Substitution: how to ignore the Nth first occurrences of a pattern?

Question

I have a file that looks as follows:

aa,bb,cc,dd,ee,ff,gg
aa,bb,cc,dd,hh
aa,bb,cc,dd,ii,jj
etc.

I would like to (preferably using one command) change occurrences of a , by a : from the 5th , until the last one.

I would thus get:

aa,bb,cc,dd,ee:ff:gg
aa,bb,cc,dd,hh
aa,bb,cc,dd,ii:jj
etc.

In other words, is there a way in vim to skip the first 4 (or n) occurrences of a pattern match and replace all other pattern matches (with the substitute command for example)?

I tried this:

1,$s/\(\([^,]*,\)\{4\}\)\([^,]*\),/\1\3:/g

but this only changes the 5th occurrence...

Answer 1

You could also try the following substitution command:

%s/\v(^(.{-},){4}.*)@<=,/:/g

The pattern is:

\v(^(.{-},){4}.*)@<=,

Which can be broken down like this:

\v(^(.{-},){4}.*)@<=,
| |||  |   |  |  |  |
| |||  |   |  |  |  +--- a comma
| |||  |   |  |  +--- the previous subpattern must be matched before what follows
| |||  |   |  +--- `.*` any text
| |||  |   +--- repeat the previous subpattern `.{-},` 4 times
| |||  |
| |||  |
| |||  +--- `.{-},` a minimum of text before a comma
| ||+--- capture the `.{-},` subpattern to apply the multi `{4}` to it
| |+--- beginning of a line (anchor)
| +--- capture the `^(.{-},){4}.*` subpattern to apply the multi `@<=` to it
+--- very magic mode

---

The @<= multi doesn't add anything to the pattern, it just asks the regex engine to make sure that at least 4 commas are found before the commas you want to substitute with a colon.

Also, there are 2 subpatterns you're capturing but you don't use any backreference to them (\1, \2, ..., \9), so you could tell Vim that you don't want it to count them as sub-expressions by prefixing a % sign before each open parenthesis:

%s/\v%(^%(.{-},){4}.*)@<=,/:/g

According to the help (:h /\%(), it makes the pattern a little bit faster.

---

I'm not sure, but more generally, it should work for other patterns too.

Suppose you wanted to replace all the occurrences of foobar, but ignoring the first N ones on any given line. Then, you could adapt the previous substitution command like this:

%s/\v(^(.{-}foobar){N}.*)@<=foobar/replacement/g

Answer 2

You'll have to play with /\=+submatch() and substitute(). First :s will select what is before the 5th coma (to ignore it until the zone start :h /\zs), then substitute() will replace the remaining comas.

Something like (untested)

%s/\v([^,]*,){4}\zs([^,]*),/\=substitute(submatch(0), ',', '::', 'g')/