JSON Slice

What is it?

JSON Slice is a Go package which allows to execute fast jsonpath queries without unmarshalling the whole data.

Sometimes you need to get a single value from incoming json using jsonpath, for example to route data accordingly or so. To do that you must unmarshall the whole data into interface{} struct and then apply some jsonpath library to it, only to get just a tiny little value. What a waste of resourses! Well, now there's jsonslice.

Simply call jsonslice.Get on your raw json data to slice out just the part you need. The []byte received can then be unmarshalled into a struct or used as it is.

Getting started

1. install

$ go get github.com/bhmj/jsonslice

2. use it

import "github.com/bhmj/jsonslice"
import "fmt"

func main() {
    var data = []byte(`
    { "sku": [ 
        { "id": 1, "name": "Bicycle", "price": 160, "extras": [ "flashlight", "pump" ] },
        { "id": 2, "name": "Scooter", "price": 280, "extras": [ "helmet", "gloves", "spare wheel" ] }
      ]
    } `)

    a, _ := jsonslice.Get(data, "$.sku[0].price")
    b, _ := jsonslice.Get(data, "$.sku[1].extras.count()")
    c, _ := jsonslice.Get(data, "$.sku[?(@.price > 200)].name")
    d, _ := jsonslice.Get(data, "$.sku[?(@.extras.count() < 3)].name")

    fmt.Println(string(a)) // 160
    fmt.Println(string(b)) // 3
    fmt.Println(string(c)) // ["Scooter"]
    fmt.Println(string(d)) // ["Bicycle"]
}

Run in Go Playground

Package functions

jsonslice.Get(data []byte, jsonpath string) ([]byte, error)

• get a slice from raw json data specified by jsonpath

Specs and references

• Originally based on Stefan Gössner's article.
• Expression evaluator designed to comply with ECMAScript Language: Expressions reference.
• A proud member of amazing Christoph Burgmer's json-path-comparison.

Syntax features

1. Classic dot notation ($.simple_key) is limited to alphanumeric characters. For more complex cases use $['complex key!'] or $.'complex key!'.

2. A single index reference returns an element, not an array; a slice reference always returns array:

$ echo '[{"name":"abc"}, {"name":"def"}]' | ./build/jsonslice '$[0].name' 
"abc"

$ echo '[{"name":"abc"}, {"name":"def"}]' | .build//jsonslice '$[0:1].name'
["abc"]

3. Indexing or slicing on root node is supported (assuming json is an array and not an object):

$ cat example/sample1.json | ./build/jsonslice '$[0].author'

Expressions

Overview

  $                   -- root node (can be either object or array)
  $.node              -- dot-notated child
  $.node.deeper       -- dot-notated child (chain of them)
  $.'some node'       -- dot-notated child (syntax extension)
  ['node']            -- bracket-notated child
  ['foo','bar']       -- bracket-notated children (aggregation)
  [5]                 -- array index
  [-5]                -- negative index means "from the end"
  [1:9]               -- array slice
  [1:9:2]             -- array slice (+step)
  .*  .[*]  .[:]      -- wildcard
  ..key               -- deepscan
  .'\''               -- escape sequences supported (\", \', \n, \r, \t, \0, \\)
  .'\x0A'            -- escaped hex bytes supported
  .'\u00F6'          -- escaped 16-bit unicode codepoints supported
  .'\U000000F6'      -- escaped 32-bit unicode codepoints supported

Functions

  $.obj.length()      -- number of elements in an array or string length, depending on the obj type
  $.obj.count()       -- same as above
  $.val.size()        -- value size in bytes (as is)

Slices

  $.arr[start:end:step]
  $.arr[start:end]

Selects elements from start (inclusive) to end (exclusive), stepping by step. If step is omitted or zero, then 1 is implied. Out-of-bounds values are reduced to the nearest bound.

If step is positive:

• empty start treated as the first element inclusive
• empty end treated as the last element inclusive
• start should be less then end, otherwise result will be empty

If step is negative:

• empty start treated as last element inclusive
• empty end treated as the first element inclusive
• start should be greater then end, otherwise result will be empty

Filters

  [?(<expression>)]  -- filter expression. Applicable to arrays only
  @                  -- the root of the current element of the array. Used only within a filter.
  @.val              -- a field of the current element of the array.

Filter operators

Operator	Description
math	+ - * / % **
===	Strict equality (mimics JavaScript). Examples: true===true, 42===42
==	Abstract equality (mimics JavaScript). Examples: true=="1", 42=="42". Use single or double quotes for string expressions. [?(@.color=='red')] or [?(@.color=="red")]
!=	Abstract not equal to [?(@.author != "Herman Melville")]
!==	Strict not equal to ?(@.tag !== "1")
>	Greater than [?(@.price > 10)]
>=	Greater than or equal to
<	Less than
<=	Less than or equal to
=~	Match a regexp [?(@.name =~ /sword.*/i]
!~ or !=~	Don't match a regexp [?(@.name !~ /sword.*/i]
&&	Logical AND [?(@.price < 10 && @isbn)]
||	Logical OR [?(@.price > 10 || @.category == 'reference')]
!	Logical NOT [?(!@.is_expensive)]
|	Bitwise OR [?(@.bits|@.pieces > 0)]
&	Bitwise AND [?(@.bits & 7 == 1)]
^	Bitwise XOR [?(@.bits ^ 1 > 0)]
~	Bitwise NOT [?(~@.bits == 0xF0)]
<<	Bitwise left shift [?(@.bits << 1 == 2)]
>>	Bitwise right shift [?(@.bits >> 1 == 0)]

Comparison details

Comparison mostly complies with JavaScript specifications, see Testing and Comparison Operations.
If you encounter wrong or inconsistent comparison behaviour please let me know by creating an issue in this repository.

Examples

Assuming sample0.json and sample1.json in the example directory:

cat example/sample0.json | ./build/jsonslice '$.store.book[0]'
cat example/sample0.json | ./build/jsonslice '$.store.book[0].title'
cat example/sample0.json | ./build/jsonslice '$.store.book[0:-1]'
cat example/sample1.json | ./build/jsonslice '$[1].author'
cat example/sample0.json | ./build/jsonslice '$.store.book[?(@.price > 10)]'
cat example/sample0.json | ./build/jsonslice '$.store.book[?(@.price > $.expensive)]'

Much more examples can be found in jsonslice_test.go

Benchmarks (Core i9-9880H)

$ go test -bench=. -benchmem -benchtime=4s
goos: darwin
goarch: amd64
pkg: github.com/bhmj/jsonslice
cpu: Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz
++ usually you need to unmarshal the whole JSON to get an object by jsonpath (for reference):
Benchmark_Unmarshal-16                    398998      11268 ns/op    4272 B/op    130 allocs/op
++ and here's a jsonslice.Get:
Benchmark_Jsonslice_Get_Simple-16        1660604       2885 ns/op      24 B/op      1 allocs/op
++ Get() involves parsing a jsonpath, here it is:
Benchmark_JsonSlice_ParsePath-16        11955015        400 ns/op       0 B/op      0 allocs/op
++ if you aggregate some non-contiguous elements, it may take a bit longer (extra mallocs involved):
Benchmark_Jsonslice_Get_Aggregated-16    1000000       4335 ns/op     313 B/op     10 allocs/op
++ usual unmarshalling a large json:
Benchmark_Unmarshal_10Mb-16                  100   40787414 ns/op     224 B/op      5 allocs/op
++ jsonslicing the same json, target element is near the start:
Benchmark_Jsonslice_Get_10Mb_First-16    3459492       1370 ns/op      24 B/op      1 allocs/op
++ jsonslicing the same json, target element is near the end: still beats Unmarshal
Benchmark_Jsonslice_Get_10Mb_Last-16         133   35731931 ns/op      54 B/op      1 allocs/op
PASS
ok      github.com/bhmj/jsonslice       52.452s

Changelog

1.1.2 (2022-01-02) -- Unicode support added. Expression parser upgrade to v0.9.1
Bugfix: indexing of array element inside expression (@[2]).
Bugfix: ecaped backslash in node key now works ($.["\\"]).
See Test_Fixes function for bugfix details.

1.1.1 (2021-10-20) -- Expression parser upgrade to v0.9.0

1.1.0 (2021-11-12) -- Expression parser/evaluator has been separated to different project and completely rewritten. Parentheses now fully supported. Exponentiation operator added (**). Bitwise operators added (|, &, ^, <<, >>). All expression calculations are JavaScript-compliant.

1.0.6 (2021-10-31) -- JS-like behaviour on string/number/boolean values comparison. === operator added for strict comparison. Strings are now comparable.

1.0.5 (2020-09-22) -- bugfix: $..many.keys used to trigger on many without recursing deeper on keys.

1.0.4 (2020-05-07) -- bugfix: $* path generated panic.

1.0.3 (2019-12-24) -- bugfix: $[0].foo [{"foo":"\\"}] generated "unexpected end of input".

1.0.2 (2019-12-07) -- nested aggregation ($[:].['a','b']) now works as expected. TODO: add option to switch nested aggregation mode at runtime!

1.0.1 (2019-12-01) -- "not equal" regexp operator added (!=~ or !~).

1.0.0 (2019-11-29) -- deepscan operator (..) added, slice with step $[1:9:2] is now supported, syntax extensions added. GetArrayElements() removed.

0.7.6 (2019-09-11) -- bugfix: escaped backslash at the end of a string value.

0.7.5 (2019-05-21) -- Functions count(), size(), length() work in filters.

$.store.bicycle.equipment[?(@.count() == 2)] -> [["light saber", "apparel"]]

0.7.4 (2019-03-01) -- Mallocs reduced (see Benchmarks section).

0.7.3 (2019-02-27) -- GetArrayElements() added.

0.7.2 (2018-12-25) -- bugfix: closing square bracket inside a string value.

0.7.1 (2018-10-16) -- bracket notation is now supported.

$.store.book[:]['price','title'] -> [[8.95,"Sayings of the Century"],[12.99,"Sword of Honour"],[8.99,"Moby Dick"],[22.99,"The Lord of the Rings"]]

0.7.0 (2018-07-23) -- Wildcard key (*) added.

$.store.book[-1].* -> ["fiction","J. R. R. Tolkien","The Lord of the Rings","0-395-19395-8",22.99]
$.store.*[:].price -> [8.95,12.99,8.99,22.99]

0.6.3 (2018-07-16) -- Boolean/null value error fixed.

0.6.2 (2018-07-03) -- More tests added, error handling clarified.

0.6.1 (2018-06-26) -- Nested array indexing is now supported.

$.store.bicycle.equipment[1][0] -> "peg leg"

0.6.0 (2018-06-25) -- Regular expressions added.

$.store.book[?(@.title =~ /(dick)|(lord)/i)].title -> ["Moby Dick","The Lord of the Rings"]

0.5.1 (2018-06-15) -- Logical expressions added.

$.store.book[?(@.price > $.expensive && @.isbn)].title -> ["The Lord of the Rings"]

0.5.0 (2018-06-14) -- Expressions (aka filters) added.

$.store.book[?(@.price > $.expensive)].title -> ["Sword of Honour","The Lord of the Rings"]

0.4.0 (2018-05-16) -- Aggregating sub-queries added.

$.store.book[1:3].author -> ["John","William"]

0.3.0 (2018-05-05) -- MVP.

Roadmap

• length(), count(), size() functions
• filters: simple expressions
• filters: complex expressions (with logical operators)
• nested arrays support
• wildcard operator (*)
• bracket notation for multiple field queries ($['a','b'])
• deepscan operator (..)
• syntax extensions: $.'keys with spaces'.price
• flexible syntax: $[0] works on both [1,2,3] and {"0":"abc"}
• JavaScript-compatible expressions
• Unicode support
• IN (), NOT IN ()
• cache parsed jsonpaths of used variables at filterMatch.varFunc(str)
• Optionally unmarshal the result
• Option to select aggregation mode (nested or plain)

Contributing

1. Fork it!
2. Create your feature branch: git checkout -b my-new-feature
3. Commit your changes: git commit -am 'Add some feature'
4. Push to the branch: git push origin my-new-feature
5. Submit a pull request :)

Licence

MIT

Author

Michael Gurov aka BHMJ