Regular Expressions

grep's -i/-v/-n flags and sed's s/find/replace/ only get you so far when "find" needs to mean "any of these shapes" instead of one exact string. A regular expression (regex) is a tiny pattern language for describing shapes of text - and it's the single most useful skill layered on top of everything you've learned about grep and sed so far.

The practice terminal's grep and sed use real, JavaScript-flavored regex syntax (the same family as grep -E, Python, and Perl) - bare parentheses for grouping, no backslash needed before +, ?, or |. That is the modern, most-transferable dialect, so every pattern in this lesson runs there exactly as written.

Anchors: where in the line

grep '^ERROR' log.txt      # lines that START WITH "ERROR"
grep 'timeout$' log.txt    # lines that END WITH "timeout"
grep '^$' log.txt          # completely empty lines

^ anchors to the start of the line, $ anchors to the end. Without an anchor, a pattern matches anywhere in the line.

Character classes: one of a set

grep '[0-9]' log.txt       # any line containing at least one digit
grep '[A-Z]' log.txt       # any line containing an uppercase letter
grep '[^0-9]' log.txt      # lines containing a NON-digit character

[...] matches exactly one character from the set inside the brackets. A leading ^ inside the brackets inverts the set - [^0-9] means "any character that isn't a digit."

The dot: any character

grep 'l.ne' log.txt        # matches "line", "lane", "l3ne" - any single character in that spot

. matches literally any one character. If you actually want to match a literal period, escape it: \..

Quantifiers: how many

grep 'ERRO*R' log.txt      # "ERR", "ERROR", "ERROOOR" - zero or more O's
grep '[0-9]+' log.txt      # one or more digits in a row
grep 'colou?r' log.txt     # matches BOTH "color" and "colour"

Alternation: this OR that

grep -E '^(ERROR|WARN)' log.txt   # lines starting with either word

| means "or," and parentheses group the alternatives so ^ applies to the whole (ERROR|WARN), not just ERROR. -E isn't strictly required in the practice terminal (its regex is always "extended"), but typing it is good muscle memory for real grep, where plain grep needs it to enable |, +, ?, and unescaped ().

Groups and backreferences with sed

Parentheses don't just group alternatives - they capture the text they matched, so you can reuse it in a replacement. sed refers to captured groups as \1, \2, and so on:

echo "John Smith" > name.txt
sed 's/([A-Z][a-z]+) ([A-Z][a-z]+)/\2 \1/' name.txt

That swaps first and last name: group 1 captures "John," group 2 captures "Smith," and the replacement \2 \1 writes them back in reverse order.

Try it

printf 'ERROR: disk full\nWARN: low memory\nline three\n' > log.txt
grep '^ERROR' log.txt
grep -E '^(ERROR|WARN)' log.txt
grep '[0-9]' log.txt
echo "John Smith" > name.txt
sed 's/([A-Z][a-z]+) ([A-Z][a-z]+)/\2 \1/' name.txt

Further reading

• "Mastering Regular Expressions" by Jeffrey Friedl (O'Reilly) - the definitive, exhaustive book on regex across every major tool and language.
• "sed & awk" by Dale Dougherty and Arnold Robbins (O'Reilly) - ties regex directly into real sed and awk usage, with the classic POSIX dialects this lesson's modern syntax evolved from.