Java regular expression - cater for an inclusive comma

Question

I'm trying to split a comma delimited pairs string but can't work out how to cater for an inclusive comma.

Here is my test case -

private void stringSplit() {
   String value = "{aaa=1111,bbb=2222,ccc=3333}";
   String regEx = "[^,]+=[^,]+";
   String separator = "=";
   final Pattern pattern = Pattern.compile(regEx);
   final Matcher matcher = pattern.matcher(value);
   while (matcher.find()) {
      final String group = matcher.group();
      final String key = group.substring(0, group.indexOf(separator));
      final String val =
         group.substring(
            group.indexOf( separator ) + separator.length(),
            group.length());
      System.out.println("key [" + key + "], val [" + val + "]");
   }
}

and here are my results -

key [{aaa], val [1111]
key [bbb], val [2222]
key [ccc], val [3333}]

All good so far...

But there may be a comma in the numeric value i.e.

"{aaa=11,11,bbb=2222,ccc=333,3}";

the results I would want are -

key [{aaa], val [11,11]
key [bbb], val [2222]
key [ccc], val [333,3}]

Could any of you regular expression guru's help me out here.

thanks!

EDIT

Following on from @bmorris591 further comments.

Ok, I have a final query - and this is a definitive list of what this crazy regex (+ a bit of java code) needs to handle.

Here is my code -

private void stringSplit() {
    String value = "{1=\"1, one\", 22=\"+t,w,o\", 333=\"three, \"3\", -33,,333,\", 4444=\"four. '4-4, (44),  -44\"}, 555=\"\", \"666\"=6666, \"777\"=\"7777\"}";
    String regex = "[^\\{,]+=([[\\w]\\(\\)\\-\\+\\.\"'\\s,]+)[,}]";
    String separator = "=";
    final Pattern pattern = Pattern.compile(regex);
    final Matcher matcher = pattern.matcher(value);
    while (matcher.find()) {
        final String group = matcher.group();
        showKeyAndValue(group, separator);
    }
}

private void showKeyAndValue(final String group, final String keyValueSeparator) {
    System.out.println("group [" + group + "]");
    final String key = removeQuotesFromString(group.substring(0,
            group.indexOf(keyValueSeparator)));
    final String val = removeQuotesFromString(group.substring(
            group.indexOf(keyValueSeparator)
                    + keyValueSeparator.length(), group.length()));
    System.out.println("key [" + key + "], val [" + val + "]");
}

private String removeQuotesFromString(final String str) {
    String returnString = str.trim();
    if (returnString.startsWith("\"")) {
        returnString = returnString.substring(
                returnString.indexOf("\"") + 1, returnString
                        .lastIndexOf("\""));
    }
    return returnString;
}

And here are the results -

group [1="1, one",]
key [1], val [1, one]
group [ 22="+t,w,o",]
key [22], val [+t,w,o]
group [ 333="three, "3", -33,,333,",]
key [333], val [three, "3", -33,,333,]
group [ 4444="four. '4-4, (44),  -44"}]
key [4444], val [four. '4-4, (44),  -44]
group [ 555="",]
key [555], val []
group [ "666"=6666,]
key [666], val [6666,]
group [ "777"="7777"}]
key [777], val [7777]

All results are correct apart from key 666. As you can see there is a trailing comma. Now I could just strip this off (for a value that is not encapsulated in quotes (basically a number)) but I was wondering if this could be acheived in the regex as this would be a 'cleaner' solution...

Many, many thanks if you can think of anything.

Answer 1

You can use the magic of negative lookahead, to split your strings on a comma not followed by a digit use

public static void main(String[] args) {
    final String s = "{aaa=11,11,bbb=2222,ccc=333,3}";
    final String[] ss = s.substring(1, s.length() -1).split(",(?!\\d)");
    for(final String str : ss) {
        System.out.println(str);
    }
}

Output

aaa=11,11
bbb=2222
ccc=333,3

You can easily expand this to yank the key=value pairs directly

public static void main(String[] args) {
    final String s = "{aaa=11,11,bbb=2222,ccc=333,3}";
    final Pattern p = Pattern.compile("([A-Za-z]++)=([\\d,]+)(?!\\d)[,}]");
    final Matcher matcher = p.matcher(s);
    while (matcher.find()) {
        System.out.println(matcher.group(1));
        System.out.println(matcher.group(2));
        System.out.println("DONE");
    }
}

Output

aaa
11,11
DONE
bbb
2222
DONE
ccc
333,3
DONE

EDIT

Following the OP's comment:

The value part of the pair is alphanumeric (including ,+-*/=()), also the value is always encapsulated in quotes, there could be multiple occurances of ,+-*/=() too...

I have revised the expression:

public static void main(String[] args) {
    final String s = "{1=\"1, one\", 22=\"+t,w,o\", 333=\"three, 3, -33,,333\", 4444=\"four. 4-4, (44), -44\"}";
    System.out.println("String is: " + s);
    final Pattern p = Pattern.compile("([^{=,\\s]++)=\"([^\"]++)\"");
    final Matcher matcher = p.matcher(s);
    while (matcher.find()) {
        System.out.println(matcher.group(1));
        System.out.println(matcher.group(2));
        System.out.println("DONE");
    }

Output:

String is: {1="1, one", 22="+t,w,o", 333="three, 3, -33,,333", 4444="four. 4-4, (44), -44"}
1
1, one
DONE
22
+t,w,o
DONE
333
three, 3, -33,,333
DONE
4444
four. 4-4, (44), -44
DONE

The pattern will now match anything that is not =,{ or whitespace followed by an = and then followed by any pattern not containing " encapsulated in ". Does this help?

Answer 2

bmorris591

Thanks for your reply.

Sorry, but looking back my original post was a little too simplistic.

The value part of the pair is alphanumeric (including ",+-*/=()"), also the value is always encapsulated in quotes, there could be multiple occurances of ",+-*/=()" too...

i.e.

"{1=\"1 one\", 22=\"two\", 333=\"three 3\"}"
"{1=\"1, one\", 22=\"+t,w,o\", 333=\"three, 3, -33,,333\", 4444=\"four. 4-4, (44), -44\"}"

Because of the complexity of this I think the most simple solution is to replace all occurances of comma with some marker character before the pair string is constructed, do the regex and then re-apply the comma to the value...

Thank you for your reply to my initial post though as it is a solution to my original question...