Equivalent to Vim's regex start and end match (`\zs` and `\ze`)?

Question

If I wanted to replace only part of a match with something else in Atom, how would I do that?

A Ruby regex solution would probably work, because Ruby uses a regex engine very much like Oniguruma (which is used by Atom).

I found an Oniguruma manual which says \G sets a match start position, but this doesn't work for me.

Answer 1

I don't think Ruby has direct equivalents to \zs and \ze, but you can use lookahead((?=pat)) and lookbehind((?<=pat)) to mimic their behavior:

[1] pry(main)> ('abc'.match /(?<=a)b(?=c)/)
=> #<MatchData "b">
[2] pry(main)> ('bc'.match /(?<=a)b(?=c)/)
=> nil
[3] pry(main)> ('ab'.match /(?<=a)b(?=c)/)
=> nil

Unfortunately, Ruby's lookbehind needs the pattern to be of a fixed size(or several alternatives of fixed size) - a limitation Vim's \zs does not have.

Answer 2

Ruby once used Oniguruma as the regex engine, but it has since switched over to Onigmo (a fork of Oniguruma) from Ruby 2.0. Onigmo supports several new constructs, such as conditional construct and routine call (which allows you to write recursive regex).

Direct equivalent to \zs is \K, which is only supported in Onigmo. Note that \K and look-behind are not equivalent.

To demonstrate one of such cases, given the string aaaa:

• a\Ka has 2 matches starting at index 1, 3, but those matches are found when the engine performs the search at index 0 and 2.
• (?<=a)a has 3 matches starting at index 1, 2, 3.

The difference comes from the fact that \K (or \zs) only modifies the starting index of the group 0 (the text matched by the whole expression). The positions where the engine starts searching for next match are the same as when the same pattern is stripped of all \K.

There is no direct equivalent to \ze. However, for simple cases where \ze is not inside any group, you can mimic the behavior by using a look-ahead, as shown in Idan Arye's answer. \ze is rather similar to look-ahead, since the engine searches for the next match based on the ending index set by \ze, rather than the index reached by the pattern after \ze.