Comments by "yapdog" (@yapdog) on "Destroyed By A Regex" video.

  1. I created my own regex engine that's not susceptible to that kind of error. This classic test regex from Rexegg: "^(A+)*B" With that pattern, a string as simple as "AAAAAAAAAAC" would require 3,584 steps. But a string with 100 A's followed by 1 C would cause a catastrophic backtracking failure, requiring 4,436,777,100,798,802,905,238,461,218,816 steps. Those numbers scared the sh!t out of me!😱 And if they don't scare you, then you haven't been paying attention. I finally just dumped all the standard regex models and built one completely from scratch (yes, in C😛). My engine significantly reduces the amount of backtracking: the 100 A string is resolved in just 1295 steps. 😁😁😁I'm still benchmarking it, but this engine is incredibly promising.........🤞
    1
  2.  @JohnDoe-yx4hi  As I stated, "^(A+)*B" is a pattern from Rexegg, and you'll find similar patterns on other regex sites, e.g. Mozilla. Yes, these are deliberately bad patterns designed for the purpose of demonstrating how engines can fail with catastrophic backtracking. This is a security issue. So, if you're designing a regex engine and you're not testing for "bad" patterns, then you shouldn't be developing a regex engine at all. Not if you intend for it to be used by others.
    1
  3.  @JohnDoe-yx4hi  I'd responded to you twice, and my comment was deleted both times. Have no idea why. Let me just say that that pattern is deliberately bad for the purposes of building a robust/secure regex engine.
    1