R strsplit with multiple unordered split arguments?

Actually strsplit uses grep patterns as well. (A comma is a regex metacharacter whereas a space is not; hence the need for double escaping the commas in the pattern argument. So the use of "\\s" would be more to improve readability than of necessity):

> strsplit(test_1, "\\, |\\,| ")  # three possibilities OR'ed
[[1]]
[1] "abc" "def" "ghi" "klm"

> strsplit(test_2, "\\, |\\,| ")
[[1]]
[1] "abc" "def" "ghi" "klm"

Without using both \\, and \\, (note extra space that SO does not show) you would have gotten some character(0) values. Might have been clearer if I had written:

> strsplit(test_2, "\\,\\s|\\,|\\s")
[[1]]
[1] "abc" "def" "ghi" "klm"

@Fojtasek is so right: Using character classes often simplifies the task because it creates an implicit logical OR:

> strsplit(test_2, "[, ]+")
[[1]]
[1] "abc" "def" "ghi" "klm"

> strsplit(test_1, "[, ]+")
[[1]]
[1] "abc" "def" "ghi" "klm"

You could go with strsplit(test_1, "\\W").

In case you don't like regular expressions, you can call strsplit() multiple times:

strsplits <- function(x, splits, ...)
{
    for (split in splits)
    {
        x <- unlist(strsplit(x, split, ...))
    }
    return(x[!x == ""]) # Remove empty values
}

strsplits(test_1, c(" ", ","))
# "abc" "def" "ghi" "klm"
strsplits(test_2, c(" ", ","))
# "abc" "def" "ghi" "klm"

Updated for the added example

strsplits(test_1, c("[[:punct:]]","[[:space:]]"))
# "abc" "def" "ghi" "klm"
strsplits(test_2, c("[[:punct:]]","[[:space:]]"))
# "abc" "def" "ghi" "klm"

But if you are going to use regular expressions, you might as well go with @DWin's approach:

strsplit(test_1, "[[:punct:][:space:]]+")[[1]]
# "abc" "def" "ghi" "klm"
strsplit(test_2, "[[:punct:][:space:]]+")[[1]]
# "abc" "def" "ghi" "klm"

 test_1<-"abc def,ghi klm"
 test_2<-"abc, def ghi klm"
 key_words <- c("abc","def","ghi")
 matches <- str_c(key_words, collapse ="|")
 str_extract_all(test_1, matches)
 str_extract_all(test_2, matches)