How to use "-c" command line option?

Question

Fully realizing the heritage of vi, I know that its predecessor (ex) could be, and was, used for command line processing of files ("sed-style", if you like).

The manpage says:

-c {command}

{command} will be executed after the first file has been read. {command} is interpreted as an Ex command. If the {command} contains spaces it must be enclosed in double quotes (this depends on the shell that is used).

And:

-e

Start Vim in Ex mode, just like the executable was called "ex".

But I was unable to find "real-life" examples, or get something like a simple search-and-replace to work myself. I like examples...

So let's assume a file.txt...

abcdef

...and I would like to open that file, remove all occurrences of e in it (:%s/e//g if I were in full vim mode), save, and close.

How would I go about it?

Answer 1

An "ex command" means anything that you type in the commandline after the :, such as :write, it doesn't really have anything to do with "ex mode", as such. ex mode is a "mode" where you only have the commandline, and not the graphical editor ("vi" stands for "visual ex mode").

So it's really as simple as:

vim -c ':%s/e//g' -c ':wq' file.txt

There's no need to use "ex mode". You can if you want:

vim -e -c ':%s/e//g' -c ':wq' file.txt

It has the advantage of not messing with your terminal so much, but other than that there's little difference.

Answer 2

Update:

The easiest portable way to do this is:

printf '%s\n' %s/e//g x | ex file

This avoids error conditions discussed below.

---

Original Answer

One of the advantages of ex as opposed to Vim is that ex is guaranteed by POSIX.

If you're in the sysadmin world and you need to script automated edits to text files (e.g. configuration files to be edited across thousands of different remote servers), you want to use something portable. Perl is great but not portable—and it's overkill for very simple file edits. sed and awk are both guaranteed by POSIX, but they're not designed for in-place file editing. (sed has a -i switch in some implementations such as GNU sed to allow for in-place file editing, but for portable scripting that option shouldn't be relied upon as it is not in the POSIX standard for sed.)

The correct tool for the job is ex.

For your example, removing all instances of the letter "e" from the file, the ex command is simple (and portable):

ex -sc '%s/e//g | x' file

The -s switch starts ex silently, for batch processing (as opposed to interactive use where messages are printed out to the terminal).

The -c switch precedes the command to be run.

The substitute command is as you wrote it.

The vertical bar separates commands to be run sequentially (unless you're using it with the global command, but we're not.)

The x command exits after writing any changes to the file. (It only writes to the file if edits were made, thus preserving the file's timestamp if no edits were necessary.)

---

There is one major caveat to the above command: If the letter "e" doesn't exist anywhere in the file, the substitute command will fail and the x command won't be executed. This will leave you staring at a silently running ex process with no error output, or output of any kind. It's not at all obvious in this case what's happening, but it's actually waiting for you to type in ex commands. (It's more obvious if you omit the -s flag, so that you are actually prompted for input with a colon.) Just type q! to exit in this case.

If I were to actually use this to edit files on remote servers, of course, I would want the command to exit correctly even if there weren't any instances of the letter "e" in the files being edited. Vim has an e flag for the substitute command which would handle this by suppressing errors if the regex weren't matched:

ex -sc '%s/e//ge | x' file

But the POSIX way would be to use the global command to run the substitute command on all lines matching the given regex—and to not run substitute at all if the regex weren't found:

ex -sc 'g/e/s///g
x' file