Question

I’m a bit lost creating a RegEx under C#.NET.
I’m doing something like parser, so I use Regex.Replace to search text for certain “variables” and replace them with their “values”.
Each variable starts with ampersand (“&”) and ends with ampersand (begining of another variable) or dot.
Each variable (as well as text surrounding variables) can only consist of alphanumerical characters and certain “special” characters, that being “$”, “@”, “#” and “-“.
Nor variables, nor the rest of the text could contain space characters (” “).

Now, the problem is that I’m trying to figure out a RegEx replacing one possible ending character (“.”), while not replacing the other possible ending character (“&”).
Which happanes to be quite an issue:

• “&”+variable+”[^A-Za-z0-9#@$]” does what I want, except for it also replaces “&” – not acceptable.
• “&”+variable+”(.)?\b” replaces dot, but only if followed by literal character – not if it’s followed by \&\@#\$\- and that could occur, so this doesn’t work either.
• “&”+variable+”(.)?(?!A-Za-z0-9)” does exactly what i want as for the ending characters, except it doesn’t recognize true end of variable – this way, search-and-replace for “&DEN” also replaces that part in another variable, called “&DENV” – of which “&DEN” is a substring. This would create false/misleading results – totally unacceptable.

These were all the possibilities I could think of (and search of); is it possible to do the task I require with one RegEx at all? Under C#.NET RegEx parser?

Just to illustrate desired function:

string variable="DEN";
string replaceWith="28";
string replText;
string regex = "<desired regex>";
replText = Regex.Replace(replText, "&"+variable+regex, replaceWith);

replText="&DEN";

=> replaced => repltext==”28″

replText="&DENV"    

=> not replaced => repltext==”&DENV”

replText="&DEN&DEN"    

=> replaced => repltext==”2828″

replText="&DEN&DENV"    

=> replaced, not replaced => repltext==”28&DENV”

replText="&DEN.anything"

=> replaced and dot removed => repltext==”28anything”

replText="&DEN..anything"

=> replaced and first dot removed => repltext==”28.anything”

variable could also be like “#DE@N-$”.

Answer 1

The following works correctly on all of your examples. I assumed that a variable &FOO should only be replaced if it’s followed by ., &, or end-of-string $. If it’s followed by anything else, it’s not replaced.

In order to match but not capture a terminating &, I used a lookahead assertion (?=&). Assertions force the string to match the regex, but they don’t consume any characters, so those characters aren’t replaced. Trailing . are still captured and replaced as part of the variable, however.

Finally, a MatchEvaluator is specified to use the captured pattern to do a lookup in the replacements dictionary for the replacement value. If the pattern (variable name) is not found, the text is effectively untouched (the full original capture is returned).

class Program
{
    static string ReplaceVariables(Dictionary<string, string> replacements, string input)
    {
        return Regex.Replace(input, @"&([\w\d$@#-]+)(\.|(?=&)|$)", m =>
        {
            string replacement = null;
            return replacements.TryGetValue(m.Groups[1].Value, out replacement)
                 ? replacement
                 : m.Groups[0].Value;
        });
    }

    static void Main(string[] args)
    {
        string[] tests = new[]
        {
            "&DEN", "&DENV", "&DEN&DEN",
            "&DEN&DENV", "&DEN.anything",
            "&DEN..anything", "&DEN Foo",
            "&DEN&FOO&DEN"
        };

        var replace = new Dictionary<string, string>
        {
            { "DEN", "28" },
            { "FOO", "42" }
        };

        foreach (var test in tests)
        {
            Console.WriteLine("{0} -> {1}", test, ReplaceVariables(replace, test));
        }
    }
}