Regular expressions

With regular expressions, you can replace and search for string patterns. You could for example show all URLs in a file or lines that start with a certain date.

Let's take the easier one as an example: This is how you show all lines that begin with the string Sep 13 in a file myfile.txt issue:

grep -E "^Sep 13" myfile.txt

In this case ^Sep 13 is your regular expression. The ^ sign says that a line must start with the following string. And there is much more you can do with regular expressions.

Escaping

The characters ^ and \ are seen as control-characters. ^ means "at the beginning of a line". With a backslash, you can escape these control-characters, meaning they act as body-characters again:

grep "^hallo" file

finds all occurrences of "hallo" at the beginning of a line in file.

grep "\^hallo" 

finds all occurrences of "^hallo" in a file

grep "\\^hallo"

finds all occurrences of "\^hallo" in a file

grep "\\\\^hallo"

finds all occurrences of "\\^hallo" in a file And so on...

Write regular expressions

For "finding a pattern defined by a regular expression", we speak of "matching".

Beginning of a line is

grep "^hallo" file

prints all occurrences of "hallo" at the beginning of a line in file.

The end of a line

grep "hallo$" file

prints all occurrences of "hallo" at the end of a line in file.

Find string1 OR string2

grep -E "Sep|Aug" file

prints all lines from file that contain "Sep" or "Aug".

Match a group of characters

grep -E "L[I,1]NUX" file

prints all lines from file that contain "LINUX" or "L1NUX"

Match a range of characters

grep -E "foo[1-9]" file

prints all lines from file that contain "foo1" or "foo2" till "foo9"

NOT the following characters

To invert matching for a group of characters

grep -E "for[^ e]" file

prints all lines from file that contain "for", but not followed by a space or an e, so not "for you" or "foresee"

Also

[^\n]*

means "all characters till the next newline". This can be useful when writing parsers.

With grep you have an additional possibility to invert matches:

grep -Ev "gettimeofday" file

prints all lines from file that do NOT contain "gettimeofday". This is a grep feature.

Any character

grep -E "L.nux" file

matches any character that is not a newline, e.g. Linux, Lenux and L7nux in file.

Match one or more times

grep -E "L[i]+nux" file

Match if i is there at least once in file The + here is a quantifier. It means, that i occurs 1 or more times. It is also possible to accept 0 or more times if you replace the + by a *.

Match n times

/etc/services is a table for protocols (services) and their port numbers. The protocols are filled up with blanks to have 16 characters. If you want to replace all protocols for port 3200 with sapdp00 you do it like this:

sed -ri "s/.{16}3200/sapdp00 3200/" /etc/services

Backreferences

Backreferences allows you to reuse matches. For example consider /var/log/nginx/access_log. It is full of lines like this:

127.0.0.1 - - [27/Dec/2020:12:07:27 -0800] "GET /wiki/load.php?lang=en&modules=jquery%2Csite%7Cjquery.client%7Cmediawiki.String%2CTitle%2Capi%2Cbase%2Ccldr%2CjqueryMsg%2Clanguage%2Cutil%7Cmediawiki.libs.pluralruleparser%7Cmediawiki.page.ready%2Cstartup%7Cskins.vector.legacy.js%7Cuser.defaults&skin=vector&version=o8vg2 HTTP/1.1" 200 277446 "http://localhost/wiki/index.php?title=Main_Page" "Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:83.0) Gecko/20100101 Firefox/83.0"

Every line tells you about a specific access request to your webserver. But maybe you are only interested in the part that starts with http, and you want to extract that part. Then you have to use backreferences:

cat access.log | sed "s;.*\(\http[^\"]*\)\(.*\);\1;"

will output for the example above

http://localhost/wiki/index.php?title=Main_Page

Now let's look at the line

cat access.log | sed "s;.*\(\http[^\"]*\)\(.*\);\1;"

what does it do?

cat access.log | : send access.log line by line to the next command (sed)
sed "s; - substitute the following pattern
.* - an arbitrary amount of characters that are not a newline. This will stop once the next pattern is found
\( - take the following match and remember it as first backreference
http[^\"]* - match a string started with http, followed by an arbitrary number of characters that is not a quote
\) - this is the end of the (first) backreference
\( - remember the following as the (second) backreference
.* - an arbitrary amount of characters that are not a newline will be matched
\) - this is the end of the (second) backreference
; - this is the end of the string that has to be substituted
\1 - substitute the string by the first backreference
; - this is the end of the substitution
" - this is the end of the sed parameter

Read regular expressions

*

An asterisk is a quantifier saying "whatever number of".

grep -E "Li*nux" file
Lnux
Linux
Liinux
Liiinux

An asterisk is placed next to an atom that can be repeated in whatever number. In the above example, the atom is the i character, but it can also be a group of characters:

grep -E "ba(na)*" file
ba
bana
banana
bananana

^

The ^ character stands for

• the beginning of a line if it stands at the beginning of a branch

# grep ^foo
barfoo
foo
foo

• "not" if it stands behind a bracket

# grep for[^e]
foresee
for each
for each

• the ^ character if it is escaped

# grep "\^"
adsf
as^df
as^df

?

The ? character stands for

• non-greedy matching:

http://.*?/

Understand regular expressions

Branches, Pieces and Atoms

A regular expression consists of one or more branches, separated by "|", the "OR" sign. If one of the branches matches, the expression matches:

grep -E "Tom|Harry"

Here, the expression is Tom|Harry, and Tom and Harry are both branches.

A branch consists of one or more pieces, seen in its particular order. A piece is an atom optionally followed by a quantifier:

grep -E "To*m"

Here, T is a piece as well as o* and m.

An atom is a character, a bracket expression or a subexpression. Each line can be an atom:

a
b
[^e]
(this is a subexpression)

quantifiers

A quantifier is used to define that an atom can exist several times. The * quantifier defines the atom in front of it can occur 0, 1 or several times:

grep -E "To*m"

Will find all lines containing Tom, Toom, Tooom and Tm.

See also

• scripting tutorial
• RegEx ComPoser
• RegEx training