jsonpath-faster

jsonpath-faster

Query JavaScript objects with JSONPath expressions. A faster compiling / cached version of jsonpath.

Compiled JSONPaths

This module is designed (and tested) to be highly compatible with jsonpath - with many extensions.

It compiles JSONpath expressions into the corresponding Javascript and caches the resulting code. For any JSONPath that is used more than a few times the speedup is considerable.

Here are some comparative benchmarks. The first two numbers are operations per second. Each test involves a mix of query, nodes and paths with and without limiting counts.

JSONPath	jsonpath	jsonpath-faster	ratio
$..*	101,974	2,321,381	19.12
$..author	88,589	1,397,474	14.70
$..[1]	71,301	1,567,123	18.91
$.store..price	64,265	1,485,605	21.17
$..book[2]	55,529	1,587,166	24.69
$..book[:2]	48,938	1,508,403	27.24
$..book[?(@.isbn)]	50,508	1,517,967	26.20
$..book[-1:]	47,421	1,518,481	28.17
$..book[?(@.price==8.95)]	47,136	1,567,365	29.05
$..book[0,1]	42,022	1,562,645	32.96
$..book[?(@.price<10)]	46,786	1,586,545	29.48
$..book[(@.length-1)]	29,671	1,612,102	49.18
$.*	182,293	9,010,209	49.51
$.store	183,276	8,728,225	47.76
$.store.*	95,512	5,338,002	55.93
$.store[*]	89,084	5,578,182	62.54
$.store.bicycle	99,969	8,550,540	85.52
$.store.book[*].author	50,153	6,814,045	135.63
$.store.book.1	73,105	8,764,344	119.91
$.store.book[1]	63,779	8,686,778	136.63
$.store.bicycle["color"]	59,106	8,244,142	139.54

With longer paths the speed advantage increases. You can also use Nests to combine multiple JSONpaths and corresponding actions into a single function eliminating the redundancy of scanning the same parts of an object multiple times.

Memory usage

For most purposes jsonpath-faster will be a drop in replacement for jsonpath however it does cache one or more implementation functions for each path it sees so in cases where there are large number of distinct paths there could be a lot of cached generated code.

Compatibility

In addition to its own test suite jsonpath-faster passes all of jsonpath's tests.

There are two known differences (and quite a few extensions):

Script and filter expressions are sanitised differently. In general jsonpath-faster is slightly more restrictive in what it will allow (no function calls for example).

Vivification is handled differently. jsonpath-faster will only vivify portions of the JSONPath that occur after any wildcards or selectors.

Query Example

This section of the documentation is copied directly from jsonpath.

var cities = [
  { name: "London", population: 8615246 },
  { name: "Berlin", population: 3517424 },
  { name: "Madrid", population: 3165235 },
  { name: "Rome", population: 2870528 }
];

var jp = require("jsonpath-faster");
var names = jp.query(cities, "$..name");

// [ "London", "Berlin", "Madrid", "Rome" ]

Install

Install from npm:

$ npm install jsonpath-faster

JSONPath Syntax

Here are syntax and examples adapted from Stefan Goessner's original post introducing JSONPath in 2007.

JSONPath	Description
$	The root object/element
@	The current object/element
.	Child member operator
..	Recursive descendant operator
*	Wildcard matching all objects/elements regardless their names
[]	Subscript operator
[,]	Union operator for alternate names or array indices as a set
[start:end:step]	Array slice operator borrowed from ES4 / Python
?()	Applies a filter (script) expression via static evaluation
()	Script expression via static evaluation

Given this sample data set, see example expressions below:

{
  "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
    }
  }
}

Example JSONPath expressions:

JSONPath	Description
$.store.book[*].author	The authors of all books in the store
$..author	All authors
$.store.*	All things in store, which are some books and a red bicycle
$.store..price	The price of everything in the store
$..book[2]	The third book
$..book[(@.length-1)]	The last book via script subscript
$..book[-1:]	The last book via slice
$..book[0,1]	The first two books via subscript union
$..book[:2]	The first two books via subscript array slice
$..book[?(@.isbn)]	Filter all books with isbn number
$..book[?(@.price<10)]	Filter all books cheaper than 10
$..book[?(@.price==8.95)]	Filter all books that cost 8.95
$..book[?(@.price<30 && @.category=="fiction")]	Filter all fiction books cheaper than 30
$..*	All members of JSON structure

Methods

jp.query(obj, pathExpression[, count][, $])

Find elements in obj matching pathExpression. Returns an array of elements that satisfy the provided JSONPath expression, or an empty array if none were matched. Returns only first count elements if specified.

var authors = jp.query(data, "$..author");
// [ 'Nigel Rees', 'Evelyn Waugh', 'Herman Melville', 'J. R. R. Tolkien' ]

A context value $ may be provided. The contents of $ may be accessed in script and filter expressions.

var bargains = jp.query(data, "$..book[?(@.price <= $.price)]", { price: 10 });
// [
//   {
//     category: 'reference',
//     author: 'Nigel Rees',
//     title: 'Sayings of the Century',
//     price: 8.95
//   },
//   {
//     category: 'fiction',
//     author: 'Herman Melville',
//     title: 'Moby Dick',
//     isbn: '0-553-21311-3',
//     price: 8.99
//   }
// ]

jp.paths(obj, pathExpression[, count][, $])

Find paths to elements in obj matching pathExpression. Returns an array of element paths that satisfy the provided JSONPath expression. Each path is itself an array of keys representing the location within obj of the matching element. Returns only first count paths if specified.

var paths = jp.paths(data, "$..author");
// [
//   ['$', 'store', 'book', 0, 'author'] },
//   ['$', 'store', 'book', 1, 'author'] },
//   ['$', 'store', 'book', 2, 'author'] },
//   ['$', 'store', 'book', 3, 'author'] }
// ]

If you'd prefer to receive paths as strings check out Pragma Chains.

jp.nodes(obj, pathExpression[, count][, $])

Find elements and their corresponding paths in obj matching pathExpression. Returns an array of node objects where each node has a path containing an array of keys representing the location within obj, and a value pointing to the matched element. Returns only first count nodes if specified.

var nodes = jp.nodes(data, "$..author");
// [
//   { path: ['$', 'store', 'book', 0, 'author'], value: 'Nigel Rees' },
//   { path: ['$', 'store', 'book', 1, 'author'], value: 'Evelyn Waugh' },
//   { path: ['$', 'store', 'book', 2, 'author'], value: 'Herman Melville' },
//   { path: ['$', 'store', 'book', 3, 'author'], value: 'J. R. R. Tolkien' }
// ]

jp.value(obj, pathExpression[, newValue[, $]])

Returns the value of the first element matching pathExpression. If newValue is provided, sets the value of the first matching element and returns the new value.

If you need to pass a context value without a newValue you must explicitly pass undefined as newValue.

var bargain = jp.value(data, "$..book[?(@.price <= $.price)]", undefined, {
  price: 10
});

jp.parent(obj, pathExpression[, $])

Returns the parent of the first matching element.

jp.apply(obj, pathExpression, fn[, $])

Runs the supplied function fn on each matching element, and replaces each matching element with any value returned from the function. The function is passed the value of each node and its path. Returns matching nodes with their updated values.

var nodes = jp.apply(data, "$..author", function (value, path) {
  return value.toUpperCase();
});
// [
//   { path: ['$', 'store', 'book', 0, 'author'], value: 'NIGEL REES' },
//   { path: ['$', 'store', 'book', 1, 'author'], value: 'EVELYN WAUGH' },
//   { path: ['$', 'store', 'book', 2, 'author'], value: 'HERMAN MELVILLE' },
//   { path: ['$', 'store', 'book', 3, 'author'], value: 'J. R. R. TOLKIEN' }
// ]

If fn returns nothing (undefined) the node's value will not be replaced.

jp.apply(data, "$..author", function (value, path) {
  console.log(value);
});
// Nigel Rees
// Evelyn Waugh
// Herman Melville
// J. R. R. Tolkien

jp.visit(obj, pathExpression, fn[, $])

Very similar to apply but instead of returning the list of visited nodes it returns the passed obj - which may have been updated by fn. This makes it possible to vivify an empty obj. None of the other methods are able to assign directly to obj.

const obj = jp.visit(undefined, "$", () => "Hello");
// obj is "Hello"

jp.parse(pathExpression)

Parse the provided JSONPath expression into path components and their associated operations.

var path = jp.parse("$..author");
// [
//   { expression: { type: 'root', value: '$' } },
//   { expression: { type: 'identifier', value: 'author' }, operation: 'member', scope: 'descendant' }
// ]

jp.stringify(path)

Returns a path expression in string form, given a path. The supplied path may either be a flat array of keys, as returned by jp.nodes for example, or may alternatively be a fully parsed path expression in the form of an array of path components as returned by jp.parse.

var pathExpression = jp.stringify(["$", "store", "book", 0, "author"]);
// "$.store.book[0].author"

Pragma chains

The methods described above potentially walk the whole of an object returning both interior and leaf nodes. When they return a path it is in the form of an array which may be stringified using jp.stringify().

Perhaps instead you'd like to get the stringified paths of all the leaf nodes in an object. The behaviour of jp can be altered using pragmatic chains:

const leaves = jp.string.leaf.paths(obj, "$..*");
// [
//   '$.store.book[0].category',
//   '$.store.book[0].author',
//   '$.store.book[0].title',
//   '$.store.book[0].price',
//   '$.store.bicycle.color',
//   '$.store.bicycle.price'
// ]

The available pragmas are leaf, interior and string.

pragma	effect
leaf	only visit leaf (non-object) nodes
interior	the opposite of leaf: only visit non-leaf (object) nodes
string	where applicable stringify paths before returning them

The order of the pragmas is unimportant but you should try to use them in a consistent order for maximum efficiency.

var n1 = jp.leaf.string.paths(obj, "$..*");
var n2 = jp.string.leaf.paths(obj, "$..*");

The two lines above will cause the path $..* to be compiled twice - once in the 'leaf.string' cache and again in the 'string.leaf' cache.

Tagged literal syntax

You can populate a 3d matix like this but because the path is dynamic it's not the most efficient way.

const matrix = [];
for (let x = 0; x < 3; x++)
  for (let y = 0; y < 3; y++)
    for (let z = 0; z < 3; z++)
      jp.value(matrix, `$[${x}][${y}][${z}]`, { x, y, z });

Because the path is different each time, every call to jp.value() has to compile and cache a new function to handle it. Only if you run the code again later will the cached versions be used.

A more efficient approach is to use a backtick literal tagged with jp.

const matrix = [];
for (let x = 0; x < 3; x++)
  for (let y = 0; y < 3; y++)
    for (let z = 0; z < 3; z++)
      jp`$[${x}][${y}][${z}]`.value(matrix, { x, y, z });

In this case the path is compiled only once (with placeholders for the bound x, y and z values).

Internally the $ context variable is used to pass the bound values to the generated path function.

Nests

Sometimes you need to run many JSONpath queries against each one of a number of objects. Instead of compiling each individual path into its own Javascript function a Nest allows multiple paths to be compiled into a single function.

const survey = [];
const authors = [];
const nest = jp.nest();
nest
  .visitor("$..*", (value, path) => survey.push({ value, path }))
  .visitor("$..author", value => authors.push(value))
  .mutator("$..price", value => value * 1.1);

nest(data); // the nest is a function
// All prices increased by 10%, survey and authors arrays populated

Calling the nest function runs all the actions that you have registered with the nest. Actions with paths that share a common prefix are efficiently compiled so that the prefix is traversed only once. In the following example the code to traverse $.assets[*]..meta is executed only once for each call to nest()

nest
  .visitor("$.assets[*]..meta.id", value => {})
  .visitor("$.assets[*]..meta.author", value => {})
  .visitor("$.assets[*]..meta.modified", value => {});

This can be written more concisely using a nested nest.

nest
  .nest("$.assets[*]..meta")
  .visitor("$.id", value => {})
  .visitor("$.author", value => {})
  .visitor("$.modified", value => {});

Execution order

A nest attempts to behave as if the visitors are executed in the order they were declared even though, as in the example above, the paths may be matched in a different order. To do this it defers all the actions until after the search of the object is complete.

That means that any mutations are executed after the object has been scanned so the following code may have surprising results.

nest
  .mutator("$..thing.seen", true)
  .visitor("$..seen", (value, path) => console.log(`Seen at ${path}`));

The visitor won't match any of the seen flags set by the mutator because the mutations only take place after the object has been scanned. If you need to work with the mutated values use a second nest.

Pragmas

Like jp nests understand pragma chains.

const nest = jp.nest();
nest.string.leaf.visitor("$..*", (value, path) =>
  console.log(`${path}: ${value}`)
);

A nest inherits pragmas from the jp that creates it, so this is equivalent to the previous example:

const nest = jp.string.leaf.nest();
nest.visitor("$..*", (value, path) => {
  console.log(`${path}: ${value}`);
});

In addition to leaf, interior and string Nests support the unordered pragma allowing actions to be fired in whatever order the generated code dictates. Using unordered with setters or mutators may lead to hard to diagnose problems but is generally OK in cases where there are no interdependencies between them.

Nest Methods

jp.nest([path])

Creates a new, empty nest. Actions may be added using its visitor, mutator, setter and at methods. Having added actions the nest may be called as a function to apply the actions to an object.

If path is supplied it will be used as a prefix for all the paths in the nest.

const nest = jp.nest();
nest.visitor("$..*", (value, path) => console.log(path));
for (const doc of docs) {
  nest(doc);
}

The nest function accepts an optional second argument that, if present will be bound to $ and may be referred to in script and filter expressions and in any code compiled using nest.at().

Its return value is the resulting object after all mutators and setters have been applied to it. Usually this is the same as the doc you passed in. However it is possible to vivify an undefined root object.

const mp = jp.nest().setter("$", { empty: false });
const obj = mp(undefined);
// obj is { empty: false }

nest.visitor(path, fn)

Register a visitor function that will be called for each matching node in the object. The function is called with three arguments: value and path and any $ context that was passed to the nest function..

nest.string.visitor(
  "$..books[*].authors[(@.length - 1)].name",
  (value, path, $) => console.log(`${path}: ${value}`)
);

If you don't need the path provide a function that accepts only a single value argument; the generated code is slightly faster if it doesn't have to track the path as it traverses the object.

nest.visitor("$..books[*].authors[(@.length - 1)].name", value =>
  console.log(value)
);

The path argument may be a single path string or an array of paths.

nest.mutator(path, fn)

Similar to visitor but the matched value is set to the return value of the callback function. A mutator does not vivify missing parts of the object.

nest.mutator("$..price", price => price * 3);

If the replacement value is a constant it may be passed directly.

nest.mutator("$..price", price => 1); // everything's a £

The path argument may be a single path string or an array of paths.

nest.setter(path, fn)

Like mutator but with vivification enabled. Like mutator it accepts either a function or a constant.

nest.setter("$.this.does.not.exist.yet", true);

The path argument may be a single path string or an array of paths.

nest.at(path, code)

Inject code directly into the generated nest function. Within the supplied code @ is a magic variable which provides access to various contextual values.

nest.at("$..vehicle", "console.log(@.value, @.path)");

The @ pseudo-variable has these properties

Property	Description
@.value	The value of the current node
@.nvalue	Non-vivifying version of @.value (see below)
@.parent	The parent of the current node
@.pathArray	The path to the current node as an array ["$", "books", 0, "author"]
@.pathString	The path to the current node as a string
@.path	The path as either an array or a string depending on the ambient string pragma

When @.value is used on the left hand side of an assignment it tells the code generator to vivify as far as possible the path leading up to this node. If vivifaction is not desired the code may assign to @.nvalue instead.

nest.at("$..flags.seen", "@.value = true"); // create `seen`
nest.at("$..flags.seen", "@.nvalue = true"); // only sets existing `seen`

The path argument may be a single path string or an array of paths.

nest.nest(path)

Add a prefix path for this call chain.

nest
  .nest("$.assets[*]..meta")
  .visitor("$.id", value => {})
  .visitor("$.author", value => {})
  .visitor("$.modified", value => {});

License

MIT