R: Joining Tibbles Using Derived Column Values - r

Question

Consider the following tibbles:
library(tidyverse)
tbl_base_ids = tibble(base_id = c("ABC", "ABCDEF", "ABCDEFGHI"), base_id_length = c(3, 6, 9), record_id_length = c(10, 12, 15))
tbl_records = tibble(record_id = c("ABC1234567", "ABCDEF123456", "ABCDEFGHI123456"))
I'd like to join matching rows to produce a third tibble:
tbl_records_with_base
record_id
base_id
base_id_length
record_id_length
As you can see, this is not a matter of joining one or more variables from each of the first two. This requires matching variable derivatives. In SQL, I'd do this:
SELECT A.record_id,
B.base_id,
B.base_id_length,
B.record_id_length
FROM tbl_records A
JOIN tbl_base_ids B
ON LENGTH(a.record_id) = B.record_id_length
AND LEFT(a.record_id, B.base_id_length) = B.base_id
I've tried variations of dplyr joins and using the match function, to but to no avail. Can someone help? Thank you.

Answer 1

You should come up with some logic to separate base_id from record_id. because joining only on record_id_length would not be enough. For this example we can get base_id if we remove all numbers from record_id. Based on your actual dataset you need to change this if needed.
Once we do that we can join tbl_records with tbl_base_ids by base_id and record_id_length.
library(dplyr)
tbl_records %>%
mutate(base_id = sub('\\d+', '', record_id),
record_id_length = nchar(record_id)) %>%
inner_join(tbl_base_ids, by = c("base_id", "record_id_length")) -> result
result
# record_id base_id record_id_length base_id_length
# <chr> <chr> <dbl> <dbl>
#1 ABC1234567 ABC 10 3
#2 ABCDEF123456 ABCDEF 12 6
#3 ABCDEFGHI123456 ABCDEFGHI 15 9

Answer 2

I suggest using the fuzzyjoin package.
library(dplyr)
library(fuzzyjoin)
tbl_base_ids %>%
mutate(record_ptn = sprintf("^%s.{%i}$", base_id, pmax(0, record_id_length - base_id_length))) %>%
regex_full_join(tbl_records, ., by = c("record_id" = "record_ptn"))
# # A tibble: 3 x 5
# record_id base_id base_id_length record_id_length record_ptn
# <chr> <chr> <dbl> <dbl> <chr>
# 1 ABC1234567 ABC 3 10 ^ABC.{7}$
# 2 ABCDEF123456 ABCDEF 6 12 ^ABCDEF.{6}$
# 3 ABCDEFGHI123456 ABCDEFGHI 9 15 ^ABCDEFGHI.{6}$
A note about this: the order of tables matters, where the regex must reside on the RHS of the by= settings. For instance, this does not work if we reverse it:
tbl_base_ids %>%
mutate(record_ptn = sprintf("^%s.{%i}$", base_id, pmax(0, record_id_length - base_id_length))) %>%
regex_full_join(., tbl_records, by = c("record_ptn" = "record_id"))
# # A tibble: 6 x 5
# base_id base_id_length record_id_length record_ptn record_id
# <chr> <dbl> <dbl> <chr> <chr>
# 1 ABC 3 10 ^ABC.{7}$ <NA>
# 2 ABCDEF 6 12 ^ABCDEF.{6}$ <NA>
# 3 ABCDEFGHI 9 15 ^ABCDEFGHI.{6}$ <NA>
# 4 <NA> NA NA <NA> ABC1234567
# 5 <NA> NA NA <NA> ABCDEF123456
# 6 <NA> NA NA <NA> ABCDEFGHI123456