捕获所有适合正则表达式的组

Question

I have a regex that does pretty much exactly what I want: \\.?(\\w+[\\s|,]{1,}\\w+[\\s|,]{1,}\\w+){1}\\.?

Meaning it captures incidences of 3 words in a row that are not separated by anything except spaces and commas (so parts of sentences only). However I want this to match every instance of 3 words in a sentence.

So in this ultra simple example:

Hi this is Bob.

There should be 2 captures - "Hi this is" and "this is Bob". I can't seem to figure out how to get the regex engine to parse the entire statement this way. Any thoughts?

Answer 1

You cannot just get overlapping texts in capturing groups, but you can obtain overlapping matches with capturing groups holding the substrings you need.

Use

(?=\b(\w+(?:[\s,]+\w+){2})\b)

See the regex demo

The unanchored positive lookahead tests for an empty string match at every position of a string. It does not consume characters, but can still return submatches obtained with capturing groups.

Regex breakdown:

• \\b - a word boundary
• (\\w+(?:[\\s,]+\\w+){2}) - 3 "words" separated with , or a whitespace.
  • \\w+ - 1 or more alphanumeric symbols followed with
  • (?:[\\s,]+\\w+){2} - 2 sequences of 1 or more whitespaces or commas followed by 1 or more alphanumeric symbols.

This pattern is just put into a capturing group (...) that is placed inside the lookahead (?=...) .

Word boundaries are important in this expression because \\b prevents matching inside a word (between two alphanumeric characters). As the lookahead is not anchored it tests all positions inside input string, and \\b serves as a restriction on where a match can be returned.

In C#, you just need to collect all match.Groups[1].Value s, eg like this:

var s = "Hi this is Bob.";
var results = Regex.Matches(s, @"(?=\b(\w+(?:[\s,]+\w+){2})\b)")
                        .Cast<Match>()
                        .Select(p => p.Groups[1].Value)
                        .ToList();

See the IDEONE demo