Streaming JSON parser for Go

jstream is a streaming JSON parser and value extraction library for Go.

Unlike most JSON parsers, jstream is document position- and depth-aware – this enables the extraction of values at a specified depth, eliminating the overhead of allocating encompassing arrays or objects; e.g:

Using the below example document:

we can choose to extract and act only the objects within the top-level array:

f, _ := os.Open("input.json")
decoder := jstream.NewDecoder(f, 1) // extract JSON values at a depth level of 1
for mv := range decoder.Stream() {
  fmt.Printf("%v\n ", mv.Value)
}

output:

map[desc:RGB colors:[red green blue]]
map[desc:CMYK colors:[cyan magenta yellow black]]

likewise, increasing depth level to 3 yields:

red
green
blue
cyan
magenta
yellow
black

optionally, kev:value pairs can be emitted as an individual struct:

decoder := jstream.NewDecoder(f, 2).EmitKV() // enable KV streaming at a depth level of 2

jstream.KV{desc RGB}
jstream.KV{colors [red green blue]}
jstream.KV{desc CMYK}
jstream.KV{colors [cyan magenta yellow black]}

Installing

go get github.com/bcicen/jstream

Commandline

jstream comes with a cli tool for quick viewing of parsed values from JSON input:

cat input.json | jstream -v -d 1
depth	start	end	type   | value

1	004	069	object | {"colors":["red","green","blue"],"desc":"RGB"}
1	073	153	object | {"colors":["cyan","magenta","yellow","black"],"desc":"CMYK"}

Options

Benchmarks

Obligatory benchmarks performed on files with arrays of objects, where the decoded objects are to be extracted.

Two file sizes are used – regular (1.6mb, 1000 objects) and large (128mb, 100000 objects)

In a real world scenario, including initialization and reader overhead from varying blob sizes, performance can be expected as below: