Question

1. How does backtracking differ from back-referencing in regular expressions?

2. How does back-referencing win limitation having with backtracking or vice-versa?

Answer 1

Backtracking is a way for a state machine to back up and retry other matches for a regular expression. It’s something that’s pretty much internal to the regex engine.

For example, say you’re trying to match the regex [a-z]*a, any number of lower case characters followed by an a.

Given the input abca, a greedy match will assign all of that to the [a-z] portion of the regex but then there’s no way to match the final a. Backtracking allows the engine to back up by returning that final a to the input stream and trying again, assigning abc to the [a-z] portion and a to the a portion.

---

Back-referencing on the other hand, is a means for a user of the regex engine to reference previously captured groups. For example,

s/^([a-z])([a-z])/\1_\2/
  \_____/\_____/
     |      |
     |      +- capture group 2
     +-------- capture group 1

may be a command to insert _ between two consecutive lower case letters at the start of each line. The \N back-reference (where N represents a number) refers back to the groups captured within ().