json-query lets you quickly select values, patterns or types from json-data. Its input requires a simple string, describing a concise query into your data

install

yarn add @sagold/json-query

• Features
• Introduction
• Breaking Changes
• API
• About patterns
• Further examples

Features

• json-pointer syntax #/list/0/id
• glob-patterns for properties (*, **)
• regex-support for properties {any.*}
• pattern-support for inifinite recursion /tree(/nodes/*)+/value
• or-patterns /node((/left), (/right))
• finite search in circular-data **
• lookahead-rules to test selected property ?property:value and regex values ?property:{\d+}
• and typechecks /value?:array

Quick introduction

Basically, a query is a json-pointer, which describes a path of properties into the json-data

import { get } from "@sagold/json-query";
const input = { object: { a: { id: "id-a" }, b: { id: "id-b" } } };

const values = get(input, "/object/a/id"); // ["id-a"]

But each property may also be a glob-pattern or a regular expression:

* selects all direct children

const input = { object: { a: { id: "id-a" }, b: { id: "id-b" } } };

const values = get(input, "/object/*/id"); // ["id-a", "id-b"]

** selects all values

const input = { object: { a: { id: "id-a" }, b: { id: "id-b" } } };

const values = get(input, "/object/**");
// [ { a: { id: "id-a" }, b: { id: "id-b" } }, { id: "id-a" }, "id-a", { id: "id-b" }, "id-b" ]

{} calls a regular expression

const input = { object: { a: { id: "id-a" }, b: { id: "id-b" } } };

const values = get(input, "/{obj.*}/{.*}/id"); // ["id-a", "id-b"]

Note. Regular expressions within strings, have to escape any backslashes, e.g. instead of {\d} you need to pass {\\d}

lookahead rules are used to validate the current value based on its properties

?child tests if a childProperty is defined

const input = { object: { a: { id: "id-a" }, b: { id: "id-b" } } };

const values = get(input, "/object/*?id"); // [{ id: "id-a" }, { id: "id-b" }]

?child:value tests if a childProperty matches a value

const input = { object: { a: { id: "id-a" }, b: { id: "id-b" } } };

const values = get(input, "/object/*?id:id-b"); // [{ id: "id-b" }]

lookahead rules can also be negated ?child:!value, tested by regex ?child:{^re+}, combined ?child&&other or joined ?child||other. Undefined may be tested with ?property:undefined, per default undefined is excluded from matches.

typechecks can be used to query certain data-types

?:<type>, where <type> may be any of ["boolean", "string", "number", "object", "array", "value"]

const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };

const values = get(input, "/**?:string"); // ["id-b"]

?:value will match all types except objects and arrays

const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };

const values = get(input, "/**?:value"); // [33, "id-b"]

patterns can be used to combine queries into a single result (OR) and to build up results from recursive queries (+)

Queries can be grouped by parenthesis, where /a/b/c = /a(/b)(/c) = /a(/b/c).

((/a), (/b)) resolves both queries on the previous result

const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };

const values = get(input, "/object((/a), (/b))"); // [{ id: 33 }, { id: "id-b" }]

and the result may be queried further

get(input, "/object((/a), (/b))/id"); // [33, "id-b"]
get(input, "/object((/a), (/b))/id?:number"); // [33]

(/a)+ will repeat the grouped query for all possible results

const input = {
    id: 1,
    a: { // first iteration
        id: 2,
        a: { // second iteration
            id: 3
            a: 4 // last iteration
        }
    }
};

const values = get(input, "/(/a)+"); // [{ id: 2, a: { id: 3, a: 4 } }, { id: 3, a: 4 }, 4]

escaping properties In case you have special characters in property-names or values, you can escape any value using doubled-quotes "<value>":

• escape property-name: '/root/*/"strange/property"' is split to ["root", "*", "strange/property"]
• escape query-property '/root/*?"strange/property":42'
• escape query-value '/root/*?id:"#/pointer/value"'

Breaking Changes

• with version v5.0.0 package has been rename to @sagold/json-query
• with version v4.0.0 (2019/10/01)
  • the api has been simplified to methods query.get and query.delete (removed run and pattern)
• with version v3.0.0
  • the syntax has changed to es6, which might require code transpilation
  • queries for root-pointer (#, #/, /) now callback root object with (rootObject, null, null, "#")
• with v2.0.0 a negated filter (lookahead), e.g. *?valid:!true will not return objects where valid === undefined. To match objects with missing properties you can still query them explicitly with *?valid:!true||valid:undefined

API

json-query exposes get, set, remove and a split-helper

get

per default, get returns a list of all values

import { get } from "@sagold/json-query";
const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };
const values = get(input, "/**?:value"); // [33, "id-b"]

Using the optional value returnType you can change the result type to the following options ["all", "value", "pointer", "map"]. The string values can also be accessed as property on get: get.ALL, get.VALUE, get.POINTER, get.MAP:

import { get } from "@sagold/json-query";
const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };

get(input, "/**?:value", get.VALUE); // [33, "id-b"]
get(input, "/**?:value", get.POINTER); // ["#/object/a/id", "#/object/b/id"]
get(input, "/**?:value", get.MAP); // { "#/object/a/id": 33, "#/object/b/id": "id-b" }

get(input, "/**?:value", get.ALL);
// [
//    [33, "id", { id: 33 }, "#/object/a/id"],
//    ["id-b", "id", { id: "id-b" }, "#/object/b/id"]
// ]

get(input, "/**?:value", (value, key, parent, pointer) => `custom-${pointer}`);
// ["custom-#/object/a/id", "custom-#/object/b/id"]

remove

remove deletes any match from the input data. Note: the input will be modified. If this is unwanted behaviour, copy your data up front.

import { remove } from "@sagold/json-query";
const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };

remove(input, "/object/*/id"); // { object: { a: {}, b: {} } };

Per default, the input object is returned. Setting the optional argument returnRemoved = true, will return a list of the removed items

import { remove } from "@sagold/json-query";
const input = { object: { a: { id: 33 }, b: { id: "id-b" } } };

remove(input, "/object/*/id", true); // [ 33, "id-b" ]

set

set inserts given input-value on result and creates missing properties and arrays. Note: Any expanding queries like * or patterns will not create any intermediate values

set has the following signature

set(input:any, query:string, value:string\|function, force?:string): any

instead of value, you can also pass a function to generate the values to set:

value(pointerOfParent:string, lastPropertyName:string, parentObject:string, pointerAtValue:string): any

Create data from simple properties

import { set } from "@sagold/json-query";

const result = set({}, "/object/id", 42); // { object: { id: 42 }}

Add properties to multiple existing objects

import { set } from "@sagold/json-query";

const result = set({ list: [ { id: 1 }, { id: 2 } ] }, "/list/*/index", 42);
// { list: [ { id: 1, index: 42 }, { id: 2, index: 42 } ] }

Or using a value-function

import { set } from "@sagold/json-query";

const result = set({ list: [ { id: 1 }, { id: 2 } ] }, "/list/*/index",
    ( _, _, parent) => `id-${parent.id}`
);
// { list: [ { id: 1, index: "id-1" }, { id: 2, index: "id-2" } ] }

Currently, set will not override simple values

import { set } from "@sagold/json-query";

const result = set({ value: 2 }, "/value/id", 3);
// { value: 2 }

And queries will not add values to the data

import { set } from "@sagold/json-query";

const result = set({ a: { id: 2 } }, "((/a), (/b))/id", true);
// { a: { id: true } }

When working with arrays, you have to choose between the following actions

• insert item at index 1: /list/[1]/id
• replace item at index 1: /list/1/id
• append item /list/[]/id

Using the force option, you can enforce insertion or replacement, independent of the syntax (same for the whole query)

set(data, "/list/[1]/id", 42, set.REPLACE_ITEMS); // will always replace index
// and
set(data, "/list/1/id", 42, set.INSERT_ITEMS); // will always insert at index

Numbers will always be interpreted as arrays

set({}, "/list/0/id", 42); // { list: [{ id: 42 }]}
set({}, "/list/[]/id", 42); // { list: [{ id: 42 }]}
set({}, "/list/[0]/id", 42); // { list: [{ id: 42 }]}

// but setting an index is respected
set({}, "/list/2/id", 42); // { list: [undefined, undefined, { id: 42 }]}

In order to treat numbers as objects, escape them using double-quotes

set({}, '/list/"2"/id', 42); // { list: { 2: { id: 42 } } }
// or "/list/\"2\"/id"

About patterns

Pattern-queries enable selection of recursive patterns and offer a way to build up a collection of data for further filterung. A pattern uses brackets () to identify repeatable structures and offers multiple selections for the same data-entry.

Using a pattern-query like #/tree((/left),(/right))* will recursively select all left and right-nodes. e.g.

const data = {
  tree: {
    left: {
      id: "1",
      left: { id: "2" },
      right: { id: "3" }
    },
    right: {
      id: "4"
    }
  }
};

const result = get(data, "#/tree((/left),(/right))*/id");
// ["1", "2", "3", "4"]

Note that each pattern-queries is resovled using query.get and thus supports all mentioned features.

One use-case for pattern-queries can be found in json-schema specification. Any definition in #/defs may reference itself or be referenced circular. A linear query cannot describe the corresponding data, but pattern-queries might be sufficient.

details

A pattern is a simple group defined by brackets: #/a(/b)/c, which is identical to #/a/b/c. But a group may also have a quantifier +: #/a(/b)+/c. Using a quantifier, the query within the pattern will be applied as long as it matches any data. Its combined result will then be passed to /c.

e.g. applying the pattern #/a(/b)+/c on the following input data:

const input = {
  a: {
    b: {
      c: "1",
      b: {
        c: "2",
        b: {}
      }
    }
  }
};

will first select property a and then repeatedly select property b: [a/b, a/b/b, a/b/b/b]. This result is filtered by c, which will return ["1", "2"] (the last b-object has no property c).

Patterns can also be used for OR-operations. An OR is identified by a semicolon , and must be within and between patterns, like ((/a/b),(/c)). Not valid patterns are (/a/b, /c) and r/(/a/b),(/c)/f.

Currently, using OR is commutative in a sense that ((/a),(/b)) = ((/b),(/a)), (with a different ordering of the resulting set), distributive so that /a((/b), (/c)) = ((/a/b), (/a/c)). Parenthesis without a quantifier are associative, e.g. #/a/b/c = #/a(/b)/c = #/a(/b/c) = #/a(/b)(/c). Thus, a pattern ((/b)(/c))+ can also be written like (/b/c)+.

further examples

for further examples refer to the unit tests

• query.delete
• query.get
• query.set
• query.split