Pfthb

Question:

I have a particular problem where I want to subset a given dataframe columnwise where the column names are stored in another dataframe.

Example using mtcars dataset:

options(stringsAsFactors = FALSE)

col_names <- c("hp,disp", "disp,hp,mpg")
df_col_names <- as.data.frame(col_names)

vec <- df_col_names[1,] # first row contains "hp" and "disp"
mtcars_new <- mtcars[, c("hp", "disp")] ## assuming that vec gives colnames

I even tried inserting double quotes to each of the words using the following:

Attempted solution:

options(stringsAsFactors = FALSE)

col_names <- c("hp,disp", "disp,hp,mpg")
df_col_names <- as.data.frame(col_names)

df_col_names$col_names <- gsub("(\w+)", '"\1"', df_col_names$col_names)
vec <- df_col_names[1,]
vec2 <- gsub("(\w+)", '"\1"', vec)

mtcars_new <- mtcars[,vec2] ## this should be same as mtcars[, c("hp", "disp")]

Expected Solution

mtcars_new <- mtcars[,vec2] is equal to mtcars_new <- mtcars[, c("hp", "disp")]

Do you need this?

lapply(strsplit(as.character(df_col_names$col_names), ","), function(x) mtcars[x])

#[[1]]
# hp disp
#Mazda RX4 110 160.0
#Mazda RX4 Wag 110 160.0
#Datsun 710 93 108.0
#Hornet 4 Drive 110 258.0
#Hornet Sportabout 175 360.0
#.....

#[[2]]
# disp hp mpg
#Mazda RX4 160.0 110 21.0
#Mazda RX4 Wag 160.0 110 21.0
#Datsun 710 108.0 93 22.8
#Hornet 4 Drive 258.0 110 21.4
#Hornet Sportabout 360.0 175 18.7
#....

Here, we split the column names on comma (",") and then subset it from the dataframe using lapply. This returns a list of dataframes with length of list which is same as number of rows in the data frame.

If you want to subset only the first row, you could do

mtcars[strsplit(as.character(df_col_names$col_names[1]), ",")[[1]]]

Here's another way to do this:

col_names <- c("hp,disp", "disp,hp,mpg")

vec2 <- unlist(str_split(col_names[[1]],','))
mtcars_new <- mtcars[,vec2]

What you are doing is picking the first element from the col_names vector, splitting it by the separator, then unlisting it (because str_split() makes a list), then you are using your new vector of names to subset the mtcars data-frame.