jsonpath

JSONPath

This is Go package providing the features that retrieves a part of the JSON objects according to the query written in the JSONPath syntax.

The core syntaxes of the JSONPath on which this package is based:

• Stefan Gössner's JSONPath - XPath for JSON
• Christoph Burgmer's json-path-comparison
• JSONPath Internet Draft Development

Note:

Please check my compare result to know which responses are adapted. Unfortunately, the proposals that is also discussing in "json-path-comparison" and the draft of the Internet Draft were not finalized at the start of development and are not adopted outright.

Getting started

go get github.com/AsaiYusuke/jsonpath

Simple example:

package main

import (
  "encoding/json"
  "fmt"

  "github.com/AsaiYusuke/jsonpath"
)

func main() {
  jsonPath, srcJSON := `$.key`, `{"key":"value"}`
  var src interface{}
  json.Unmarshal([]byte(srcJSON), &src)
  output, _ := jsonpath.Retrieve(jsonPath, src)
  outputJSON, _ := json.Marshal(output)
  fmt.Println(string(outputJSON))
  // Output:
  // ["value"]
}

Basic design

• PEG separated the JSONPath syntax analyzer from functionality itself to simplify the source.
• The error specification allows package users to handle errors appropriately.
• Adopted more of the consensus behavior from the Christoph Burgmer's json-path-comparison. Adapted my own behavior to the other part of the such consensus behavior that found difficult to use.
• Equipped with numerous unit tests and tried to eliminate the bugs that return strange result.

How to use

* Retrieve one-time, or successively

The Retrieve function returns a retrieved JSON object by a one-time sequential operation (analyzing syntax and retrieving objects) using the given JSONPath and the source JSON object :

output, err := jsonpath.Retrieve(jsonPath, src)

The Parse function returns a parser-function that completed to analyze the JSONPath syntax. By using this returned parser-function it can be performed successively a retrieve with the same JSONPath syntax :

jsonPath, err := jsonpath.Parse(jsonPath)
output1, err1 := jsonPath(src1)
output2, err2 := jsonPath(src2)
:

* Error handling

If there is a problem with the execution of the Retrieve, Parse or prepared parser-functions, an error type is returned. These error types define the corresponding symptom, as listed below:

Syntax analyze errors from Retrieve, Parse

Runtime errors from Retrieve, parser-functions

The type checking is convenient to recognize which error happened.

  :
  _,err := jsonpath.Retrieve(jsonPath, srcJSON)
  if err != nil {
    switch err.(type) {
    case jsonpath.ErrorIndexOutOfRange:
      fmt.printf(`retry with other srcJSON: %v`, err)
      continue
    case jsonpath.ErrorInvalidArgumentFormat:
      return nil, fmt.errorf(`specified invalid argument: %v`, err)
    }
    :
  }

* Function syntax

Function is a feature that allows you to format JSONPath results by using pre-registered user functions and the instruction syntaxes at the end of the JSONPath statement.

There are two ways to use function:

Filter function

The filter function applies a user function to each values in the JSONPath result to get converted.

  config := jsonpath.Config{}
  config.SetFilterFunction(`twice`, func(param interface{}) (interface{}, error) {
    if floatParam, ok := param.(float64); ok {
      return floatParam * 2, nil
    }
    return nil, fmt.Errorf(`type error`)
  })
  jsonPath, srcJSON := `$[*].twice()`, `[1,3]`
  var src interface{}
  json.Unmarshal([]byte(srcJSON), &src)
  output, _ := jsonpath.Retrieve(jsonPath, src, config)
  outputJSON, _ := json.Marshal(output)
  fmt.Println(string(outputJSON))
  // Output:
  // [2,6]

Aggregate function

Aggregate function converts all values in the JSONPath result into a single value by applying them to a user function.

  config := jsonpath.Config{}
  config.SetAggregateFunction(`max`, func(params []interface{}) (interface{}, error) {
    var result float64
    for _, param := range params {
      if floatParam, ok := param.(float64); ok {
        if result < floatParam {
          result = floatParam
        }
        continue
      }
      return nil, fmt.Errorf(`type error`)
    }
    return result, nil
  })
  jsonPath, srcJSON := `$[*].max()`, `[1,3]`
  var src interface{}
  json.Unmarshal([]byte(srcJSON), &src)
  output, _ := jsonpath.Retrieve(jsonPath, src, config)
  outputJSON, _ := json.Marshal(output)
  fmt.Println(string(outputJSON))
  // Output:
  // [3]

* Accessing JSON

You can get a collection of accessors ( Getters / Setters ) to the input JSON instead of the retrieved values by giving Config.SetAccessorMode(). These accessors can be used to update the original nodes retrieved by JSONPath in the input JSON. See the Example for usage.

Note:

It is not possible to use Setter for some execution results, such as including function syntax.

Also, operations using accessors follow the map/slice manner of Go language, so if you use accessors after changing the structure of JSON, you need to pay attention to the behavior caused by the operation. If you want to handle it casually, you may want to retrieve the accessor again each time you change the structure of JSON.

Differences

Some behaviors that differ from the consensus exists in this package. For the entire comparisons, please check this result to see which responses are different. These behaviors will be changed in the future if appropriate ones are found.

Character types

The following character types can be available for identifiers in dot-child notation.

Character types of printable symbols other than hyphens and underscores can be used by escaping them.

JSONPath : $.abc\.def
srcJSON  : {"abc.def":1}
Output   : 1

Wildcard in qualifier

The wildcard in qualifier can be specified mixed with other subscript syntaxes.

JSONPath : $[0,1:3,*]
srcJSON  : [0,1,2,3,4,5]
Output   : [0,1,2,0,1,2,3,4,5]

Regular expression

The regular expression syntax works as a regular expression in Go lang. In particular, you can use "(?i)" to specify the regular expression as the ignore case option.

JSONPath : $[?(@.a=~/(?i)CASE/)]
srcJSON  : ["Case","Hello"]
Output   : ["Case"]

JSONPaths in the filter-qualifier

In the case of the comparators and regular expressions in the filter qualifier, the following JSONPaths that return a value group cannot be specified. On the other hand, in the case of the existence check in the filter qualifier, it can be specified.

Benchmarks

I benchmarked two JSONPaths using several libraries for the Go language. What is being measured is the cost per job for a job that loops a lot after all the prep work is done. There was a difference in execution performance between the libraries, but if the number of queries using JSONPaths is little, I don't think there will be a big difference between any of them.

Also, the combination of the JSON and JSONPath syntax input to the library will yield drastically different results, so this benchmark is for informational purposes only and should be re-measured at every opportunity.

• BenchmarkAsaiYusukeJSONPath
• BenchmarkOhler55Ojg
• BenchmarkBhmjJSONSlice
• BenchmarkPaesslerAGJSONPath
• BenchmarkOliveagleJsonpath

JSONPath for comparison with more libraries

This is the result of a JSONPath that all libraries were able to process. The fastest library was oliveagle/jsonpath, and the JSONPath in this example gave the most ideal score.

JSONPath : $.store.book[0].price

BenchmarkAsaiYusukeJSONPath_threeLevelsWithIndex-4         	 5674974	       207 ns/op	      56 B/op	       3 allocs/op
BenchmarkOhler55Ojg_threeLevelsWithIndex-4                 	 1728166	       683 ns/op	    1040 B/op	       2 allocs/op
BenchmarkBhmjJSONSlice_threeLevelsWithIndex-4              	  546034	      2213 ns/op	      32 B/op	       1 allocs/op
BenchmarkPaesslerAGJSONPath_threeLevelsWithIndex-4         	 1875916	       648 ns/op	     208 B/op	       7 allocs/op
BenchmarkOliveagleJsonpath_threeLevelsWithIndex-4          	13331599	        90.3 ns/op	       0 B/op	       0 allocs/op

A slightly complex JSONPath

Libraries that can handle complex syntax are limited to a few. Among these libraries, my library is the fastest at the moment.

JSONPath : $..book[?(@.price > $.store.bicycle.price)]

BenchmarkAsaiYusukeJSONPath_recursiveDescentWithFilter-4   	  387717	      3325 ns/op	     656 B/op	      26 allocs/op
BenchmarkOhler55Ojg_recursiveDescentWithFilter-4           	  231213	      5237 ns/op	    5240 B/op	      20 allocs/op
BenchmarkBhmjJSONSlice_recursiveDescentWithFilter-4        	   52902	     22661 ns/op	    3032 B/op	      57 allocs/op
BenchmarkPaesslerAGJSONPath                                	  not supported
BenchmarkOliveagleJsonpath                                 	  not supported

{ "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
  }
  }
}

Processor  : Intel Core i5-6267U 2.90GHz
Memory     : 16.0 GB
OS         : Windows 10
Go version : go1.15.6 windows/amd64

Project progress

• Syntax
  • Identifier
    • identifier in dot notations
    • identifier in bracket notations
    • wildcard
    • multiple-identifier in bracket
    • recursive retrieve
  • Qualifier
    • index
    • slice
    • wildcard
    • Filter
      • logical operation
      • comparator
      • JSONPath retrieve in filter
    • script
  • Function
    • filter
    • aggregate
  • Refer to the consensus behaviors
• Archtecture
  • PEG syntax analyzing
  • Error handling
  • Function
  • Accessing JSON
• Go language manner
  • retrieve with the object in interface unmarshal
  • retrieve with the json.Number type
• Source code
  • Release version
  • Unit tests
    • syntax tests
    • benchmark
    • coverage >80%
  • Examples
  • CI automation
  • Documentation
    • README
    • API doc
  • comparison result (local)
• Development status
  • determine requirements / functional design
  • design-based coding
  • testing
  • documentation
• Future ToDo
  • Refer to the something standard
  • Go language affinity
    • retrieve with the object in struct unmarshal
    • retrieve with the struct tags
    • retrieve with the user defined objects