当连字符包围单个内部字符时，如何替换字内连字符

Question

I want to preserve intra-word hyphens in text prior to tokenizing it. The strategy involves substituting the hyphens for a unique character, then replacing that unique character with hyphens after tokenizing. Note: I'll ultimately use the Unicode class of Pd to catch all forms of dash character, but here I'm keeping it simple since I don't think that part is pertinent to the problem.

Problem: It fails when a word contains multiple inner hyphens separating a single character.

Examples and desired outcomes:

replaceDash <- function(x) gsub("(\\w)-(\\w)", "\\1§\\2", x)

# these are all OK
replaceDash("Hawaii-Five-O")  
## [1] "Hawaii§Five§O"
replaceDash("jack-of-all-trades")  
## [1] "jack§of§all§trades"
replaceDash("A-bomb")         
## [1] "A§bomb"
replaceDash("freakin-A")      
## [1] "freakin§A"

# not the desired outcome
replaceDash("jack-o-lantern")  # FAILS - should be "jack§o§lantern"
## [1] "jack§o-lantern"
replaceDash("Whack-a-Mole")    # FAILS - should be "Whack§a§Mole"
## [1] "Whack§a-Mole"

What regex patterns do I need for the first and second expressions of the gsub() ?

Answer 1

You can use a PCRE regex with a look-ahead that would check if a word character appears right after a hyphen, but would not consume it.

replaceDash <- function(x) gsub("(\\w)-(?=\\w)", "\\1§", x, perl=T)

See IDEONE demo

So, (\\\\w) captures an alphanumeric symbol into Group 1 that is later inserted into the replacement result with the help of \\\\1 backreference, and with (?=\\\\w) we only make sure there is a word character, but the regex index stays at the hyphen, thus allowing the next match from that word character.

Answer 2

you did not specify what kind of regex capabilities are allowed. Here is a pattern using zero-with look around:

gsub("(?<=\\w)-(?=\\w)", "§", "jack-o-lantern");
# jack§o§trade