It also looks like double comma before a ""  like this ,,"" -- the ,,
counts as 1 token

I do not believe that regular expressions (even the weird "enhanced" regular
expressions that Java has apparently borrowed from Perl) are a suitable tool
for this job.  Even if you could find some way of expressing the complex logic
of CSV parsing using the extended regexps, your resulting code would be totally
incomprehensible.

A useful test case for your parser is that /any/ CSV file can (treated as
a single string) be used as the contents of a field within another CSV.
Embedded newlines, arbitrarily deep nesting, etc, are all allowed.  That power
hardly makes it any harder to parse CVS /unless/ you've started out by relying
on an inadequate tool like regexps.

Also, don't forget to test your code with files generated from Excel in places
where the numeric thousands separator is not ','.

   -- chris

P.S.  Although I don't much like Java's regexps (I think they are /grossly/
overused), it seems that I like them rather better than OE's spellchecker,
which suggests that replace "regexp" with "rage" and "regexps" with "creeps".
True AI is clearly only just around the corner ;-)

Cross posting is fine, but you should add a follow up header.  Which I
have (to comp.lang.java.programmer)

I'm not sure that a regex is good enough to do everything...

Anyway, here are the cases that I can think of, ignoring the
delimiters.

Field value:
Field value: ,
Field value: "
Field value: a,b
Field value: "a and b"
Field value: 6"3

What are the encodings of this?

I'm guessing that

Field value:
Encoded value:
Field value: ,
Encoded value: ","
Field value: "
Encoded value: ""
Field value: a,b
Encoded value: "a,b"
Field value: "a and b"
Encoded value: ""a and b""
Field value: 6"3
Encoded value: 6""3

You can verify these cases in excel.
If those ARE the correct cases, then this would work:

import java.util.List;
import java.util.ArrayList;
import java.util.regex.Pattern;
import java.util.regex.Matcher;

public class ParseCSV {
   final static String quoted = "\"(:?[^\"]|\"\")+\"";
   public static List<String> parseCSV(String csv, String delim) {
       final  Pattern NEXT_COLUMN = nextColumnRegex(delim);
       final List<String> strings = new ArrayList<String>();
       final Matcher matcher = NEXT_COLUMN.matcher(csv);
       while (!matcher.hitEnd() && matcher.find()) {
           String match = matcher.group(1);
           if (match.matches(quoted)) {
               match = match.substring(1, match.length() - 1);
           }
           match = match.replaceAll("\"\"", "\"");
           strings.add(match);
       }
       return strings;
   }

   private static Pattern nextColumnRegex(String comma) {
       String unquoted = "(:?[^\"" + comma + "]|\"\")*";
       String ending = "(:?" + comma +"|$)";
       return Pattern.compile('(' + quoted + '|' + unquoted + ')' +
ending);
   }

   public static void main(String[] args) {
       String csv = ",\",\",\"\",\"a,b\",\"\"a and b\"\",6\"\"3";
       List<String> result = parseCSV(csv, ",");
       for (String col : result) {
           System.out.println("Field value:" + col);
       }
   }

}

Text::CSV should do the job quite nicely for regular CSV files.

That doesn't suprise me. While REs are quite powerful indeed, there are
many, many tasks, where you need a more complex logic than REs can provide.
Even if you have custom requirements and you need a custom version of a CSV
parser I would still start with one of the Text::CSV modules and customize
that instead of rolling my own code from scratch and repeating all the
mistakes that have been fixed in the ready-made modules a long time ago.

jue

 Whatever you want to parse, you need a specification of
 the syntax - which usually is given in EBNF.

 This is not such a specification.

 Correct accordint to which syntax specification?

 What is a »continous string«? Is »continuous string or with
 space« supposed to be English?

 You do not seem to be able to describe what you need to parse
 in English, nor in ENBF. So how are you ever going to implement it?

 My (German) version of Excel does not use the comma as a
 separator at all, but the semicolon. So, a CSV file generated
 by it for two cells containing »1« and »2«, respectively, is:

1;2

 I suggest the following reading:

http://secretgeek.net/csv_trouble.asp