How do I lock the first digits of the 'by' column in a stringdist join?-CodePudding

I am trying to use stringdist_join to merge two tables. I have built my 'by' variable as the concatenation of three variables which are named as such:

UAI : a serial number nom : surname prenom : name

The code below works well, however I'd like to have a perfect match on the UAI part which is always the first 8 characters of the variable UAInomprenom. How can I do that?

stringdist_join(Ech_final_nom, BSA_affect_nom, 
                by = "UAInomprenom",
                mode = "left",
                ignore_case = FALSE, 
                method = "jw", 
                max_dist = 0.1117, 
                distance_col = "dist")

Thank you for your help!

CodePudding user response：

I am taking the following two datasets as an example:

df1 <- structure(list(V1 = c("abcNum1Num1Num1Num1", "abc1Num1Num1Num1Num", 
"accArv", "accbrf"), V2 = c(1L, 4L, 5L, 2L)), class = "data.frame", row.names = c(NA, 
-4L))

df2 <- structure(list(V1 = c("abcNun1Nun1Nun1Nun1", "abc1Nun1Nun1Nun1Nun", 
"accArv", "accNun1Nun1Nun1Nun1"), V2 = c(2L, 5L, 4L, 1L)), class = "data.frame", row.names = c(NA, 
-4L))

In these two dataframes, the variable V1 is the join by field, in which the 3 first characters are not fuzzy (in your case, there are 8 not fuzzy characters).

Now, separate the column V1 to have an isolated column with the referred 3 first characters:

library(fuzzyjoin)
library(tidyverse)

df1 <- df1 %>% 
  extract(V1, into = c("V1A","V1B"), "(.{3})(.*)")

df2 <- df2 %>% 
  extract(V1, into = c("V1A","V1B"), "(.{3})(.*)")

Finally, apply the fuzzy join and remove the rows where the values of the two columns with the 3-character field are different:

stringdist_join(df1, df2, 
                by = "V1B",
                mode = "left",
                ignore_case = FALSE, 
                method = "jw", 
                max_dist = 0.5) %>% 
  filter(V1A.x == V1A.y) %>% 
  unite("V1",c("V1A.x","V1B.x"),sep="") %>% 
  select(V1,V2=V2.x,V3=V2.y)