agrep: only return best match(es)

问题

I'm using the 'agrep' function in R, which returns a vector of matches. I would like a function similar to agrep that only returns the best match, or best matches if there are ties. Currently, I am doing this using the 'sdist()' function from the package 'cba' on each element of the resulting vector, but this seems very redundant.

/edit: here is the function I'm currently using. I'd like to speed it up, as it seems redundant to calculate distance twice.

library(cba)
word <- 'test'
words <- c('Teest','teeeest','New York City','yeast','text','Test')
ClosestMatch <- function(string,StringVector) {
  matches <- agrep(string,StringVector,value=TRUE)
  distance <- sdists(string,matches,method = "ow",weight = c(1, 0, 2))
  matches <- data.frame(matches,as.numeric(distance))
  matches <- subset(matches,distance==min(distance))
  as.character(matches$matches)
}

ClosestMatch(word,words)

回答1:

RecordLinkage package was removed from CRAN, use stringdist instead:

library(stringdist)

ClosestMatch2 = function(string, stringVector){

  stringVector[amatch(string, stringVector, maxDist=Inf)]

}

回答2:

The agrep package uses Levenshtein Distances to match strings. The package RecordLinkage has a C function to calculate the Levenshtein Distance, which can be used directly to speed up your computation. Here is a reworked ClosestMatch function that is around 10x faster

library(RecordLinkage)

ClosestMatch2 = function(string, stringVector){

  distance = levenshteinSim(string, stringVector);
  stringVector[distance == max(distance)]

}

来源：https://stackoverflow.com/questions/5721883/agrep-only-return-best-matches