R - Merging two dataframe by text - r

Question

I have two datasets which I want to merge :
df1 <- data.frame( title =
c("residence mozart",
"les hesperides auteuil mirabeau",
"chaillot",
"jouvenet",
"retraite dosne"))
df2 <- data.frame(title = c("terrasses mozart", "chaillot",
"villa jules janin", "retraites dosne"))
And I would like to have something like this :
1 residence mozart NA (or terrasses mozart)
2 les hesperides auteuil mirabeau NA
3 chaillot chaillot
4 jouvenet NA
5 retraite dosne retraites dosne
Here is what I did :
x = data.frame(title_df2 = matrix(ncol = 1, nrow = nrow(df1)))
for (i in nbr){
x[i, ] <- grep(df1$title[i], df2$title, value = T)
}
It does not work at all ! Even though grep(df1$title[5], df2$title, value = T) works and return "chaillot"!

Answer 1

If I understand correctly
df1 <- data.frame( title =
c("residence mozart",
"les hesperides auteuil mirabeau",
"chaillot",
"jouvenet",
"retraite dosne"))
df2 <- data.frame(title = c("terrasses mozart", "chaillot",
"villa jules janin", "retraites dosne"))
library(dplyr)
library(fuzzyjoin)
stringdist_left_join(x = df1, y = df2, method = "jw", distance_col = "d") %>%
filter(d < 0.25) %>%
right_join(df1, by = c("title.x" = "title"))
#> Joining by: "title"
#> title.x title.y d
#> 1 residence mozart terrasses mozart 0.23863636
#> 2 chaillot chaillot 0.00000000
#> 3 retraite dosne retraites dosne 0.09206349
#> 4 les hesperides auteuil mirabeau <NA> NA
#> 5 jouvenet <NA> NA
Created on 2021-04-19 by the reprex package (v2.0.0)

Answer 2

The issue is that grep returns a vector of length 0 when there is no match.
grep('a', 'hello', value = TRUE)
#character(0)
If we want to make use of the same for loop, make an adjustment in the code to return NA whereever there is no match
nbr <- seq_len(nrow(df1))
for (i in nbr){
x[i, ] <- c(grep(df1$title[i], df2$title, value = TRUE), NA_character_)[1]
}
-output
x
# title_df2
#1 <NA>
#2 <NA>
#3 chaillot
#4 <NA>
#5 <NA>

Answer 3

You could do:
a <-Vectorize(agrep, "pattern")(df1$title, df2$title, value=TRUE)
is.na(a)<- lengths(a) == 0
cbind(df1,df2_title=unlist(a, use.names = FALSE))
title df2_title
1 residence mozart <NA>
2 les hesperides auteuil mirabeau <NA>
3 chaillot chaillot
4 jouvenet <NA>
5 retraite dosne retraites dosne

Answer 4

To achieve your goal, you need a matching on each word of your strings within df1 title.
As used in your example, Grep will return an output only if there is a match on the full string.
In order to do that, you'll need to grep on possible words on df1 that are also contained in df2. This can be achieved by implementing an or condition on the full word contained in each string.
nbr <- 1:nrow(x)
for (i in nbr){
pattern <- paste("\\b",unlist(strsplit(as.character(df1$title[i]), " ")), "\\b", collapse = "|", sep = "") # here you create a regex expression whereby you can check if one of the words contained in 1 is also in df2. the \\b \\b escape makes sure that there is a full match on the single word.
fitInDataFrame <- grep(pattern, as.character(df2$title), value = T) # here you grep on the constructed regex expression
x[i, ] <- ifelse(length(fitInDataFrame) == 0, NA, fitInDataFrame)
}
Here the output:
> x
title_df2
1 terrasses mozart
2 <NA>
3 chaillot
4 <NA>
5 retraites dosne

Answer 5

You can do a left_join(df1, df2, by = c('title' = 'title'), keep = TRUE), specifying keep = TRUE so it doesn't drop df2's join column.
Or, for this particular case, you could do this:
df1$newcol <- ifelse(df1$title %in% df2$title, df1$title, NA)
This adds a new column to df1 which is filled out by going through each title in df1, checking if that title is in df2, if so writing that title in the second column and if not writing NA in that row of the second column. You could choose to put something else there instead, like:
df1$newcol <- ifelse(df1$title %in% df2$title, 'Title in DF2', 'Not in DF2')