NOTE

I'm currently working with the SymbolixAU team to add the code/functionality from this package into their CRAN package jsonify. If that all works out as planned, then all future development of this code will be happening in that repo.

jsonparse

R package for parsing JSON. There are already a few R packages that parse JSON data ( jsonlite, rjson ), the intent behind this one is to try to build a package that is faster than the existing options. This project is very young, currently the functions can only handle values of type int, double, logical, and character.

This package is built using the rapidjson C++ library (via the rapidjsonr R package), and Rcpp.

As an additional resource, check out the jsonify package, which uses the rapidjson library to convert R objects to json.

Please report issues, comments, or feature requests.

Installation

Install from this repo:

# install.packages("devtools")
devtools::install_github("ChrisMuir/jsonparse")

Example Usage

library(jsonparse)
library(jsonify)

# Create json string, using package jsonify
json_str <- jsonify::to_json(
  list(
    "string_key" = "cats", 
    "int_key" = 5L, 
    "double_key" = 99.4, 
    "bool_key" = TRUE, 
    "vector_key" = c(9L, 10L, 11L, 12L), 
    "list_key" = list("dogs", 55.3)
  )
)

# print json_str
json_str

#> {"string_key":["cats"],"int_key":[5],"double_key":[99.4],"bool_key":[true],"vector_key":[9,10,11,12],"list_key":[["dogs"],[55.3]]}

jsonparse::from_json(json_str)

#> $string_key
#> [1] "cats"

#> $int_key
#> [1] 5

#> $double_key
#> [1] 99.4

#> $bool_key
#> [1] TRUE

#> $vector_key
#> [1]  9 10 11 12

#> $list_key
#> $list_key[[1]]
#> [1] "dogs"

#> $list_key[[2]]
#> [1] 55.3

Benchmarks

library(jsonlite)
jl_fromJSON <- jsonlite::fromJSON
library(rjson)
rj_fromJSON <- rjson::fromJSON

Test 1

json_str <- jsonify::to_json(
  list(
    "ints" = 1L:100000L, 
    "doubles" = rnorm(100000), 
    "strings" = stringi::stri_rand_strings(100000, 8), 
    "bools" = sample(c(TRUE, FALSE), size = 100000, replace = TRUE)
  )
)

microbenchmark::microbenchmark(
  jsonparse = from_json(json_str), 
  rjson = rj_fromJSON(json_str), 
  jsonlite = jl_fromJSON(json_str, simplifyVector = FALSE)
)

#> Unit: milliseconds
#>      expr       min        lq      mean    median       uq       max neval
#> jsonparse  24.01423  27.23423  29.97406  29.60571  32.0372  44.45918   100
#>     rjson 100.33898 109.40579 119.47500 117.18489 126.3026 226.08668   100
#>  jsonlite 207.57313 219.68605 230.28911 226.09717 239.4743 277.65422   100

Test 2

json_str <- lapply(1:10000, function(x) {
  list(
    "string_key" = "cats", 
    "int_key" = 5L, 
    "double_key" = 99.4, 
    "bool_key" = TRUE, 
    "vector_key" = c(9L, 10L, 11L, 12L), 
    "list_key" = list("dogs", 55.3)
  )
})
json_str <- jsonify::to_json(json_str)

microbenchmark::microbenchmark(
  jsonparse = from_json(json_str), 
  rjson = rj_fromJSON(json_str), 
  jsonlite = jl_fromJSON(json_str, simplifyVector = FALSE)
)

#> Unit: milliseconds
#>      expr       min        lq      mean    median        uq      max neval
#> jsonparse  15.64111  17.31473  22.00976  19.56679  22.38043 105.1796   100
#>     rjson  39.75470  47.17560  57.10193  52.63964  60.26783 168.0407   100
#>  jsonlite 106.23252 111.21012 118.63235 115.01797 119.71804 238.5919   100