How to parse JSON in a DataFrame column using R

Question

How do I come from here ...

| ID | JSON Request                                                          |
==============================================================================
|  1 | {"user":"xyz1","weightmap": {"P1":0,"P2":100}, "domains":["a1","b1"]} |
------------------------------------------------------------------------------
|  2 | {"user":"xyz2","weightmap": {"P1":100,"P2":0}, "domains":["a2","b2"]} |
------------------------------------------------------------------------------

to here (The requirement is to make a table of JSON in column 2):

| User | P1 | P2 | domains | 
============================
| xyz1 |  0 |100 | a1, b1  |
----------------------------
| xyz2 |100 | 0  | a2, b2  |
----------------------------

Here is the code to generate the data.frame:

raw_df <- 
  data.frame(
    id   = 1:2,
    json = 
      c(
        '{"user": "xyz2", "weightmap": {"P1":100,"P2":0}, "domains": ["a2","b2"]}', 
        '{"user": "xyz1", "weightmap": {"P1":0,"P2":100}, "domains": ["a1","b1"]}'
      ), 
    stringsAsFactors = FALSE
  )

Answer 1

Here's a tidyverse solution (also using jsonlite) if you're happy to work in a long format (for domains in this case):

library(jsonlite)
library(dplyr)
library(purrr)
library(tidyr)

d <- data.frame(
  id = c(1, 2),
  json = c(
    '{"user":"xyz1","weightmap": {"P1":0,"P2":100}, "domains":["a1","b1"]}',
    '{"user":"xyz2","weightmap": {"P1":100,"P2":0}, "domains":["a2","b2"]}'
  ),
  stringsAsFactors = FALSE
)

d %>% 
  mutate(json = map(json, ~ fromJSON(.) %>% as.data.frame())) %>% 
  unnest(json)
#>   id user weightmap.P1 weightmap.P2 domains
#> 1  1 xyz1            0          100      a1
#> 2  1 xyz1            0          100      b1
#> 3  2 xyz2          100            0      a2
#> 4  2 xyz2          100            0      b2

• mutate... is converting from a string to column of nested data frames.
• unnest... is unnesting these data frames into multiple columns

Answer 2

I would go for the jsonlite package in combination with the usage of mapply , a transformation function and data.table's rbindlist .

# data 
raw_df <- data.frame(id = 1:2, json = c('{"user": "xyz2", "weightmap": {"P1":100,"P2":0}, "domains": ["a2","b2"]}', '{"user": "xyz1", "weightmap": {"P1":0,"P2":100}, "domains": ["a1","b1"]}'), stringsAsFactors = FALSE)

# libraries
library(jsonlite)
library(data.table)


# 1) First, make a transformation function that works for a single entry
f <- function(json, id){
  # transform json to list
  tmp    <- jsonlite::fromJSON(json)

  # transform list to data.frame
  tmp    <- as.data.frame(tmp)

  # add id
  tmp$id <- id

  # return
  return(tmp)
}


# 2) apply it via mapply 
json_dfs <- 
  mapply(f, raw_df$json, raw_df$id, SIMPLIFY = FALSE)


# 3) combine the fragments via rbindlist
clean_df <- 
  data.table::rbindlist(json_dfs)

# 4) et-voila
clean_df
##    user weightmap.P1 weightmap.P2 domains id
## 1: xyz2          100            0      a2  1
## 2: xyz2          100            0      b2  1
## 3: xyz1            0          100      a1  2
## 4: xyz1            0          100      b1  2

Answer 3

Could not get the flatten parameter to work as I expected so needed to unlist and then "re-list" before rbinding with do.call:

library(jsonlite)
 do.call( rbind, 
          lapply(raw_df$json, 
                  function(j) as.list(unlist(fromJSON(j, flatten=TRUE)))
        )       )
     user   weightmap.P1 weightmap.P2 domains1 domains2
[1,] "xyz2" "100"        "0"          "a2"     "b2"    
[2,] "xyz1" "0"          "100"        "a1"     "b1"    

Admittedly, this will require further processing since it coerces all the lines to character.

Answer 4

library(jsonlite)
json = c(
        '{"user":"xyz1","weightmap": {"P1":0,"P2":100}, "domains":["a1","b1"]}',
        '{"user":"xyz2","weightmap": {"P1":100,"P2":0}, "domains":["a2","b2"]}'
        )

json <- lapply( paste0("[", json ,"]"), 
                function(x) jsonlite::fromJSON(x))

df <- data.frame(matrix(unlist(json), nrow=2, ncol=5, byrow=T))

df <- df %>% unite(Domains, X4, X5, sep = ", ") 
colnames(df) <- c("user", "P1", "P2", "domains")
head(df)

The output is:

  user  P1  P2 domains
1 xyz1   0 100  a1, b1
2 xyz2 100   0  a2, b2

Answer 5

Using tidyjson https://cran.r-project.org/web/packages/tidyjson/vignettes/introduction-to-tidyjson.html

install.packages("tidyjson")

library(tidyjson)

json_as_df <- raw_df$json %>% spread_all

# retain columns
json_as_df <- raw_df %>% as.tbl_json(json.column = "json") %>% spread_all