CODEWITHSUNDEEP
rlistdataframe
CptNemo
CptNemo

Reputation: 6755

R: Count objects in column-list

Let me define a data frame with one column id formed by a vector of integer

df <- data.frame(id = c(1,2,2,3,3))

and a column objects which instead is list of character vectors. Let''s create the column with the following function

randomObjects <- function(argument) {
  numberObjects <- sample(c(1,2,3,4), 1)
  vector <- character()
  for (i in 1:numberObjects) {
    vector <- c(vector, sample(c("apple","pear","banana"), 1))
  }
  return(vector)
}

which is then called with lapply

set.seed(28100)
df$objects <- lapply(df$id, randomObjects)

The resulting data frame is

df
#   id                 objects
# 1  1            apple, apple
# 2  2     apple, banana, pear
# 3  2                  banana
# 4  3    banana, pear, banana
# 5  3 pear, pear, apple, pear

Now I want to count the number of objects corresponding to each id with a data frame like this

summary <- data.frame(id = c(1, 2, 3),
                      apples = c(2, 1, 1), 
                      bananas = c(0, 2, 2),
                      pears = c(0, 1, 4))

summary
#   id apples bananas pears
# 1  1      2       0     0
# 2  2      1       2     1
# 3  3      1       2     4

How should I collapse the information of df into a more compact data frame such as summary without using a for loop?

Upvotes: 5

Views: 1118

Answers (4)

Colonel Beauvel
Colonel Beauvel

Reputation: 31181

An alternative way than using apply:

library(data.table)

vals = unique(do.call('c', df[,2]))

setDT(df)[,as.list(table(factor(do.call('c',objects), levels=vals))),by=id]

#   id apple banana pear
#1:  1     2      0    0
#2:  2     1      2    1
#3:  3     1      2    4

Upvotes: 1

A5C1D2H2I1M1N2O1R2T1
A5C1D2H2I1M1N2O1R2T1

Reputation: 193687

Here is a "data.table" approach:

library(data.table)
dcast.data.table(as.data.table(df)[
  , unlist(objects), by = id][
    , .N, by = .(id, V1)], 
  id ~ V1, value.var = "N", fill = 0L)
#    id apple banana pear
# 1:  1     2      0    0
# 2:  2     1      2    1
# 3:  3     1      2    4

unlist the values by ID, count them using .N, and reshape wide with dcast.data.table.

Initially, I had thought of mtabulate from "qdapTools", but that doesn't do the aggregation step. Still, you can try something like:

library(data.table)
library(qdapTools)
data.table(cbind(df[1], mtabulate(df[[-1]])))[, lapply(.SD, sum), by = id]
#    id apple banana pear
# 1:  1     2      0    0
# 2:  2     1      2    1
# 3:  3     1      2    4

Upvotes: 4

Roland
Roland

Reputation: 132999

library(plyr)

ddply(df, .(id), function(d, lev) {
  x <- factor(unlist(d$objects), levels = lev)
  t(as.matrix(table(x)))
}, lev = unique(unlist(df$objects)))
#  id apple banana pear
#1  1     2      0    0
#2  2     1      2    1
#3  3     1      2    4

Upvotes: 3

Frank
Frank

Reputation: 66819

First, aggregate to id and convert to factor

id_objs <- lapply(tapply(df$obj,df$id,unlist),factor,levels=unique(unlist(df$obj)))

Then tabulate

tab <- sapply(id_objs,table)

For your desired output, transpose the result: t(tab)

  apple banana pear
1     2      0    0
2     1      2    1
3     1      2    4

Upvotes: 1

Related Questions