Regex match with lookbehind and lookahead with named groups

Question

I'm trying to match on the following text:

"abc" matches "b" and field[cba] = "cba" or (field[cba] matches "c") and "cc" = "bb"

the parts before and after "matches" into named groups.

I need to match "abc" as ${left} and "b" as ${right} , and then "field[cba]" / "c" on the second match.

I need to give bounds to ${left} and ${right} so that they break if:

Left:

• should be preceded by any of: " and " , " or " , "(" when not in doublequotes (")
• if none of those is present then it could be the start of the string

Right:

• should be followed by any of: " and " , " or " , ")" when not in doublequotes (")
• if none of those is present then it could be the end of the string

The replacement regex pattern I would like to use is:

RegExpMatch(${left}, ${right})

So to get the following output:

RegExpMatch("abc","b") and field[cba] = "cba" or (RegExpMatch(field[cba],"c")) and "cc" = "bb"

I tried with:

(?<=^|\\(| or | and )(?<left>.*?) matches (?<right>.*?)(?=\\)|$| and | or )

This has a couple of issues:

• using ^ for start of string seems to make the lookbehind greedy and it captures from start of string even if there is an " or " or " and " before, which is weird because $ seems to work ok
• I don't know how to tell the " or " , " and " , "(" or ")" to match only when not in quotes (in a literal)

Can you please help me in figuring out the correct regular pattern to apply?

Answer 1

The problem is it sees and in your lookahead, and then you use .*? (which will suck up everything until matches : field[cba] = "cba" or (field[cba] ). We need a more strict definition of left/right, it can't just be "any character".

(?<=^|\(| or | and )(?<left>\S+) matches (?<right>\S+?)(?=\)|$| and | or )

I changed .*? to \\S+ which matches anything but whitespace ( [^\\r\\n\\t\\f ] ). Now it won't suck up all the unnecessary characters in left/right capture groups. \\S+ may not be the right definition for you, but it should get you started.

Demo: Regex101

Answer 2

I'm not entirely sure how your data is, but I suggest this regex, which is independent of the bounds:

(?:(?<left>"[^"]*")|\b(?<left>\S*)) matches (?:(?<right>"[^"]*")|(?<right>\S*[^)\s]))

I'm exploiting the fact that C# allows captures with the same name here. The left and right parts are almost the same.

(?:            => Non-capture group
  (?<left>     => Left capture begin
    "[^"]*"    => Double quotes, non-quote characters then double quotes
  )            => End left capture 
|              => OR
  \b           => Word boundary
  (?<left>     => Begin other left capture if first failed
    \S*        => Capture non-space characters (if your parts break on multiple lines, you can use [^"]* instead
  )            => End left capture
)              => End non-capture group

regex101 demo (I changed the named captures because PCRE doesn't support same name capture groups)

If the word boundary is causing problems (eg when you have a part that doesn't start with " or a \\w character, you might use the following regex instead:

(?:(?<left>"[^"]*")|\s\(?(?<left>\S*)) matches (?:(?<right>"[^"]*")|(?<right>\S*[^)\s]))

Which is using \\s\\(? instead of the \\b

---

If you want to stick to the bounds you mentioned, you will have to know what exactly can be in the parts or what cannot. For instance, if

field["abc"] in field matches field["cba"] in field

is valid and the parts are field["abc"] in field and field["cba"] in field respectively, then it's another complication.