Learn R Programming

RJSONIO (version 1.3-1.9)

toJSON: Convert an R object to a string in Javascript Object Notation

Description

This function and its methods convert an R object into a string that represents the object in Javascript Object Notation (JSON).

The different methods try to map R's vectors to JSON arrays and associative arrays. There is ambiguity here as an R vector of length 1 can be a JSON scalar or an array with one element. When there are names on the R vector, the descision is clearer. We have introduced the emptyNamedList variable to identify an empty list that has an empty names character vector and so maps to an associative array in JSON, albeit an empty one.

Objects of class AsIs in R, i.e. that are enclosed in a call to I() are treated as containers even if they are of length 1. This allows callers to indicate the desired representation of an R "scalar" as an array of length 1 in JSON

Usage

toJSON(x, container = isContainer(x, asIs, .level), 
        collapse = "\n", ..., .level = 1L,
         .withNames = length(x) > 0 && length(names(x)) > 0, .na = "null",
         .escapeEscapes = TRUE, pretty = FALSE, asIs = NA, .inf = " Infinity")

Value

A string containing the JSON content.

Arguments

x

the R object to be converted to JSON format

...

additional arguments controlling the formatting of the JSON.

container

a logical value indicating whether to treat the object as a vector/container or a scalar and so represent it as an array or primitive in JavaScript.

collapse

a string that is used as the separator when combining the individual lines of the generated JSON content

.level

an integer value. This is not a parameter the caller is supposed to supply. It is a value that is passed in recursive calls to identify the top-level and sub-level serialization to JSON and so help to identify when a scalar needs to be in a container and when it is legitimate to output a scalar value directly.

.withNames

a logical value. If we are dealing with a named vector/list, we typically generate a JSON associative array/dictionary. If there are no names, we create a simple array. This argument allows us to explicitly control whether we use a dictionary or to ignore the names and use an array.

.na

a value to use when we encounter an NA value in the R objects. This allows the caller to convert these to whatever makes sense to them. For example, we might specify this as "null" and then the NA values will appear as null in the JSON output. One can also specify an unusual numeric value, e.g. -9999999 to indicate a missing value!

.escapeEscapes

a logical value that controls how new line and tab characters are serialized. If this is TRUE, we preserve them symbolically by escaping the \. Otherwise, we replace them with their literal value.

pretty

a logical value that controls if extra processing is done on the result to make it indented for easier human-readability. At present, this reparses the generated JSON content and re-formats it (using libjson). This means that there can be three copies of the data in memory simultaneously - the original data, the JSON text and the pretty-printed version of the JSON text. For large objects, this can require a lot of memory.

asIs

a logical value that, if TRUE causes R vectors of length 1 to be represented as arrays in JSON, but if FALSE to be represented as scalars, where appropriate (i.e. not the top level of the JSON content). This avoids having to explicitly mark sub-elements in an R object as being of class AsIs.

.inf

how to represent infinity in JSON. This should be a string.

Author

Duncan Temple Lang <duncan@wald.ucdavis.edu>

References

http://www.json.org

See Also

fromJSON

Examples

Run this code
 toJSON(1:10)
 toJSON(rnorm(3))
 toJSON(rnorm(3), digits = 4)

 toJSON(c("Duncan", "Temple Lang"))

 toJSON(c(FALSE, FALSE, TRUE))

   # List of elements
 toJSON(list(1L, c("a", "b"), c(FALSE, FALSE, TRUE), rnorm(3)))
   # with digits controlling formatting of sub-elements
 toJSON(list(1L, c("a", "b"), c(FALSE, FALSE, TRUE), rnorm(3)),
          digits = 10)

   # nested lists
 toJSON(list(1L, c("a", "b"), list(c(FALSE, FALSE, TRUE), rnorm(3))))

   # with names
 toJSON(list(a = 1L, c("a", "b"), c(FALSE, FALSE, TRUE), rnorm(3)))

 setClass("TEMP", representation(a = "integer", xyz = "logical"))
 setClass("TEMP1", representation(one = "integer", two = "TEMP"))

 new("TEMP1", one = 1:10, two = new("TEMP", a = 4L, xyz = c(TRUE, FALSE)))


 toJSON(list())
 toJSON(emptyNamedList)
 toJSON(I(list("hi")))
 toJSON(I("hi"))


 x = list(list(),
          emptyNamedList,
          I(list("hi")),
          "hi",
          I("hi"))
 toJSON(x)

  # examples of specifying .withNames
 toJSON(structure(1:3, names = letters[1:3]))
 toJSON(structure(1:3, names = letters[1:3]), .withNames = FALSE)


  # Controlling NAs and mapping them to whatever we want.
 toJSON(c(1L, 2L, NA), .na = "null")
 toJSON(c(1L, 2L, NA), .na = -9999)

 toJSON(c(1, 2, pi, NA), .na = "null")

 toJSON(c(TRUE, FALSE, NA), .na = "null")

 toJSON(c("A", "BCD", NA), .na = "null")

 toJSON( factor(c("A", "B", "A", NA, "A")), .na = "null" )

 toJSON(list(TRUE, list(1, NA), NA), .na = "null")



 setClass("Foo", representation(a = "integer", b = "character"))
 obj = new("Foo", a = c(1L, 2L, NA, 4L), b = c("abc", NA, "def"))
 toJSON(obj)
 toJSON(obj, .na = "null")

  # hexmode example with .na ?

 toJSON(matrix(c(1, 2, NA, 4), 2, 2), .na = "null")
 toJSON(matrix(c(1, 2, NA, 4), 2, 2), .na = -9999999)


 x = '"foo\tbar\n\tagain"'
 cat(toJSON(x))
 cat(toJSON(list(x)))

  # if we want to expand the new lines and tab characters
 cat(toJSON(x), .escapeEscapes = FALSE)


  # illustration of the asIs argument
  cat(toJSON(list(a = 1, b = 2L, c = TRUE,
                  d = c(1, 3),
                  e = "abc"), asIs = TRUE))

  cat(toJSON(list(a = 1, b = 2L, c = TRUE,
                  d = c(1, 3),
                  e = "abc"), asIs = FALSE))

   # extra level
  cat(toJSON(list(a = c(x = 1), b = 2L, c = TRUE,
                  d = list(1, 3),
                  e = "abc"), asIs = FALSE, pretty = TRUE))


   # data frame by row as arrays
  twoRows = data.frame(a = 1:2, b = as.numeric(1:2))
  j = toJSON(twoRows, byrow = TRUE)
  r = data.frame(do.call(rbind, fromJSON(j)))

   # here we keep the names of the columns on each row
   # which allows us to round-trip the object back to R
  j = toJSON(twoRows, byrow = TRUE, colNames = TRUE)
  r = data.frame(do.call(rbind, fromJSON(j)))

Run the code above in your browser using DataLab