CODEWITHSUNDEEP
rlapply
vino88
vino88

Reputation: 163

Combine Series of Columns in R

I have a dataframe where I need to combine every tenth column. As an example my dataframe looks like this:

V1    V2    V3    V4    V5    V6
A     B     C     D     E     F

What I'd like is:

first second third
AD    BE     CF

to do this I've been trying to use lapply and a custom function:

col_unite <- c(c('V1', 'V4', 'first'), c('V2', 'V5', 'second'), c('V3', 'V6', 'third'))
#function that combines columns and gives a new name
unite_cols <- function(dataframe, col_list){
  dataframe[[col_list[3]]] <- paste0(dataframe[[col_list[1]]], ', ', 
  dataframe[[col_list[2]]])
  dataframe
}

new_df <- lapply(col_unite, function(x){
   df <- unite_cols(orig_df, x)
   # df[x[3]] <- paste0(test[x[1]], ', ', test[x[2]]) # a different attempt
   return(df)
})

I know I could do this with a bunch of unite statements, or maybe there is a way to do this with a loop and unite. But given the number of columns I need to loop through I'd like to try to do it in a more elegant way.

Upvotes: 1

Views: 716

Answers (3)

G. Grothendieck
G. Grothendieck

Reputation: 270195

1) Convert the data frame to a 3d array, apply paste over its first two dimensions giving matrix m and convert back to data frame:

a <- array(as.matrix(d), c(nrow(d), 3, ncol(d)/3))
m <- apply(a, 1:2, paste, collapse = "") 
as.data.frame(m, stringsAsFactors = FALSE)

giving

  V1 V2 V3
1 ad be cf
2 gj hk il

2) Another approach if we have a grouping vector g is the following. This does not require evenly spaced columns as we can define g arbitrarily.

g <- rep(1:3, length = ncol(d))
Paste0 <- function(x) do.call("mapply", c(paste0, x))
as.data.frame(lapply(split(as.list(d), g), Paste0), stringsAsFactors = FALSE)

giving:

   a  b  c
a ad be cf
g gj hk il

Note

We used this 2x6 data frame as test input:

d <- structure(list(V1 = c("a", "g"), V2 = c("b", "h"), V3 = c("c", 
  "i"), V4 = c("d", "j"), V5 = c("e", "k"), V6 = c("f", "l")), 
  class = "data.frame", row.names = c(NA, -2L))

which looks like this:

> d
  V1 V2 V3 V4 V5 V6
1  a  b  c  d  e  f
2  g  h  i  j  k  l

Upvotes: 2

Dan
Dan

Reputation: 12084

First, I define the data frame.

df <- read.table(text = "V1    V2    V3    V4    V5    V6
A     B     C     D     E     F", header = TRUE, colClasses = "character")

Then, I use lapply to go through and paste each column with the column 3 places over and then recast the resulting list as a data frame.

lapply(1:(ncol(df)/2), function(x)paste0(df[,c(x,x+3)], collapse = "")) %>% 
  data.frame

The pipe (%>%) requires the dplyr package.

#    X.AD. X.BE. X.CF.
# 1    AD    BE    CF

Edit

Alternatively, I create a list of paired column names to combine. Much the same approach otherwise.

MyList <- list(c("V1", "V4"), c("V2", "V5"), c("V3", "V6"))

lapply(MyList, function(x)paste0(df[,x], collapse = "")) %>% 
    data.frame

#    X.AD. X.BE. X.CF.
# 1    AD    BE    CF

Upvotes: 2

Maurits Evers
Maurits Evers

Reputation: 50728

How about something like this

n <- 4
rbind.data.frame(apply(matrix(df, nrow = n - 1), 1, paste0, collapse = ""))
#   X.AD. X.BE. X.CF.
#1    AD    BE    CF

Here we combine every n = 4th entry by casting the data.frame as a matrix with n - 1 rows, paste0ing rows, and converting the resulting matrix back into a data.frame.

The only thing left is to come up with suitable column names.

Sample data

df <- read.table(text =
    "V1    V2    V3    V4    V5    V6
A     B     C     D     E     F", header = T, colClasses = "character")

Upvotes: 3

Related Questions