Which terminal commands do data people actually need?

The server has no buttons to click. These 24 commands are the whole survival kit: move around, look at data, slice columns, search, pipe, and avoid the three classic accidents. The print-ready A4 PDF is at the bottom.

Move around

• pwd: where am I.
• ls -lah: sizes + hidden files.
• cd -: back to the last directory.
• mkdir -p a/b: nested directories in one go.

Look at data

• head -5 f.csv: first rows.
• tail -f app.log: follow a log live.
• wc -l f.csv: row count, any size, instantly.
• less f.csv: scroll without loading, q quits.

Slice and count

• cut -d, -f1 f.csv: column 1.
• sort | uniq -c: value counts.
• awk -F, '{print $2}': column 2, with logic if needed.
• tr ',' '\t': swap delimiters.

Search

• grep -r 'TODO' .: find in files.
• grep -c error app.log: count matches.
• find . -name '*.csv': find files.
• history | grep ssh: what did I run last time.

Pipes

• a | b: output of a feeds input of b.
• > f.txt: write (overwrites!).
• >> f.txt: append.
• cmd1 && cmd2: run 2 only if 1 worked.

Survive

• ctrl+c / ctrl+r: kill the command / search history.
• man cmd: the manual.
• chmod +x run.sh: make a script runnable.
• rm: no trash bin. Think first.

The one-liner worth stealing

Value counts of column 3, no pandas, any machine:

cut -d, -f3 sales.csv | sort | uniq -c | sort -rn | head

Take the column, group identical values, count each, biggest first, top 10. Works on a 10GB file over ssh.

The traps: the terminal forgives nothing

There is no undo and no trash. The pros are not braver, they just echo before they run.