Execute bash commands one at a time

Question

I have a list of bash commands that I'd like to execute one at a time. They could be in a file, or the terminal, or whatever is necessary. But I'd like to execute the first one then I'd check the results or whathaveyou, and then execute the next one, etc. I don't want a timed pause between each. Better still is if I could select the next command to execute. Can this be done?

I know copy/paste would work, and I'll do that if I have to, but I'd like something more efficient if possible.

Answer 1

Well, you could always manually read the commands from a file, and run them after getting confirmation. If you want to support the full shell syntax and e.g. pipelines, you'll need to use eval. Assuming each command is on a single line, you could use a something like this script:

#!/bin/bash

while IFS=  read -r line; do
        # skip empty lines
        if [[ "$line" =~ ^[[:space:]]*$ ]]; then
                continue
        fi
        printf "Next command:\n"
        printf "  %s\n" "$line"
        printf "Hit enter to run or ^C to break"
        read x < /dev/tty
        echo -------------------------------------------------
        eval "$line"
        ret=$?
        echo -------------------------------------------------
        printf "Previous command exited with status %d\n" "$ret"
        echo
done

call that single-step.sh, and run as single-step.sh < file-with-commands.txt

Answer 2

Usually, when you have a list of commands, the main idea is to create a shell script where you can check more deeply error handling.

But just for fun, this is what I would do:

file:

uptime
pwd
whoami

command:

set -eo pipefail
xargs < file -n1 | sh

The part set -eo pipefail is taking care of failed commands and exit. Better put this in a script to avoir quitting your interactive shell in case of error.

Answer 3

You'll find that "stepping through my script, inspecting the system after every line": That's the classical job of a debugger!

How to debug a bash script? has quite a few options (and I've used none of them. If you prefer the command line, this answer mentions and illustrates bashdb; sadly the reference documentation seems offline, but the man page seems available. The bad news is that it's not available for any debian, fedora, suse and was dropped from Ubuntu nearly a decade ago, which often is an indication of packaging trouble. Might have become much better!

---

Assuming this holds true:

could you tell us what the most complex bash script that you want to run is? Is it really just one command after the other, as your choice of words "a list of bash commands" suggests? or might it involve calling bash functions (and would you want that function to be one "command" or "step" through the function's contents as well? how deeply?) or even loops like for f in *.mp3; do id3tag "$f"; done?

@MarcusMüller Yes, they're simple commands. Almost all will be "mv file1 file2" commands.

Alternatively, you could just start an interactive shell after every line. A line gets executed, you do whatever you want to do (and quit the "inspection" shell through ctrl+d or exit), and the next line gets executed.

You can do something like (assuming GNU sed, which is more likely than BSD sed for someone who asks about GNU bash):

sed \
 's/.*/echo "Next line is:"\ncat << END_ORIGINAL\n&\nEND_ORIGINAL\nbash -i;&;echo "Last line was:"\ncat << END_ORIGINAL\n&\nEND_ORIGINAL\n\n/' \
  originalscript.sh > steppablescript.sh

which is a way to take each original line (matching the whole line with .*), surround it in a HEREDOC with

<< delimiter_string
&
delimiter_string

outputting that through cat (which is a horrible way to not have to figure out how to correctly quote the line). (& in sed refers to "replace this with the whole match", i.e. here with the original line)

Then, it runs an interactive bash (bash -i); then it the actual line (& again!), then it outputs the "last line" through the same trick as before.

you can then run steppablescript.sh as you would have originalscript.sh.