在R中将数据帧转换为JSON

Question

I have a data frame df that I'd like to convert to json format:

df <- data.frame(`ArtId` = rep(c(50,70),each=4), 
  `Year` = rep(1990:1993,times=2), 
  `IndexBV` = c(132,94,83,101,100,100,67,133), 
  `SE` = c(20,15,12,13,NA,NA,NA,NA))

ArtId Year IndexBV SE
50    1990 132     20
50    1991  94     15
50    1992  83     12
50    1993 101     13
70    1990 100     NA
70    1991 110     NA
70    1992  67     NA
70    1993 133     NA

The data frame contains index-values (IndexBV) and their standard errors (SE) for two different species (ArtId 50 and 70) for the years 1990-1993.

When I do:

cat(jsonlite::toJSON(df, pretty=T))

I get this (only the first two elements are shown):

[
  {
    "ArtId": 50,
    "Year": 1990,
    "IndexBV": 132,
    "SE": 20
  },
  {
    "ArtId": 50,
    "Year": 1991,
    "IndexBV": 94,
    "SE": 15
  },
... 

But what I need as desired.res is a structure that looks as follows when printed with cat(jsonlite::toJSON(desired.res, pretty=T)) :

{
    "Year":["1990","1991","1992","1993"],
    "ArtId": {
        "50":[
            {"IndexBV": 132, "SE": 20},
            {"IndexBV": 94, "SE": 15},
            {"IndexBV": 83, "SE": 12},
            {"IndexBV": 101, "SE": 13}
        ],
        "70":[
            {"IndexBV": 100, "SE": NA},
            {"IndexBV": 110, "SE": NA},
            {"IndexBV": 67, "SE": NA,
            {"IndexBV": 133, "SE": NA}
        ]
    }
}

How can I convert df to desired.res ? I think the problem is about reshaping df into a nested list to take into account its nested structure.

Answer 1

I can give a very specific solution, but to generalise will take more work, as its a strange format to me. Try this out, and adapt to make more robust accordingly eg the year going from length 8 to length 4.

# I find data.table easier
library(data.table)
dt <- as.data.table(df)

# manually can do it, find levels of L, split data and name
L <- unique(df$ArtId)
ArtList <- lapply(L, function(x){
  x.dt <- dt[ArtId == x, .(IndexBV, SE)]
  x.dt
})
names(ArtList) <- L

# combine in to one list
X <- list(Year=unique(dt$Year),
          ArtId = ArtList)

# convert to Json, need either "null" or "string" for na, else dropped
jsonlite::toJSON(X, pretty = TRUE, na="null")