Check for string in dataframe containing list of strings - r

Question

I'm trying to find a way to lookup multiple values in a dataframe and return a value. Simplified example:
df1 <- read.table(text="chk1 chk2 chk3 value
xx aa;bb;cc jj 1
xx;yy dd;ee;ff kk 2
zz gg;hh;ii ll;nn 3", header=T)
df2 <- read.table(text="val1 val2 val3
xx bb jj
xx dd kk
yy ee kk
zz hh jj
", header=T)
Lookup values val1, val2, and val3 from df2 in df1, return value from df1.
Desired results:
df2 <- read.table(text="
val1 val2 val3 value
xx bb jj 1
xx dd kk 2
yy ee kk 2
zz hh jj NA
")
Tried match x %in% y and looping over the rows, can't get it to work.

Answer 1

Here is one possibility:
library(tidyverse)
df3 <- df2 %>% rowwise %>%
mutate(rowmatch=which(grepl(val1, df1$chk1) &
grepl(val2, df1$chk2) &
grepl(val3, df1$chk3))[1],
value=df1$value[rowmatch])
Result:
# A tibble: 4 x 5
val1 val2 val3 rowmatch value
<chr> <chr> <chr> <int> <int>
1 xx bb jj 1 1
2 xx dd kk 2 2
3 yy ee kk 2 2
4 zz hh jj NA NA
Notes:
the [1] is to ensure that only first of the matching rows is used.
note that although rowmatch and value are identical in this example this is only because df1$value is equal to the row number.
tibble behaves like a data.frame, but if you really prefer a data frame, add %>% as.data.frame
The same can be done with base R and apply:
df2$rowmatch <- with(df1, apply(df2, 1, function(x)
which(grepl(x["val1"], chk1) &
grepl(x["val2"], chk2) &
grepl(x["val3"], chk3))[1]))
df2$value <- df1$value[df2$rowmatch]

Answer 2

another option would be splitting the values first:
df1 <- df1 %>%
splitstackshape::cSplit("chk1", ";", fixed = TRUE, direction = "long", drop = FALSE, type.convert = FALSE) %>%
splitstackshape::cSplit("chk2", ";", fixed = TRUE, direction = "long", drop = FALSE, type.convert = FALSE) %>%
splitstackshape::cSplit("chk3", ";", fixed = TRUE, direction = "long", drop = FALSE, type.convert = FALSE)
and then using join

Answer 3

You can also do it using two nested for loops. The logic is to take first row of df2 and then start going through rows of df1 to see if df2$val1 matches df1$chk, df2$val2 matches df1$chk2 and df2$val3 matches df1$chk3. I consider all values a match if there is at least one match per column. The caverat here is that if df2 does not have unique rows, the last matching row from df1 will be written to df2. But this can be changed by breaking out of the loop as soon as the match is found.
for (i in 1:nrow(df2)) {
for (j in 1:nrow(df1)) {
# Take i-th row and split by ;. Result is a vector of strings against
# which we'll use match.
i.split <- strsplit(as.character(unlist(df1[j, , drop = TRUE][-4])), ";")
# Pairwise check columns from df1 and df2.
all.ok <- all(mapply(FUN = function(x, y) {
any(x %in% y)
}, x = i.split, y = as.list(df2[i, 1:3])
))
if (all.ok) {
# If a match is found, write the value to df2.
df2[i, "value"] <- df1[j, "value"]
}
}
}
Output:
val1 val2 val3 value
1 xx bb jj 1
2 xx dd kk 2
3 yy ee kk 2
4 zz hh jj NA