0

I have a list of more than 100,000 json files from which I want to get a data.table with only a few variables. Unfortunately the files are complex. The content of each json file looks like:

Sample 1

$id
[1] "10.1"
$title
$title$value
[1] "Why this item"
$itemsource
$itemsource$id
[1] "AA"
$date
[1] "1992-01-01"
$itemType
[1] "art"
$creators
list()

Sample 2

$id
[1] "10.2"
$title
$title$value
[1] "We need this item"
$itemsource
$itemsource$id
[1] "AY"
$date
[1] "1999-01-01"
$itemType
[1] "art"
$creators
    type                name firstname    surname affiliationIds
1 Person Frank W. Cornell.  Frank W. Cornell.             a1
2 Person David A. Chen.  David A. Chen.             a1

$affiliations
  id                                          name
1 a1 Foreign Affairs Desk, New York Times

What I need from this set of files is a table with creator names, item ids and dates. For the two sample files above:

id           date            name                firstname lastname  creatortype
"10.1"      "1992-01-01"      NA                    NA        NA      NA
"10.2"      "1999-01-01"  Frank W. Cornell.      Frank W.   Cornell.  Person
"10.2"      "1999-01-01"  David A. Chen.         David A.   Chen.     Person

What I have done so far:

library(parallel)
library(data.table)
library(jsonlite)
library(dplyr)

filelist = list.files(pattern="*.json",recursive=TRUE,include.dirs =TRUE)
parsed = mclapply(filelist, function(x) fromJSON(x),mc.cores=24)
data = rbindlist(mclapply(1:length(parsed), function(x) { 
  a = data.table(item = parsed[[x]]$id, date = list(list(parsed[[x]]$date)), name = list(list(parsed[[x]]$name)), creatortype = list(list(parsed[[x]]$creatortype))) #ignoring the firstname/lastname fields here for convenience
  b = data.table(id = a$item, date = unlist(a$date), name=unlist(a$name), creatortype=unlist(a$creatortype))
  return(b)
},mc.cores=24))

However, on the last step, I get this error: 

"Error in rbindlist(mclapply(1:length(parsed), function(x){:
Item 1 of list is not a data.frame, data.table or list"

Thanks in advance for your suggestions. Related questions include: Extract data from list of lists [R] R convert json to list to data.table I want to convert JSON file into data.table in r How can read files from directory using R? Convert R data table column from JSON to data table

Community
  • 1
  • 1
user2627717
  • 344
  • 3
  • 14

1 Answers1

1

from the error message, i suppose this basically means that one of the results from mclapply() is empty, by empty I mean either NULL or data.table with 0 row, or simply encounters an error within the parallel processing.

what you could do is:

  1. add more checks inside the mclapply() like try-error or check the class of b and nrow of b, whether b is empty or not

  2. when you use rbindlist, add argument fill = T

hope this solves ur problem.

sinalpha
  • 383
  • 1
  • 9