Submatch has been sped up by implementing a modified Boyer–Moore–Horspool algorithm with an average-case complexity of O(N) on random text. Worst case, it behaves similarly to the previous implementation O(MN), where M is the length of the boundary and N is the length of the buffer to operate on.
Method SubMatch looks for two things:
1. Whether the byte array segment fully contains the boundary, or
2. Whether the byte array ends with the start of the boundary.
Case 1 is now a lot faster than the previous implementation. Case 2 remains using the same code as before. The method will do Case 1 until the matchOffset is equal to N-M. It then switches to Case 2, unless a match is found.
The code can be further sped up with a full Boyer–Moore implementation, or something more sophisticated. This however can be evaluated in the case that this implementation is insufficiently performant for our main scenarios.
This commit resolves issue #575.
Chris R
David Obando
N. Taylor Mullen