Complex lag inheritance with dplyr - r

Question

Firstly "complex lag inheritance" may not be the clearest title, so suggestions welcome. I have a large dataset of ordered segmented strings that I need to group by stem matching of segments. This looping example demonstrates the required logic:
require(tidyverse)
x = data_frame(name = c('smith', 'smith.james', 'smith.jill',
'taylor', 'taylor.ian', 'walker', 'walker.john', 'walker.john.sid',
'reed.snow', 'reed.snow.harry', 'reed.snow.helen.jane'),
family_name = NA_character_)
x$family_name[1] = x$name[1]
for(i in 2:nrow(x)){
# if current record matches previous record's family assignment..
family_match = str_detect(string = x$name[i], pattern = paste0('^', x$family_name[i-1], '[.]'))
x$family_name[i] = ifelse(family_match, x$family_name[i-1], x$name[i])
}
print(x)
#> # A tibble: 11 x 2
#> name family_name
#> <chr> <chr>
#> 1 smith smith
#> 2 smith.james smith
#> 3 smith.jill smith
#> 4 taylor taylor
#> 5 taylor.ian taylor
#> 6 walker walker
#> 7 walker.john walker
#> 8 walker.john.sid walker
#> 9 reed.snow reed.snow
#> 10 reed.snow.harry reed.snow
#> 11 reed.snow.helen.jane reed.snow
I have tried using this looping approach and it does not seem feasible given the data size, so the alternative is a vectored dplyr approach or python.
The heart of the problem is that each family_name assignment is based on match of either the current record's name (when inferring new family names), or the previous record's family_name. I don't see how to reconcile this logic with an approach using e.g. pmap_chr, but if I'm wrong I'd love to know how.

Answer 1

This uses no regular expressions or explicit loops although internally Reduce would be using a loop. No packages are used.
Names <- paste0(x$name, ".")
iter <- function(x, y) if (startsWith(y, x)) x else y
Reduce(iter, Names, acc = TRUE)
giving:
[1] "smith." "smith." "smith." "taylor." "taylor."
[6] "walker." "walker." "walker." "reed.snow." "reed.snow."
[11] "reed.snow."