There are two types of R objects we need to handle when converting to JSON, simple and complex.
I’ve categorised them this way because ‘simple’ objects don’t include any form of recursion. That is, a vector can’t contain a data.frame or a list. But a list or data.frame can contain other data.frames, vectors, matrices, scalars, lists, and any combination thereof.
Simple objects ( scalars, vectors and matrices ) get converted to JSON ARRAYS
## scalar -> single array
to_json( 1 )
# [1.0]
to_json( "a" )
# ["a"]
## scalar (unboxed) -> single value
to_json( 1, unbox = TRUE )
# 1.0
to_json( "a", unbox = TRUE )
# "a"
## vector -> array
to_json( 1:4 )
# [1,2,3,4]
to_json( letters[1:4] )
# ["a","b","c","d"]
## named vector - array (of the elements)
to_json( c("a" = 1, "b" = 2) )
# [1.0,2.0]
## matrix -> array of arrays (by row)
to_json( matrix(1:4, ncol = 2) )
# [[1,3],[2,4]]
to_json( matrix(letters[1:4], ncol = 2))
# [["a","c"],["b","d"]]
## matrix -> array of arrays (by column)
to_json( matrix(1:4, ncol = 2), by = "column" )
# [[1,2],[3,4]]
to_json( matrix(letters[1:4], ncol = 2 ), by = "column" )
# [["a","b"],["c","d"]]List of unnamed vectors gives an ARRAY of ARRAYS (since a vector gets converted to an array)
to_json( list( 1:2, c("a","b") ) )
# [[1,2],["a","b"]]A list with named elements gives an OBJECT with named ARRAYS
## List of vectors -> object with named arrays
to_json( list( x = 1:2 ) )
# {"x":[1,2]}A combination of named and unnamed list elements gives both
to_json( list( x = 1:2, y = list( letters[1:2] ) ) )
# {"x":[1,2],"y":[["a","b"]]}A data.frame will, by default, treat each row as an object (to maintain the relationship inherent in a row of data )
## data.frame -> array of objects (by row)
to_json( data.frame( x = 1:2, y = 3:4) )
# [{"x":1,"y":3},{"x":2,"y":4}]
to_json( data.frame( x = c("a","b"), y = c("c","d")))
# [{"x":"a","y":"c"},{"x":"b","y":"d"}]You can set by = "column" to parse the data.frame by
columns. And as each column (in this example) is a vector, each vector
gets converted to an array. And since the vectors have names (the column
names), we get an object of named arrays
## data.frame -> object of arrays (by column)
to_json( data.frame( x = 1:2, y = 3:4), by = "column" )
# {"x":[1,2],"y":[3,4]}
to_json( data.frame( x = c("a","b"), y = c("c","d") ), by = "column" )
# {"x":["a","b"],"y":["c","d"]}A data.frame where one columns is ‘AsIs’ a list
## data.frame where one colun is a list
df <- data.frame( id = 1, val = I(list( x = 1:2 ) ) )
to_json( df )
# [{"id":1.0,"val":{"x":[1,2]}}]The data.frame is being parsed ‘by row’, so we get an array of
objects. The second column is a list of a named vector, so the
val column contains an object of a named array.
Here are the individual components to show how it’s put together
## which we see is made up of
to_json( data.frame( id = 1 ) )
# [{"id":1.0}]
## and
to_json( list( x = 1:2 ) )
# {"x":[1,2]}If we take the same example and parse it ‘by column’ we get the
id column treated as a vector, but the list column remains
the same
to_json( df, by = "column" )
# {"id":[1.0],"val":{"x":[1,2]}}We can build up a more complex example with nested lists inside
columns of data.frames
df <- data.frame( id = 1, val = I(list(c(0,0))))
df
# id val
# 1 1 0, 0
to_json( df )
# [{"id":1.0,"val":[0.0,0.0]}]
df <- data.frame( id = 1:2, val = I(list( x = 1:2, y = 3:4 ) ) )
df
# id val
# x 1 1, 2
# y 2 3, 4
to_json( df )
# [{"id":1,"val":{"x":[1,2]}},{"id":2,"val":{"y":[3,4]}}]
df <- data.frame( id = 1:2, val = I(list( x = 1:2, y = 3:6 ) ), val2 = I(list( a = "a", b = c("b","c") ) ) )
df
# id val val2
# x 1 1, 2 a
# y 2 3, 4, 5, 6 b, c
pretty_json( df )
# [
# {
# "id": 1,
# "val": {
# "x": [
# 1,
# 2
# ]
# },
# "val2": {
# "a": [
# "a"
# ]
# }
# },
# {
# "id": 2,
# "val": {
# "y": [
# 3,
# 4,
# 5,
# 6
# ]
# },
# "val2": {
# "b": [
# "b",
# "c"
# ]
# }
# }
# ]
df <- data.frame( id = 1:2, val = I(list( x = 1:2, y = 3:6 ) ), val2 = I(list( a = "a", b = c("b","c") ) ), val3 = I(list( l = list( 1:3, l2 = c("a","b")), 1)) )
df
# id val val2 val3
# x 1 1, 2 a 1:3, c("....
# y 2 3, 4, 5, 6 b, c 1
pretty_json( df )
# [
# {
# "id": 1,
# "val": {
# "x": [
# 1,
# 2
# ]
# },
# "val2": {
# "a": [
# "a"
# ]
# },
# "val3": {
# "l": {
# "": [
# 1,
# 2,
# 3
# ],
# "l2": [
# "a",
# "b"
# ]
# }
# }
# },
# {
# "id": 2,
# "val": {
# "y": [
# 3,
# 4,
# 5,
# 6
# ]
# },
# "val2": {
# "b": [
# "b",
# "c"
# ]
# },
# "val3": {
# "": [
# 1.0
# ]
# }
# }
# ]Use from_json() to convert from JSON to an R object.
## scalar / vector
js <- '[1,2,3]'
from_json( js )
# [1] 1 2 3
## matrix
js <- '[[1,2],[3,4],[5,6]]'
from_json( js )
# [,1] [,2]
# [1,] 1 2
# [2,] 3 4
# [3,] 5 6
## data.frame
js <- '[{"x":1,"y":"a"},{"x":2,"y":"b"}]'
from_json( js )
# x y
# 1 1 a
# 2 2 bBy default from_json() will try and simplify
If an array contains objects with different keys, for example
'[{"x":1},{"y":2}]', from_json() will not
simplify this to a data.frame, because it would have to assume and
insert NAs in rows where data is missing.
js <- '[{"x":1},{"y":2}]'
from_json( js )
# [[1]]
# [[1]]$x
# [1] 1
#
#
# [[2]]
# [[2]]$y
# [1] 2You can override this default and use fill_na = TRUE to
force it to a data.frame with NAs in place of missing
values
js <- '[{"x":1},{"y":2}]'
from_json( js, fill_na = TRUE )
# x y
# 1 1 NA
# 2 NA 2