Vim regexesEdit

Created 6/30/2014, updated 5/18/2017

Basics

  • \1, \2 etc: backreference to previous capturing group (works in both search pattern and replacement)
  • &: backreference the entire match (same as \0)

Gotchas

  • * is special when not escaped but...
  • + is special when escaped
  • \{x,y} (escaping only the opening bracket) works but...
  • You have to use \( \) (escape both parens)
  • [] is special when both are unescaped

Pro-Tips

  • \v: everything is special unless escaped ("very magic")
  • \V: turn off very-magic
  • You can use (somewhat) arbitrary delimiters, like @, instead of /
  • \_: match anything, including newlines (cf: .)
  • \zs: consider match to start at this point (ie. match but don't capture); useful for partial substitutions:
  • \ze: consider match to end at this point
    • :%s/foo\zsbar/baz/g will change "foobar" to "foobaz"
    • :%s/\v^(foobar)(baz)/\1/ (same with explicit capturing and restoration)
    • :%s/\v(^foobar)@<=baz// (same with zero-width look-behind assertion)
  • \=: evaluate an expression on right side of regex; eg: :%s/\v(\S+)/\=expand(submatch(1))/g (turns ~/foo into full path)
  • \C: make match case sensitive
  • \u: uppercase next letter in replacement
  • \l: lowercase next letter in replacement
  • \U: uppercase replacement from here on (until \e or \E)
  • \L: lowercase replacement from here on (until \e or \E)

Making Vim regexen more sane

Always be "very magic" with this in your ~/.vimrc:

nnoremap / /\v
vnoremap / /\v

Or just use Loupe.

Using Ruby to do heavy lifting instead

:rubydo can be used to transform lines; eg:

:rubydo $_ = $_.split(' ').reverse.join(' <- ')

Sources