The Shell: The Twenty Commands That Cover Almost Everything

• Everything is a program, and a program reads input and writes output. The shell just connects them.
• Pipes (`|`) send one command's output to the next as input. This is why small tools compose into something powerful.
• Redirection (`>`, `>>`, `<`) changes where input comes from and where output goes.

cat access.log | grep ' 500 ' | sort | uniq -c | sort -rn | head

That one line counts the most common server errors, sorted, top ten. None of those tools knows what it is part of. That is the whole idea.

• ls, cd, pwd — move around.
• cp, mv, rm — copy, move, delete. Learn the flags before you use rm on anything large.
• cat, head, tail, less — look at things. `tail -f` on a log file is the one you will use daily.
• grep — search inside files. The most used command on any machine.
• find — search for files by name, size or date.
• chmod — permissions. 644 for files and 755 for scripts is a reasonable default.
• ps, kill, top — see what is running and stop it.
• curl — make HTTP requests. This is how you test an API.
• ssh, scp — remote machines.
• tar, zip — archive and extract.

• Start with `#!/usr/bin/env bash` and a shebang, then make it executable with chmod +x.
• Add `set -euo pipefail` at the top. It stops the script on the first error instead of ploughing on and corrupting something.
• Quote every variable: "$var" not $var. This alone prevents a large class of nasty bugs.
• Run it with bash -x for a trace of every command, which is the fastest way to see where it went wrong.
• Add an echo of what it is about to do before anything destructive.

#!/usr/bin/env bash
set -euo pipefail

target="${1:?usage: clean.sh <dir>}"
echo "about to remove old logs in $target"
find "$target" -name '*.log' -mtime +30 -print -delete