C# Regex whitespace between capturing groups

Question

So basically, my input string is some kind of text containing keywords that I want to match, provided that:

each keyword may have whitespace/non-word chars pre/appended, or none (|\s\W)
there must be exactly one non-word/whtiespace char seperating multiple keywords, or keyword is at begining/end of line
Keyword simply ocurring as a substring does not count, e.g. bar does not match foobarbaz

E.g.:

input:    "#foo barbazboo tree car"
keywords: {"foo", "bar", "baz", "boo", "tree", "car"}

I am dynamically generating a Regex in C# using a enumerable of keywords and a string-builder

StringBuilder sb = new();
foreach (var kwd in keywords)
{
   sb.Append($"((|[\\s\\W]){kwd}([\\s\\W]|))|");
}
sb.Remove(sb.Length - 1, 1); // last '|'
_regex = new Regex(sb.ToString(), RegexOptions.Compiled | RegexOptions.IgnoreCase);

Testing this pattern on regexr.com, given input matches all keywords. However, I do not want {bar, baz, boo} included, since there is no whitespace between each keyword. Ideally, I'd want my regex to only match {foo, tree, car}.

Modifying my pattern like (( |[\s\W])kwd([\s\W]| )) causes {bar, baz, boo} not to be included, but produces bogus on {tree, car}, since for that case there must be at least two spaces between keywords.

How do I specify "there may be only one whitespace seperating two keywords", or, to put it differently, "half a whitespace is ok", preserving the ability to create the regex dynamically?

This looks too messy and convoluted. Are you trying to match the keywords as whole words? `_regex = new Regex($@"\b(?:{string.Join("|", keywords.Select(Regex.Escape))})\b", RegexOptions.Compiled | RegexOptions.IgnoreCase);`? — Wiktor Stribiżew, Feb 20 '22 at 11:56
One question though: can the `keywords` start/end with non-word chars? — Wiktor Stribiżew, Feb 20 '22 at 12:08
Yes, but at most one. So `{"#foo%", "#foo", "foo%"}` matches keyword `foo` — sir_photch, Feb 20 '22 at 12:12
Well, I meant if the `keywords` can start/end with a non-word char, not the input text. — Wiktor Stribiżew, Feb 20 '22 at 12:13
But only for whitespaces in a keyword, if I understand correctly? — sir_photch, Feb 20 '22 at 12:21
Can the keywords look like `keywords: {"$foo", "^bar^", "[baz]", "(boo)", "tree+", "+car"}`? — Wiktor Stribiżew, Feb 20 '22 at 12:25
Okay, I misread your question again. No, they cant; At most they'd look like `{"foo bar", "tree car"}`. — sir_photch, Feb 20 '22 at 12:29
Aha, so `\b` will work, but you need to sort the keywords before putting them into the pattern. — Wiktor Stribiżew, Feb 20 '22 at 12:32
I posted the two solutions, the first one is for you, the second one is a generic one. — Wiktor Stribiżew, Feb 20 '22 at 12:44

score 2 · Accepted Answer · answered Feb 20 '22 at 12:44

In your case, you need to build the

var pattern = $@"\b(?:{string.Join("|", keywords.OrderByDescending(x => x.Length).Select(Regex.Escape))})\b";
_regex = new Regex(pattern, RegexOptions.Compiled | RegexOptions.IgnoreCase);

Here, you are getting the longer keywords before shorter ones, so, if you have foo, bar and foo bar, the pattern will look like \b(?:foo\ bar|foo|bar)\b and will match foo bar, and not foo and bar once there is such a match.

In case your keywords can look like keywords: {"$foo", "^bar^", "[baz]", "(boo)", "tree+", "+car"}, i.e. they can have special chars at the start/end of the keyword, you can use

_regex = new Regex($@"(?!\B\w)(?:{string.Join("|", keywords.Select(Regex.Escape))})(?<!\w\B)", RegexOptions.Compiled | RegexOptions.IgnoreCase);

The $@"(?!\B\w)(?:{string.Join("|", keywords.OrderByDescending(x => x.Length).Select(Regex.Escape))})(?<!\w\B)" is an interpolated verbatim string literal that contains

(?!\B\w) - left-hand adaptive dynamic word boundary
(?: - start of a non-capturing group:
- {string.Join("|", keywords.OrderByDescending(x => x.Length).Select(Regex.Escape))} - sorts the keywords by lenght in descending order, escapes them and joins with |
) - end of the group
(?<!\w\B) - right-hand adaptive dynamic word boundary.

C# Regex whitespace between capturing groups

1 Answers1