`str_match` does not work well with non-capturing groups about stringr HOT 6 CLOSED

That's only need for perl compatible regexps, right?

from stringr.

Not entirely sure. The TRE POSIX 1003.2 library seems to have some bonus features. For example, the code above does appropriately capture what you would expect (i.e. just the ending group that gets the last letter), even though we use a non-capturing group to get the states that start with Ala or Mas only. Here is the version that uses regexec/regmatches:

lst.match <- regmatches(state.name, regexec("^(?:Ala|Mas).*(.)$", state.name))
lst.match[ifelse(sapply(lst.match, length) > 0, T, F)]
[[1]]
[1] "Alabama" "a"

[[2]]
[1] "Alaska" "a"

[[3]]
[1] "Massachusetts" "s"

Which works as expected (i.e. only the last letter is captured, but the non-capturing group works as it should). I haven't looked deeper in the source of str_match_*, but presumably they use regexec, so it seems this would work.

from stringr.

sorry, new to github, had no idea what "close & comment" did.

from stringr.

Just to follow up on this, it looks like the TRE ERE POSIX library definitely supports non capturing groups, in particular to specify options. See docs at: laurikari.net/tre/documentation/regex-syntax/. In particular the definition of atoms under section "Catenation of REs" which describes the following three as valid atoms:

• "(?#" comment-text ")"
• "(?" options ")" extended-regexp
• "(?" options ":" extended-regexp ")"
  What isn't explicitly mentioned is whether the above types of groups/options are capturing, but based on the results above it seems they are not. I'll admit I'm not 100% familiar with the documentation syntax, so I may be misinterpreting.

As an aside, you would have to modify my fix to just remove any "(?" (as opposed to "(?:")from consideration as an indication of the number of matches, or have some other mechanism for detecting number of matches.

from stringr.

I've just fallen over this bug too. I was using R's internal regex engine rather than the Perl engine.

The fix I found was to locate all instances of open parentheses not followed by a question mark and a colon.

Replace the line

n <- str_length(str_replace_all(tmp, "[^(]", "")) + 1

with

n <- nrow(str_locate_all(pattern, perl("\\((?!\\?:)"))[[1]]) + 1

from stringr.

Fixed in stringi branch

from stringr.