zujson is a JSON parser and writer that converts between JSON and ordinary R vectors and lists by a small, documented type mapping.
- It is built on yyjson, whose sources it bundles, so it needs no system library and no package outside base R.
- No value changes type to fit a vector: an array of numbers and strings stays a list, unless you pass
simplify = "coerce". - It reads and writes raw bytes as well as strings, which is how an HTTP body arrives and leaves.
- Every failure is a condition that inherits from
zujson_error, so one handler catches them all.
Installation
Install the released version from CRAN:
install.packages("zujson")Or the development version from GitHub:
# install.packages("pak")
pak::pak("pedrobtz/zujson")Usage
json_parse() reads JSON into R. An object becomes a named list, and an array whose elements share a type becomes an atomic vector:
str(json_parse('{"ok": true, "ids": [1, 2, 3], "next": null}'))
#> List of 3
#> $ ok : logi TRUE
#> $ ids : int [1:3] 1 2 3
#> $ next: NULLjson_write() goes the other way. A fully named list becomes an object, a vector an array, and a length-one vector a bare value:
cat(json_write(list(query = "cats", limit = 10L, tags = c("a", "b"))))
#> {"query":"cats","limit":10,"tags":["a","b"]}zujson also reads files and NDJSON, builds data frames from arrays of records, and checks whether input is valid JSON; Get started shows each of these.
Learn more
- Get started takes one response body through parsing, writing, raw bytes, errors and NDJSON.
- Read a large API document works through the 3.3 MB Kubernetes API specification.
- Handle malformed or untrusted input covers the condition classes and the limits that bound the work.
- Coming from jsonlite lists where the two packages give different results, and why.
- The reference has the full type mapping in both directions.
zujson is narrower than jsonlite by design; use jsonlite when you need matrices, custom serializers or input too large to hold in memory.