Iterate over column names and separate fields recursively with dplyr - r

Question

I want to iterate over column names of the data frame, then using dplyr, separate fields using a delimiter(->) found among the row fields. This is how the dataset looks like :
dput(df)
structure(list(v1 = c("Silva->Mark", "Brandon->Livo", "Mango->Apple"),
v2 = c("Austin", "NA ", "Orange"),
v3 = c("James -> Jacy","NA->Jane", "apple -> Orange")),
class = "data.frame", row.names = c(NA, -3L))
Now I wrote a code that filters out column names with delimiter(->) on rows which are column v1 and column v3. Here is the code:
rows_true <- apply(df,2,function(x) any(sapply(x,function(y)grepl("->",y))))
ss<-df[,rows_true]
Then I tried to loop through those column names so that I can separate using the delimiter using this code but it ain't working
cols<- names(df)
if (names %in% df){
splitcols <- ss %>%
tidyr::separate(cols, into = c(paste0(names,+ "old"), "paste0(names,+ "New")"), sep = "->")
}
The reason I am using paste0 is because I do want the columns split into two using the delimiter then the newly formed columns should be named using the original name plus suffix Old for the first one and New for second split column
End result after looping through column names and recursively separating them should look like this
dput(df)
structure(list(v1_Old = c("Silva", "Brandon", "Mango"),
v1_New = c("Mark", "Livo", "Apple"),
v3_Old = c("James","NA", "apple"),
v3_New = c("Jacy","Jane", "Orange")),
class = "data.frame", row.names = c(NA, -3L))

Answer 1

For the sake of completeness, here is also a solution which uses data.table().
There are some differences to the other answers posted so far:
It is not required to identify the columns to be split beforehand. Instead, columns without "->" are dropped from the result on the fly.
The regular expression which is used for splitting includes surrounding white space (if any)
" *-> *". This avoids to call trimws() on the resulting pieces afterwards or to remove white space beforehand.
.
library(data.table)
library(magrittr) # piping used to improve readability
setDT(df)
lapply(names(df), function(x) {
mDT <- df[, tstrsplit(get(x), " *-> *")]
if (ncol(mDT) == 2L) setnames(mDT, paste0(x, c("_Old", "_New")))
}) %>% as.data.table()
v1_Old v1_New v3_Old v3_New
1: Silva Mark James Jacy
2: Brandon Livo NA Jane
3: Mango Apple apple Orange

Answer 2

One possibility involving dplyr and tidyr could be:
df %>%
select(v1, v3) %>%
rowid_to_column() %>%
gather(var, val, -rowid) %>%
separate_rows(val, sep = "->", convert = TRUE) %>%
group_by(rowid) %>%
mutate(val = trimws(val),
var = make.unique(var)) %>%
ungroup() %>%
spread(var, val) %>%
select(-rowid)
v1 v1.1 v3 v3.1
<chr> <chr> <chr> <chr>
1 Silva Mark James Jacy
2 Brandon Livo <NA> Jane
3 Mango Apple apple Orange
Or to further match the expected output:
df %>%
select(v1, v3) %>%
rowid_to_column() %>%
gather(var, val, -rowid) %>%
separate_rows(val, sep = "->", convert = TRUE) %>%
group_by(rowid, var) %>%
mutate(val = trimws(val),
var2 = if_else(row_number() == 2, paste0(var, "_old"), paste0(var, "_new"))) %>%
ungroup() %>%
select(-var) %>%
spread(var2, val) %>%
select(-rowid)
v1_new v1_old v3_new v3_old
<chr> <chr> <chr> <chr>
1 Silva Mark James Jacy
2 Brandon Livo <NA> Jane
3 Mango Apple apple Orange

Answer 3

A different approach with dplyr, purr, and stringr is the following.
library(dplyr)
library(purrr)
library(stringr)
# Detect the columns with at least on "->"
my_df_cols <- map_lgl(my_df, ~ any(str_detect(., "->")))
my_df %>%
# Select only the columns with at least "->"
select(which(my_df_cols)) %>%
# Mutate these columns and only keep the mutated columns with new names
transmute_all(list(old = ~ str_split(., "->", simplify = TRUE)[, 1],
new = ~ str_split(., "->", simplify = TRUE)[, 2]))
# v1_old v3_old v1_new v3_new
# 1 Silva James Mark Jacy
# 2 Brandon NA Livo Jane
# 3 Mango apple Apple Orange

Answer 4

We can also use cSplit from splitstackshape
#Detect columns with "->"
cols <- names(df)[colSums(sapply(df, grepl, pattern = "->")) > 1]
#Remove unwanted whitespaces before and after "->"
df[cols] <- lapply(df[cols], function(x) gsub("\\s+", "", x))
#Split into new columns specifying sep as "->"
splitstackshape::cSplit(df[cols], cols, sep = "->")
# v1_1 v1_2 v3_1 v3_2
#1: Silva Mark James Jacy
#2: Brandon Livo <NA> Jane
#3: Mango Apple apple Orange