7

I need to match @anything_here@ from a string @anything_here@dhhhd@shdjhjs@. So I'd used following regex.

^@.*?@

or

^@[^@]*@

Both way it's work but I would like to know which one would be a better solution. Regex with non-greedy repetition or regex with negated character class?

Pranav C Balan
  • 113,687
  • 23
  • 165
  • 188

2 Answers2

8

Negated character classes should usually be prefered over lazy matching, if possible.

If the regex is successful, ^@[^@]*@ can match the content between @s in a single step, while ^@.*?@ needs to expand for each character between @s.

When failing (for the case of no ending @) most regex engines will apply a little magic and internally treat [^@]* as [^@]*+, as there is a clear cut border between @ and non-@, thus it will match to the end of the string, recognize the missing @ and not backtrack, but instantly fail. .*? will expand character for character as usual.

When used in larger contexts, [^@]* will also never expand over the borders of the ending @ while this is very well possible for the lazy matching. E.g. ^@[^@]*a[^@]*@ won't match @bbbb@a@ while ^@.*?a.*?@ will.

Note that [^@] will also match newlines, while . doesn't (in most regex engines and unless used in singleline mode). You can avoid this by adding the newline character to the negation - if it is not wanted.

Rory O'Kane
  • 29,210
  • 11
  • 96
  • 131
Sebastian Proske
  • 8,255
  • 2
  • 28
  • 37
  • 2
    *Note that `[^@]` will also match newlines, while `.` doesn't* is not true without specifying the regex flavor. In POSIX, TRE and Tcl (Henry Spencer's regex library) regex flavors a dot matches line break symbols by default. – Wiktor Stribiżew Dec 21 '16 at 19:38
7

It is clear the ^@[^@]*@ option is much better.

The negated character class is quantified greedily which means the regex engine grabs 0 or more chars other than @ right away, as many as possible. See this regex demo and matching:

enter image description here

When you use a lazy dot matching pattern, the engine matches @, then tries to match the trailing @ (skipping the .*?). It does not find the @ at Index 1, so the .*? matches the a char. This .*? pattern expands as many times as there are chars other than @ up to the first @.

See the lazy dot matching based pattern demo here and here is the matching steps:

enter image description here

Wiktor Stribiżew
  • 607,720
  • 39
  • 448
  • 563