JSONPath} = require('jsonpath-plus');
const result = JSONPath({path: '...', json});
Browser
For browser usage you can directly includedist/index-browser-umd.cjs
; no
Browserify magic is necessary:<script src="node_modules/jsonpath-plus/dist/index-browser-umd.cjs"></script>
<script>
const result = JSONPath.JSONPath({path: '...', json: {}});
</script>
ESM (Modern browsers)
You may also use ES6 Module imports (for modern browsers):<script type="module">
import {
JSONPath
} from './node_modules/jsonpath-plus/dist/index-browser-esm.js';
const result = JSONPath({path: '...', json: {}});
</script>
ESM (Bundlers)
Or if you are bundling your JavaScript (e.g., with Rollup), just use, noting thatmainFields
should include browser
for browser builds (for Node, the default, which
checks module
, should be fine):import {JSONPath} from 'jsonpath-plus';
const result = JSONPath({path: '...', json});
Usage
The full signature available is:const result = JSONPath([options,] path, json, callback, otherTypeCallback);
The arguments
path
, json
, callback
, and otherTypeCallback
can alternatively be expressed (along with any other of the
available properties) on options
.Note that
result
will contain all items found (optionally
wrapped into an array) whereas callback
can be used if you
wish to perform some operation as each item is discovered, with
the callback function being executed 0 to N times depending
on the number of independent items to be found in the result.
See the docs below for more on JSONPath
's available arguments.See also the API docs.
Properties
The properties that can be supplied on the options object or evaluate method (as the first argument) include:- path (required) - The JSONPath expression as a (normalized
- json (required) - The JSON object to evaluate (whether of
- autostart (default: true) - If this is supplied as
false
,
evaluate
method manually.- flatten (default: false) - Whether the returned array of results
- resultType (default: "value") - Can be case-insensitive form of
- sandbox (default: {}) - Key-value map of variables to be
- wrap (default: true) - Whether or not to wrap the results
wrap
is set to false
, and no results are found,
undefined
will be returned (as opposed to an empty array when
wrap
is set to true). If wrap
is set to false
and a single
non-array result is found, that result will be the only item returned
(not within an array). An array will still be returned if multiple
results are found, however. To avoid ambiguities (in the case where
it is necessary to distinguish between a result which is a failure
and one which is an empty array), it is recommended to switch the
default to false
.- eval (default: "safe") - Script evaluation method.
safe
: In browser, it will use a minimal scripting engine which doesn't
use eval
or Function
and satisfies Content Security Policy. In NodeJS,
it has no effect and is equivalent to native as scripting is safe there.
native
: uses the native scripting capabilities. i.e. unsafe eval
or
Function
in browser and vm.Script
in nodejs. false
: Disable JavaScript
evaluation expressions and throw exceptions when these expressions are attempted.
callback [ (code, context) => value]
: A custom implementation which is called
with code
and context
as arguments to return the evaluated value.
class
: A class which is created with code
as constructor argument and code
is evaluated by calling runInNewContext
with context
.
``- ignoreEvalErrors (default: false) - Ignore errors encountered during
- parent (default: null) - In the event that a query could be
- parentProperty (default: null) - In the event that a query
parentProperty
of that root node to be returned within results.- callback (default: (none)) - If supplied, a callback will be
resultType
),
the type of the payload (whether it is a normal "value" or a "property"
name), and a full payload object (with all resultType
s).
when @other() is encountered\>) - In the current absence of JSON
Schema support, one can determine types beyond the built-in types by
adding the operator @other()
at the end of one's query. If such a
path is encountered, the otherTypeCallback
will be invoked with the
value of the item, its path, its parent, and its parent's property name,
and it should return a boolean indicating whether the supplied value
belongs to the "other" type or not (or it may handle transformations and
return false).Instance methods
- evaluate(path, json, callback, otherTypeCallback) OR
autostart
property is set to false
. It
can be used for repeated evaluations using the same configuration.
Besides the listed properties, the latter method pattern can
accept any of the other allowed instance properties (except
for autostart
which would have no relevance here).Class properties and methods
- JSONPath.cache - Exposes the cache object for those who wish
- JSONPath.toPathArray(pathAsString) - Accepts a normalized or
['$', 'aProperty', 'anotherProperty']
.- JSONPath.toPathString(pathAsArray) - Accepts a path array and
$['aProperty']['anotherProperty][0]
. The JSONPath terminal
constructions ~
and ^
and type operators like @string()
are
silently stripped.- JSONPath.toPointer(pathAsArray) - Accepts a path array and
/aProperty/anotherProperty/0
(with any ~
and /
internal characters escaped as per the JSON
Pointer spec). The JSONPath terminal constructions ~
and ^
and
type operators like @string()
are silently stripped.Syntax through examples
Given the following JSON, taken from :{
"store": {
"book": [
{
"category": "reference",
"author": "Nigel Rees",
"title": "Sayings of the Century",
"price": 8.95
},
{
"category": "fiction",
"author": "Evelyn Waugh",
"title": "Sword of Honour",
"price": 12.99
},
{
"category": "fiction",
"author": "Herman Melville",
"title": "Moby Dick",
"isbn": "0-553-21311-3",
"price": 8.99
},
{
"category": "fiction",
"author": "J. R. R. Tolkien",
"title": "The Lord of the Rings",
"isbn": "0-395-19395-8",
"price": 22.99
}
],
"bicycle": {
"color": "red",
"price": 19.95
}
}
}
and the following XML representation:
<store>
<book>
<category>reference</category>
<author>Nigel Rees</author>
<title>Sayings of the Century</title>
<price>8.95</price>
</book>
<book>
<category>fiction</category>
<author>Evelyn Waugh</author>
<title>Sword of Honour</title>
<price>12.99</price>
</book>
<book>
<category>fiction</category>
<author>Herman Melville</author>
<title>Moby Dick</title>
<isbn>0-553-21311-3</isbn>
<price>8.99</price>
</book>
<book>
<category>fiction</category>
<author>J. R. R. Tolkien</author>
<title>The Lord of the Rings</title>
<isbn>0-395-19395-8</isbn>
<price>22.99</price>
</book>
<bicycle>
<color>red</color>
<price>19.95</price>
</bicycle>
</store>
Please note that the XPath examples below do not distinguish between retrieving elements and their text content (except where useful for comparisons or to prevent ambiguity). Note: to test the XPath examples (including 2.0 ones), this demo may be helpful (set to
xml
or xml-strict
).| XPath | JSONPath | Result | Notes | | ----------------- | ---------------------- | ------------------------------------- | ----- | /store/book/author | $.store.book\[.author | The authors of all books in the store | Can also be represented without the
$.
as store.book[*].author
(though this is not present in the original spec); note that some character literals ($
and @
) require escaping, however
//author | $..author | All authors |
/store/ | $.store. | All things in store, which are its books (a book array) and a red bicycle (a bicycle object).|
/store//price | $.store..price | The price of everything in the store. |
//book\[3 | $..book\[2 | The third book (book object) |
//book\[last() | $..book\[(@.length-1)$..book\[-1: | The last book in order.| To access a property with a special character, utilize
[(@['...'])]
for the filter (this particular feature is not present in the original spec)
//book\[position()<3| $..book\[0,1$..book\[:2| The first two books | //book/\self::category\|self::author or //book/(category,author) in XPath 2.0 | $..book\[0\category,author| The categories and authors of all books | //book\[isbn | $..book\[?(@.isbn) | Filter all books with an ISBN number | To access a property with a special character, utilize
[?@['...']]
for the filter (this particular feature is not present in the original spec)
//book\[price<10 | $..book\[?(@.price<10) | Filter all books cheaper than 10 |
| //\*\name() = 'price' and . != 8.95 | $..\*\?(@property === 'price' && @ !== 8.95) | Obtain all property values of objects whose property is price and which does not equal 8.95 | With the bare @
allowing filtering objects by property value (not necessarily within arrays), you can add ^
after the expression to get at the object possessing the filtered properties
/ | $ | The root of the JSON object (i.e., the whole object itself) | To get a literal $
(by itself or anywhere in the path), you must use the backtick escape
//\*/\*\|//\*/\*/text() | $.. | All Elements (and text) beneath root in an XML document. All members of a JSON structure beneath the root. |
// | $.. | All Elements in an XML document. All parent components of a JSON structure including root. | This behavior was not directly specified in the original spec
//\price>19/.. | $..\[?(@.price>19)^ | Parent of those specific items with a price greater than 19 (i.e., the store value as the parent of the bicycle and the book array as parent of an individual book) | Parent (caret) not present in the original spec
/store//name() (in XPath 2.0) | $.store.~ | The property names of the store sub-object ("book" and "bicycle"). Useful with wildcard properties. | Property name (tilde) is not present in the original spec
/store/book\[not(. is /store/book\[1\])\] (in XPath 2.0) | $.store.book\[?(@path !== "$\[\'store\'\\'book\'\0") | All books besides that at the path pointing to the first | @path not present in the original spec
//book\[parent::\*/bicycle/color = "red"/category | $..book\[?(@parent.bicycle && @parent.bicycle.color === "red").category | Grabs all categories of books where the parent object of the book has a bicycle child whose color is red (i.e., all the books) | @parent is not present in the original spec
//book/\name() != 'category' | $..book.\?(@property !== "category") | Grabs all children of "book" except for "category" ones | @property is not present in the original spec
//book\[position() != 1 | $..book\[?(@property !== 0) | Grabs all books whose property (which, being that we are reaching inside an array, is the numeric index) is not 0 | @property is not present in the original spec
/store/\*/\*\name(parent::) != 'book' | $.store.\?(@parentProperty !== "book") | Grabs the grandchildren of store whose parent property is not book (i.e., bicycle's children, "color" and "price") | @parentProperty is not present in the original spec
//book\[count(preceding-sibling::\*) != 0/\*/text() | $..book.\?(@parentProperty !== 0) | Get the property values of all book instances whereby the parent property of these values (i.e., the array index holding the book item parent object) is not 0 | @parentProperty is not present in the original spec
//book\[price = /store/book\[3/price | $..book\[?(@.price === @root.store.book\[2.price) | Filter all books whose price equals the price of the third book | @root is not present in the original spec
//book/../\*\. instance of element(\*, xs:decimal)\] (in XPath 2.0) | $..book..\*@number() | Get the numeric values within the book array | @number(), the other basic types (@boolean(), @string()), other low-level derived types (@null(), @object(), @array()), the JSONSchema-added type, @integer(), the compound type @scalar() (which also accepts undefined
and non-finite numbers for JavaScript objects as well as all of the basic non-object/non-function types), the type, @other(), to be used in conjunction with a user-defined callback (see otherTypeCallback
) and the following non-JSON types that can nevertheless be used with JSONPath when querying non-JSON JavaScript objects (@undefined(), @function(), @nonFinite()) are not present in the original spec
//book/name() = 'category' and matches(., 'tion$') (XPath 2.0) | $..book.\?(@property === "category" && @.match(/TION$/i)) | All categories of books which match the regex (end in 'TION' case insensitive) | @property is not present in the original spec.
//book/matches(name(), 'bn$')/parent:: (XPath 2.0) | $..book.\?(@property.match(/bn$/i))^ | All books which have a property matching the regex (end in 'TION' case insensitive) | @property is not present in the original spec. Note: Uses the parent selector \^ at the end of the expression to return to the parent object; without the parent selector, it matches the two isbn
key values.
| | `
` (e.g.,
$` to match a property literally named
$) | Escapes the entire sequence following (to be treated as a literal) |
`` is not present in the original spec; to get a literal backtick, use an additional backtick to escapeAny additional variables supplied as properties on the optional "sandbox" object option are also available to (parenthetical-based) evaluations.
Potential sources of confusion for XPath users
- In JSONPath, a filter expression, in addition to its
@
being a
- In JSONPath, array indexes are, as in JavaScript, 0-based (they begin
- In JSONPath, equality tests utilize (as per JavaScript) multiple equal signs
Command line interface
A basic command line interface (CLI) is provided. Access it usingnpx jsonpath-plus <json-file> <jsonpath-query>
.Ideas
- Support OR outside of filters (as in XPath
|
) and grouping. - Create syntax to work like XPath filters in not selecting children?
- Allow option for parentNode equivalent (maintaining entire chain of
parent-and-parentProperty objects up to root)
Development
Running the tests on Node:npm test
For in-browser tests:
- Serve the js/html files:
npm run browser-test
- Visit http://localhost:8082/test/.