longest common substring in R finding non-contiguous matches between the two strings

Question

I have a question regarding finding the longest common substring in R. While searching through a few posts on StackOverflow, I got to know about the qualV package. However,

Answer 1

Leveraging @RichardScriven's insight that adist could be used, but this function combines it all,

EDIT This was tricky because we needed to get the longest_string in two contexts, so I made this function:

longest_string <- function(s){return(s[which.max(nchar(s))])}

This combines @RichardSriven's work using the library...

library(stringi)
library(stringdist)
lcsbstr <- function(a,b) { 
  sbstr_locations<- stri_locate_all_regex(drop(attr(adist(a, b, counts=TRUE), "trafos")), "M+")[[1]]
  cmn_sbstr<-stri_sub(longest_string(c(a,b)), sbstr_locations)
  longest_cmn_sbstr <- longest_string(cmn_sbstr)
   return(longest_cmn_sbstr) 
}

We can rewrite it to avoid the use of any external libraries (but still using the adist)...

lcsbstr_no_lib <- function(a,b) { 
    matches <- gregexpr("M+", drop(attr(adist(a, b, counts=TRUE), "trafos")))[[1]];
    lengths<- attr(matches, 'match.length')
    which_longest <- which.max(lengths)
    index_longest <- matches[which_longest]
    length_longest <- lengths[which_longest]
    longest_cmn_sbstr  <- substring(longest_string(c(a,b)), index_longest , index_longest + length_longest - 1)
    return(longest_cmn_sbstr ) 
}

All of identify only 'hello ' as the longest common substring, instead of 'hello r':

identical('hello ', 
    lcsbstr_no_lib('hello world', 'hello there'), 
    lcsbstr(       'hello world', 'hello there'))

EDIT And since the edit, works regardless of which argument is the longer of the two:

identical('hello',
    lcsbstr_no_lib('hello', 'hello there'), 
    lcsbstr(       'hello', 'hello there'),
    lcsbstr_no_lib('hello there', 'hello'), 
    lcsbstr(       'hello there', 'hello'))

LAST EDIT But this is only good if you accept this behavior. Notice this result:

lcsbstr('hello world', 'hello')
#[1] 'hell'

I was expecting 'hello', but since the transformation actually moves (via deletion) the world to become the hello, so only the hell part is considered a match according to the M:

drop(attr(adist('hello world', 'hello', counts=TRUE), "trafos"))
#[1] "MMMMDDDMDDD"
#[1]  vvvv   v
#[1] "hello world"

This behavior is observed using [this Levenstein tool] -- it gives two possible solutions, equivalent to these two transforms; can we tell adist which one we prefer? (the one with the greater number of consecutive M)

#[1] "MMMMDDDMDDD"
#[1] "MMMMMDDDDDD"

Finally, don't forget adist allows you to pass in ignore.case = TRUE (FALSE is the default)