Regular expressions, part 2: classes and groups with egrep · Lesson 34

The dot matches any character, which is often too generous. Character classes let you say "a digit", "a vowel" or "anything but a space". And when you want "this or that", or to repeat a whole group rather than a single character, you need the extended regular expression syntax, which grep -E enables; egrep is a traditional alias for the same thing, and this VM has both.

Character classes

Square brackets match one character from the set inside. [aeiou] is a vowel, [0-9] a digit, [a-zA-Z] a letter, [^0-9] anything that is *not* a digit (the caret negates when it is first inside the brackets). POSIX names make this readable and locale-safe: [[:digit:]], [[:alpha:]], [[:space:]], [[:upper:]]; the double brackets are correct, the inner pair is the class name and the outer pair the set. Inside brackets most special characters lose their meaning, so [.] is a literal dot.

Classes work in basic grep too, and combine with *: grep '^[0-9]* matches lines made only of digits (including empty ones, because * allows zero). grep '[A-Z]' finds lines with at least one capital letter, and grep -v '[^a-z]' finds lines with nothing but lowercase letters.

Extended syntax: |, ( ), +, ?, {n,m}

Syntax	Meaning	Example (grep -E)
`a\\|b`	a or b	`gmail\\|yahoo`
`( )`	group, to apply \| or a repetition to several characters	`^(GET\\|POST)`
`+`	one or more of the preceding item	`[0-9]+` a whole number
`?`	zero or one (optional)	`colou?r`
`{n}` `{n,}` `{n,m}`	exactly n / at least n / between n and m	`[0-9]{3}` three digits

The difference between + and * matters: [0-9]* can match nothing at all, so it never filters lines out; [0-9]+ requires at least one digit. Groups make repetitions apply to sequences: (ha)+ is one or more ha, (ha){2,} is haha or longer. | has the lowest precedence, so ^cat|dog$ means "starts with cat, OR ends with dog"; write ^(cat|dog)$ for "exactly cat or exactly dog".

~% cd /root/lab/l34
l34% cat contacts.txt
ana ana@gmail.com 600123456
luis luis@yahoo.es
maria maria@empresa.org 912345678
pepe no-email
l34% grep '[0-9]' contacts.txt
ana ana@gmail.com 600123456
maria maria@empresa.org 912345678
l34% grep -E 'gmail|yahoo' contacts.txt
ana ana@gmail.com 600123456
luis luis@yahoo.es
l34% grep -oE '[0-9]{9}' contacts.txt
600123456
912345678
l34% grep -E '@[a-z]+\.(com|org)' contacts.txt
ana ana@gmail.com 600123456
maria maria@empresa.org 912345678
l34% grep -E '(ha){2,}' feelings.txt
haha!
hahaha

Basic versus extended

In basic syntax (plain grep, sed) the characters + ? | ( ) { } are ordinary, and you must escape them to get the special meaning: $ha$\{2,\}. In extended syntax (grep -E, egrep, sed -E, awk) they are special as written. Both understand . * ^ $ [ ] the same way. When a pattern needs any of the extended operators, just reach for -E; there is no downside.

\d for digit, \w for word character and similar shortcuts come from Perl, not POSIX. BusyBox grep does not understand them; write [0-9] or [[:digit:]] and [A-Za-z0-9_] instead, which work everywhere.

Commands in this lesson

Command	What it does
`grep '[0-9]' file`	Lines containing a digit.
`grep '[^a-z]' file`	Lines containing a non-lowercase character.
`grep '[[:upper:]]' file`	POSIX class: lines with a capital letter.
`grep -E 'cat\|dog' file`	Lines with cat or dog.
`grep -E '^(GET\|POST) ' file`	Group the alternatives.
`grep -E '[0-9]+' file`	One or more digits.
`grep -E 'colou?r' file`	Optional character.
`grep -oE '[0-9]{3}' file`	Print just the three-digit matches.

Quiz

What does `[0-9]` match?

The text `0-9`
Exactly one digit
Any number of digits

Which command finds lines containing `error` or `warning`?

`grep 'error|warning' log`
`grep -E 'error|warning' log`
`grep 'error warning' log`

What is the difference between `[0-9]*` and `[0-9]+`?

None
`*` allows zero digits, so it matches every line; `+` requires at least one
`+` only works with letters

What does `grep -E '^(yes|no) ` match?

Lines that are exactly `yes` or exactly `no`
Lines starting with yes or ending with no
Lines containing `yes|no`

Which pattern matches a Spanish postcode, exactly five digits, as a whole line?

`grep -E '^[0-9]{5} `
`grep -E '[0-9]5'`
`grep '^[0-9]* `

Practice

/root/lab/l34/contacts.txt has one contact per line; some include a phone number. Save the lines that contain at least one digit into /root/lab/l34/digits.txt.

Save into /root/lab/l34/mail.txt the lines of /root/lab/l34/contacts.txt that contain either `gmail` or `yahoo`, using a single pattern with an alternative.

/root/lab/l34/feelings.txt has short lines such as `ha`, `hat`, `haha!` and `hahaha`. Using a group with a repetition count, save the lines where `ha` appears two or more times in a row into /root/lab/l34/laugh.txt.

Open this lesson in the app to do the tasks in a real Linux machine and have them checked.