Identifying patterns in two strings in R - r

Question

I want to evaluate if ColA contains a new string than ColB. However, I am not interested in certain types of string, for example, oil. I would like to have an indicator variable as follow:
ColA ColB Ind
-------------------------- ------------------------ -----
coconut+grape+pine grape+coconut TRUE
orange+apple+grape+pine grape+coconut TRUE
grape+pine grape+oil TRUE
oil+grape grape+apple FALSE
grape grape+oil FALSE
grape+pine grape+orange+pine FALSE
Any Suggestions using R?
Many thanks!

Answer 1

Since we need to split the strings, we'll start with strsplit,
strsplit(dat$ColA, '+', fixed = TRUE)
# [[1]]
# [1] "coconut" "grape" "pine"
# [[2]]
# [1] "orange" "apple" "grape" "pine"
# [[3]]
# [1] "grape" "pine"
# [[4]]
# [1] "oil" "grape"
# [[5]]
# [1] "grape"
# [[6]]
# [1] "grape" "pine"
From here, we want to determine what is in ColA that is not in ColB. I'll use Map to run setdiff on each set (ColA's [[1]] with ColB's [[1]], etc).
Map(setdiff, strsplit(dat$ColA, '+', fixed = TRUE), strsplit(dat$ColB, '+', fixed = TRUE))
# [[1]]
# [1] "pine"
# [[2]]
# [1] "orange" "apple" "pine"
# [[3]]
# [1] "pine"
# [[4]]
# [1] "oil"
# [[5]]
# character(0)
# [[6]]
# character(0)
To determine which one has "new words", we can just check for non-zero length using lengths(.) > 0:
lengths(Map(setdiff, strsplit(dat$ColA, '+', fixed = TRUE), strsplit(dat$ColB, '+', fixed = TRUE))) > 0
# [1] TRUE TRUE TRUE TRUE FALSE FALSE
But since you don't care about oil, we need to remove that as well.
lapply(Map(setdiff, strsplit(dat$ColA, '+', fixed = TRUE), strsplit(dat$ColB, '+', fixed = TRUE)), setdiff, "oil")
# [[1]]
# [1] "pine"
# [[2]]
# [1] "orange" "apple" "pine"
# [[3]]
# [1] "pine"
# [[4]]
# character(0)
# [[5]]
# character(0)
# [[6]]
# character(0)
lengths(lapply(Map(setdiff, strsplit(dat$ColA, '+', fixed = TRUE), strsplit(dat$ColB, '+', fixed = TRUE)),
setdiff, "oil")) > 0
# [1] TRUE TRUE TRUE FALSE FALSE FALSE
#akrun suggested a tidyverse variant:
library(dplyr)
library(purrr) # map2_lgl
library(stringr) # str_extract_all
dat %>%
mutate(
new = map2_lgl(
str_extract_all(ColB, "\\w+"), str_extract_all(ColA, "\\w+"),
~ !all(setdiff(.y, "oil") %in% .x)
)
)
# ColA ColB Ind new
# 1 coconut+grape+pine grape+coconut TRUE TRUE
# 2 orange+apple+grape+pine grape+coconut TRUE TRUE
# 3 grape+pine grape+oil TRUE TRUE
# 4 oil+grape grape+apple FALSE FALSE
# 5 grape grape+oil FALSE FALSE
# 6 grape+pine grape+orange+pine FALSE FALSE
Data
dat <- structure(list(ColA = c("coconut+grape+pine", "orange+apple+grape+pine", "grape+pine", "oil+grape", "grape", "grape+pine"), ColB = c("grape+coconut", "grape+coconut", "grape+oil", "grape+apple", "grape+oil", "grape+orange+pine"), Ind = c(TRUE, TRUE, TRUE, FALSE, FALSE, FALSE)), class = "data.frame", row.names = c(NA, -6L))

Answer 2

Here's a solution similar to r2evans's that calls strsplit only once with the help of do.call.
rid <- function(x) x[!x %in% z] ## helper FUN to get rid of the oil
z <- "oil"
L <- sapply(unname(dat), strsplit, "\\+")
dat$ind <- sapply(1:nrow(L), function(x) length(do.call(setdiff, rev(Map(rid, L[x,]))))) > 0
dat
# V1 V2 ind
# 1 grape+coconut coconut+grape+pine TRUE
# 2 grape+coconut orange+apple+grape+pine TRUE
# 3 grape+oil grape+pine TRUE
# 4 grape+apple oil+grape FALSE
# 5 grape+oil grape FALSE
# 6 grape+orange+pine grape+pine FALSE
Data:
dat <- structure(list(V1 = c("grape+coconut", "grape+coconut", "grape+oil",
"grape+apple", "grape+oil", "grape+orange+pine"), V2 = c("coconut+grape+pine",
"orange+apple+grape+pine", "grape+pine", "oil+grape", "grape",
"grape+pine")), row.names = c(NA, -6L), class = "data.frame")