How to remove unbalanced/unpartnered double quotes (in Java)

Question

I thought to share this relatively smart problem with everyone here. I am trying to remove unbalanced/unpaired double-quotes from a string.

My work is in progress, I might be close to a solution. But, I didn't get a working solution yet. I am not able to delete the unpaired/unpartnered double-quotes from the string.

Example Input

string1=injunct! alter ego."
string2=successor "alter ego" single employer"  "proceeding "citation assets"

Output Should be

string1=injunct! alter ego.
string2=successor "alter ego" single employer  proceeding "citation assets"

This problem sound similar to Using Java remove unbalanced/unpartnered parenthesis

Here is my code so far(it doesn't delete all the unpaird double-quotes)

private String removeUnattachedDoubleQuotes(String stringWithDoubleQuotes) {
    String firstPass = "";

    String openingQuotePattern = "\\\"[a-z0-9\\p{Punct}]";
    String closingQuotePattern = "[a-z0-9\\p{Punct}]\\\"";

    int doubleQuoteLevel = 0;
    for (int i = 0; i < stringWithDoubleQuotes.length() - 3; i++) {
        String c = stringWithDoubleQuotes.substring(i, i + 2);
        if (c.matches(openingQuotePattern)) {
            doubleQuoteLevel++;
            firstPass += c;
        }
        else if (c.matches(closingQuotePattern)) {
            if (doubleQuoteLevel > 0) {
                doubleQuoteLevel--;
                firstPass += c;
            }
        }
        else {
            firstPass += c;
        }
    }

    String secondPass = "";
    doubleQuoteLevel = 0;
    for (int i = firstPass.length() - 1; i >= 0; i--) {
        String c = stringWithDoubleQuotes.substring(i, i + 2);
        if (c.matches(closingQuotePattern)) {
            doubleQuoteLevel++;
            secondPass = c + secondPass;
        }
        else if (c.matches(openingQuotePattern)) {
            if (doubleQuoteLevel > 0) {
                doubleQuoteLevel--;
                secondPass = c + secondPass;
            }
        }
        else {
            secondPass = c + secondPass;
        }
    }

    String result = secondPass;

    return result;
}

Answer 1

It could probably be done in a single regex if there is no nesting.
There is a concept of delimeters roughly defined, and it is possible to 'bias'
those rules to get a better outcome.
It all depends on what rules are set forth. This regex takes into account
three possible scenario's in order;

1. Valid Pair
2. Invalid Pair (with bias)
3. Invalid Single

It also doesen't parse "" beyond end of line. But it does do multiple
lines combined as a single string. To change that, remove \n where you see it.

---

global context - raw find regex
shortened

(?:("[a-zA-Z0-9\p{Punct}][^"\n]*(?<=[a-zA-Z0-9\p{Punct}])")|(?<![a-zA-Z0-9\p{Punct}])"([^"\n]*)"(?![a-zA-Z0-9\p{Punct}])|")

replacement grouping

$1$2 or \1\2

Expanded raw regex:

(?:                            // Grouping
                                  // Try to line up a valid pair
   (                                 // Capt grp (1) start 
     "                               // "
      [a-zA-Z0-9\p{Punct}]              // 1 of [a-zA-Z0-9\p{Punct}]
      [^"\n]*                           // 0 or more non- [^"\n] characters
      (?<=[a-zA-Z0-9\p{Punct}])         // 1 of [a-zA-Z0-9\p{Punct}] behind us
     "                               // "
   )                                 // End capt grp (1)

  |                               // OR, try to line up an invalid pair
       (?<![a-zA-Z0-9\p{Punct}])     // Bias, not 1 of [a-zA-Z0-9\p{Punct}] behind us
     "                               // "
   (  [^"\n]*  )                        // Capt grp (2) - 0 or more non- [^"\n] characters
     "                               // "
       (?![a-zA-Z0-9\p{Punct}])      // Bias, not 1 of [a-zA-Z0-9\p{Punct}] ahead of us

  |                               // OR, this single " is considered invalid
     "                               // "
)                               // End Grouping

Perl testcase (don't have Java)

$str = '
string1=injunct! alter ego."
string2=successor "alter ego" single employer "a" free" proceeding "citation assets"
';

print "\n'$str'\n";

$str =~ s
/
  (?:
     (
       "[a-zA-Z0-9\p{Punct}]
        [^"\n]*
        (?<=[a-zA-Z0-9\p{Punct}])
       "
     )
   |
       (?<![a-zA-Z0-9\p{Punct}])
       " 
     (  [^"\n]*  )
       " (?![a-zA-Z0-9\p{Punct}])
   |
       "
  )
/$1$2/xg;

print "\n'$str'\n";

Output

'
string1=injunct! alter ego."
string2=successor "alter ego" single employer "a" free" proceeding "citation assets"
'

'
string1=injunct! alter ego.
string2=successor "alter ego" single employer "a" free proceeding "citation assets"
'

Answer 2

You could use something like (Perl notation):

s/("(?=\S)[^"]*(?<=\S)")|"/$1/g;

Which in Java would be:

str.replaceAll("(\"(?=\\S)[^\"]*(?<=\\S)\")|\"", "$1");