Optional regular expression groups

Question

Hi,
I wonder if you could help me parse some name value pairs with RE?
I have a string of name value pairs and would like to parse out certain parts of certain pairs efficiently.
The string is like:
"name1=value1; name2=value2; FOO=value3:xyz; BAR_123=value4;"
or
The FOO and BAR pairs may or may not exist.
I am trying to extract the 'xyz' and '123' strings into groups.
I have tried umpteen capturing group and non-capturing group possibilities but none seem to either match or give null values.
One try that didn't seem to work at all was:
String regex =
               "\"" +
                    "(BAR[\\d]+=[\\S&&[^;]]+; |" +
                    "-|" +
                    "FOO=[\\S&&[^:]]+:[\\w]+; |" +
                    "[\\S&&[^=]]+=[\\S&&[^;]]+; )+" +
                    "" +
another slightly more successful one that returned nulls in the BAR pair was:
          String regex =
               "\"[\\w\\d\\s;_=]*?" +
               "(?:(?:FOO=[\\S&&[^:]]+:)([\\w]+))?" +
                    "[\\w\\d\\s;_=]*?" +
                    "(?:(?:BAR_)([\\d]*)=)?" +
                    "[\\w\\d\\s;_=]*?" +
I just couldn't seem to get any values out for the 'BAR' nvp....
Thanks,

Answer 1

Let's start by getting some obvious errors out of the way. First, the lookingAt() method returns true iff the regex matches starting at the beginning of the target string. That works okay with this test data (as would the matches() method), but it sounds like you're going to be searching for these quote-delimited sequences within a larger string; for that you need to use find(). Second, in your FOO and BAR alternatives, you're trying to match "FOO" or "BAR" immediately after the opening quote, but there seems to be other text in there that has to be matched first. There's also text following the part you're interested in, but preceding the closing quote, which you should also match. Finally, the pipe (or alternation operator) has lower precedence than any other regex metacharacter, so you need to enclose it in parentheses to limit its effect to just the text within the quotes. Taking your regex as a starting point, this is how that would look: "\"(?:([-])|FOO=\\w+:([^;]+)|BAR_(\\d+))\"" Notice also how I used non-capturing parens so as not to mess up the capturing-group numbers.
Now to the actual problem. The first three cases can be handled with a relatively simple regex (ugly as hell, but conceptually pretty straightforward). "\"(?:(-)|[^\"]*?(?:FOO=\\w+:([^;]+)|BAR_(\\d+))[^\"]*)\"" Plug that into your code and you should see that it yields the correct result for the first three strings, but not the fourth. In fact, it isn't really handling the second and third strings correctly. It's just testing for "X or Y" when it's supposed to be testing for "X and not Y" or "Y and not X" or "X and Y". Testing for the absence of a multicharacter sequence is fairly tricky, especially when you're also testing for the presence of another sequence. However, if you know that the FOO segment will always appear earlier in the string than the BAR segment, it will simplify things a great deal. The regex can take the form "X optionally followed by Y, or just Y": "\"(?:(-)|[^\"]*?(?:FOO=\\w+:([^;]+)(?:[^\"]*?BAR_(\\d+))?|BAR_(\\d+))[^\"]*)\""{code} This regex always captures the FOO value in group(2).  If there is also a BAR value, it will be captured in group(3); otherwise that group will be null.  If there is only a BAR value, you'll find it in group(4).
If you don't know in what order the FOO and BAR segments will appear, the regex will be twice as ugly.