用模式替换R中的字符串并替换两个向量

Question

Let's say I have two vectors like so:

a <- c("this", "is", "test")
b <- c("that", "was", "boy")

I also have a string variable like so:

string <- "this is a story about a test"

I want to replace values in string so that it becomes the following:

string <- "that was a story about a boy"

I could do this using a for loop but I want this to be vectorized. How should I do this?

Answer 1

If you're open to using a non-base package, stringi will work really well here:

stringi::stri_replace_all_fixed(string, a, b, vectorize_all = FALSE)
#[1] "that was a story about a boy"

Note that this also works the same way for input strings of length > 1.

To be on the safe side, you can adapt this - similar to RUser's answer - to check for word boundaries before replacing:

stri_replace_all_regex(string, paste0("\\b", a, "\\b"), b, vectorize_all = FALSE)

This would ensure that you don't accidentally replace his with hwas , for example.

Answer 2

Here are some solutions. They each will work even if string is a character vector of strings in which case substitutions will be done on each component of it.

1) Reduce This uses no packages.

Reduce(function(x, i) gsub(paste0("\\b", a[i], "\\b"), b[i], x), seq_along(a), string)
## [1] "that was a story about a boy"

2) gsubfn gsubfn is like gsub but the replacement argument can be a list of substitutions (or certain other objects).

library(gsubfn)

gsubfn("\\w+", setNames(as.list(b), a), string)
## [1] "that was a story about a boy"

3) loop This isn't vectorized but have added for comparison. No packages are used.

out <- string
for(i in seq_along(a)) out <- gsub(paste0("\\b", a[i], "\\b"), b[i], out)
out
## [1] "that was a story about a boy"

Note: There is some question of whether cycles are possible. For example, if

a <- c("a", "A")
b <- rev(a)

do we want

• "a" to be replaced with "A" and then back to "a" again, or
• "a" and "A" to be swapped.

All the solutions shown above assume the first case. If we wanted the second case then perform the operation twice. We will illustrate with (2) because it is the shortest but the same idea applies to them all:

# swap "a" and "A"
a <- c("a", "A")
b <- rev(a)

tmp <- gsubfn("\\w+", setNames(as.list(seq_along(a)), a), string)
gsubfn("\\w+", setNames(as.list(b), seq_along(a)), tmp)
## [1] "this is A story about A test"

Answer 3

> library(stringi)
> stri_replace_all_regex(string, "\\b" %s+% a %s+% "\\b", b, vectorize_all=FALSE)
#[1] "that was a story about a boy"

Answer 4

Chipping in as well with a little function that relies only on R base :

repWords <- function(string,toRep,Rep,sep='\\s'){

  wrds <- unlist(strsplit(string,sep))
  ix <- match(toRep,wrds)
  wrds[ix] <- Rep  
  return(paste0(wrds,collapse = ' '))

}

a <- c("this", "is", "test")
b <- c("that", "was", "boy")

string <- "this is a story about a test"

> repWords(string,a,b)
[1] "that was a story about a boy"

Note:

This assumes you have a matching number of replacements. You can define the separator with sep .

Answer 5

Talking of external packages, here's another one:

a <- c("this", "is", "test")
b <- c("that", "was", "boy")
x <- "this is a story about a test"


library(qdap)
mgsub(a,b,x)

which gives:

 "that was a story about a boy"