How to replace a character by a newline in Vim

Question

I'm trying to replace each , in the current file by a new line:

:%s/,/\n/g 

But it inserts what looks like a ^@ instead of an actual newline. The file is not in DOS mode or anything.

What should I do?

If you are curious, like me, check the question Why is \r a newline for Vim? as well.

Solution 1

Use \r instead of \n.

Substituting by \n inserts a null character into the text. To get a newline, use \r. When searching for a newline, you’d still use \n, however. This asymmetry is due to the fact that \n and \r do slightly different things:

\n matches an end of line (newline), whereas \r matches a carriage return. On the other hand, in substitutions \n inserts a null character whereas \r inserts a newline (more precisely, it’s treated as the input CR). Here’s a small, non-interactive example to illustrate this, using the Vim command line feature (in other words, you can copy and paste the following into a terminal to run it). xxd shows a hexdump of the resulting file.

echo bar > test
(echo 'Before:'; xxd test) > output.txt
vim test '+s/b/\n/' '+s/a/\r/' +wq
(echo 'After:'; xxd test) >> output.txt
more output.txt

Before:
0000000: 6261 720a                                bar.
After:
0000000: 000a 720a                                ..r.

In other words, \n has inserted the byte 0x00 into the text; \r has inserted the byte 0x0a.

Solution 2

Here's the trick:

First, set your Vi(m) session to allow pattern matching with special characters (i.e.: newline). It's probably worth putting this line in your .vimrc or .exrc file:

:set magic

Next, do:

:s/,/,^M/g

To get the ^M character, type Ctrl + V and hit Enter. Under Windows, do Ctrl + Q, Enter. The only way I can remember these is by remembering how little sense they make:

A: What would be the worst control-character to use to represent a newline?

B: Either q (because it usually means "Quit") or v because it would be so easy to type Ctrl + C by mistake and kill the editor.

A: Make it so.

Solution 3

In the syntax s/foo/bar, \r and \n have different meanings, depending on context.

---

Short:

For foo:

\r == "carriage return" (CR / ^M)
\n == matches "line feed" (LF) on Linux/Mac, and CRLF on Windows

For bar:

\r == produces LF on Linux/Mac, CRLF on Windows
\n == "null byte" (NUL / ^@)

When editing files in linux (i.e. on a webserver) that were initially created in a windows environment and uploaded (i.e. FTP/SFTP) - all the ^M's you see in vim, are the CR's which linux does not translate as it uses only LF's to depict a line break.

---

Longer (with ASCII numbers):

NUL == 0x00 == 0 == Ctrl + @ == ^@ shown in vim
LF == 0x0A == 10 == Ctrl + J
CR == 0x0D == 13 == Ctrl + M == ^M shown in vim

Here is a list of the ASCII control characters. Insert them in Vim via Ctrl + V,Ctrl + ---key---.

In Bash or the other Unix/Linux shells, just type Ctrl + ---key---.

Try Ctrl + M in Bash. It's the same as hitting Enter, as the shell realizes what is meant, even though Linux systems use line feeds for line delimiting.

To insert literal's in bash, prepending them with Ctrl + V will also work.

Try in Bash:

echo ^[[33;1mcolored.^[[0mnot colored.

This uses ANSI escape sequences. Insert the two ^['s via Ctrl + V, Esc.

You might also try Ctrl + V,Ctrl + M, Enter, which will give you this:

bash: $'\r': command not found

Remember the \r from above? :>

This ASCII control characters list is different from a complete ASCII symbol table, in that the control characters, which are inserted into a console/pseudoterminal/Vim via the Ctrl key (haha), can be found there.

Whereas in C and most other languages, you usually use the octal codes to represent these 'characters'.

If you really want to know where all this comes from: The TTY demystified. This is the best link you will come across about this topic, but beware: There be dragons.

---

TL;DR

Usually foo = \n, and bar = \r.

Solution 4

You need to use:

:%s/,/^M/g

To get the ^M character, press Ctrl + v followed by Enter.

Solution 5

\r can do the work here for you.

View more solutions

Comments

[Aleksander Grzyb]

/r is treated as pressing the Enter/Return key. It works on all platforms.

[Chris Phillips]

I'm using GVim on Windows, and I need neither the :set magic (it's not in my ~/_vimrc either) or ctrl-q. Just a simple ctrl-v followed by enter creates the ^M character for me just fine.

[tipu]

for whatever reason, replacing all '\n with ',\n works when doing: %s/'\n/',\r/g

[Evan Donovan]

Note that this requires GNU sed. Try printf 'foo\\nbar\n' | sed 's/\\n/\n/g' to see if it will work on your system. (Credit to the good people of #bash on freenode for this suggestion.)

[Andrew Marshall]

See also Why is \r a newline for Vim?.

[Jim Stewart]

C-v doesn't represent a newline; it's the "escape next literal character" command. I dunno what C-v is a mnemonic for either, but there's a reason it doesn't mentally map to newline.

[eksortso]

I wish this worked for classic vi. On AIX v6.1, \r doesn't work like this. But you can press Ctrl-V Enter in place of typing \r, and it works.

[Tomasz Gandor]

Ctrl-v is a mnemonic for "verbatim" - i.e. escape next key pressed to its "verbatim" keycode/character. In Windows it's paste: to keep things familiar. Ctrl-Q is for "(un)Quote" maybe. Quite stupid, anyway - but you can use it in binary files - e.g. to search for Ctrl-A through Ctrl-Z (Ascii 1-26 I guess).

[Jezen Thomas]

I have to do <C-v><C-m> to get the ^M character.

[Nagaraju]

Even this works Type V followed by M holding control key which looks like this ^M which can be used instead of \r or \n

[codeshot]

So I'm intrigued how you would substitute a character with a carriage return

[sjas]

@codeshot :s/x/^M/g should do. Insert the ^M via ctrl-v followed by ctrl-m.

[codeshot]

Thanks sjas, You know this question is one of the weirdest of all time. 1008 votes for the answer which basically says nothing more than "vim does what you found. That's because vim does what you found. Never forget that vim does what you found." I'd hoped to find a shortlist of codes for interesting characters in the pattern, the replacement and the reason for the weirdness so its easy to remember and predict other similar weirdness. That would have got my vote.

[sjas]

@codeshot a list of ascii control characters might help you. See cs.tut.fi/~jkorpela/chars/c0.html for further reference. I will update my answer to include two links.

[Mr. Llama]

I'm late to the party. If \r inserts <CR> and \n inserts a null, how would I replace something with a carriage return?

[0x0]

@KonradRudolph: Ages ago when people were using teletypes, before computers, when you reach the end of the line it took the time of inserting two characters for the print head to reach from the far right to the left to begin a new line. So if any character came when the head was moving from right to left, it was lost. To solve this problem people inserted a Carriage Return (CR) and a Line Feed (LF). After the speed of the printhead movement was reduced, different people took different routes- unix used LF, DOS used CR and LF, Apple used CR.

[0x0]

So, to maintain compatibility, DOS removes CR when the file is opened as ASCII, and has both when opened as binary. I'm not sure why this is a problem in unix, but this probably also has some quirks like that.

[Konrad Rudolph]

@Sunil I’m aware of the history of carriage return and line feed in different systems; however, it doesn’t explain the asymmetry between the meaning of \r and \n in searching and replacing.

[00dani]

Ctrl-C doesn't actually kill the editor, although it can cancel you back to Normal mode. Ctrl-V means verbatim, and Ctrl-Q means that someone made the mistake of loading the $VIMRUNTIME/mswin.vim configuration file. You don't need mswin. Just use your own vimrc instead.

[Akavall]

It is interesting that to go back I would have to do: %s/\n/,/g, not %s/\r/,/g.

[LangeHaare]

This answer has really come along way since the beginning stackoverflow.com/revisions/71334/1

[Mikko Rantalainen]

@SunnyRaj are you sure about the history of CR and LF? I was under impression that originally LF moved paper one row forward but did not move the printhead, and CR moved the printhead but did not move the paper. As a result, if your OS did not convert the input before printing, you could not just use just LF nor CR to get the correct output. MS DOS used raw printer data as the text file format, Mac OS used CR and converted from that to printer's raw format and UNIX used LF and converted from that to printer's raw format.

[tormodatt]

I had a long line containing a lot of explicit '\n' which I wanted to expand to a line each. after reading this thread, %s/\\n/\r/g did the trick! (notice the extra \, otherwise it searches for actual newlines)

[Chromium]

Actually, I always do something like: :%s/text/^M/g, where I just press ^V and Enter for the second part. Hope that help someone

[Don Hatch]

@Chromium I'd always done it like that too, but I'm glad to learn this new way, which I think is better. The advantage is it can be described in a recipe, which is more-or-less copy-pastable directly into vim, without having to explain the "where I just press ^V and Enter for the second part" part.

[Peter Mortensen]

Yes, but the question was about Vim. There is Stack Overflow question How can I replace a newline (\n) using sed?.

[arpit]

wow----this is amazing. I used to do sed -n l till now. Good to know that the same can be achieved with Ctrl-v in vim.

[Ben Farmer]

So how do I match the null characters I just inserted so that I can replace them with \r?

[Konrad Rudolph]

@BenFarmer You can enter a null character in command mode by pressing Ctrl+V, 0. So, to replace all null characters by newlines, type: Esc, :, %, s, Ctrl+V, 0, /, \ , r, /, g