Regular Expressions are search patterns for “Regular Text”. They are used by many different tools and languages, including the Linux grep command, the Windows findstr command, less, vi/vim, sed, awk, perl, python, and many others.

• Video Lecture on Regular Expressions (This is similar to the lecture given in class and is made available for review).

Regular Expressions can be a little daunting to learn: they often look like someone was just bashing their head against the keyboard (or, like a cat was lying on the keyboard). But they are very powerful - a well-written regular expression can replace many pages of code in a programming language such as C or C++ - and so it is worth investing some time to understand them.

In a regular expression (regexp), any character that doesn't otherwise have a special meaning matches that character. So the digit "5", for example, matches the digit "5"; similarly "cat" matches the letters "c", "a", and "t" in sequence.

A backslash can be used to remove any special meaning which a character has. The period character "." is a type of wildcard (see below), so to search for a literal period, we place a backslash in front of it: "\."

A period "." will match any single character. Similarly, three periods "..." will match any three characters.

Bracket Expressions or Character Classes are contained in square brackets "[[|]]"

• A list of characters in square brackets will match any one character from the list of characters: "[abc]" will match "a", "b", or "c"
• A range of characters in square brackets, written as a starting character, a dash, and an ending character, will match any character in that range: "[0-9]" will match any one digit.
• There are some pre-defined named character classes. These are selected by specifying the name of the character class surrounded by colons and square brackets, placed within outer square brackets, like "[[:digits:]]". The available names are:
  • alnum - alphanumeric
  • alpha - alphabetic characters
  • blank - horizontal whitespace (space, tab)
  • cntrl - control characters
  • digit - digits
  • graph - letters, digits, and punctuation
  • print - letters, digits, punctuation, and space
  • punct - punctuation marks
  • space - horizontal and vertical whitespace (space, tab, vertical tab, form feed)
  • upper - UPPERCASE letters
  • lower - lowercase letters
  • xdigit - hexidecimal digits (digits plus a-f and A-F)
• Ranges, lists, and named character classes may be combined - e.g., "[[[:digit:]]+-.,]" "[[:digit:]][:punct:]]" "[0-9_*]"
• To invert a character class, add a carat ^ character as the first character after the opening square bracket: "[^[:digit:]]" matches any non-digit character, and "[^:]" matches any character that is not a colon.
• To include a literal carat, place it at the end of the character class. To include a literal dash or closing square bracket, place it at the start of the character class.

• A repeat count can be placed in curly brackets. It applies to the previous element: "x{3}" matches "xxx"
• A repeat can be a range, written as min,max in curly brackets: "x{2,5}" will match "xx", "xxx", "xxxx", or "xxxxx"
• The maximum value in a range can be omitted: "x{2,}" will two or more "x" characters in a row
• There are short forms for some commonly-used ranges:
  • "*" is the same as "{0,}" (zero or more)
  • "+" is the same as "{1,}" (one or more)
  • "?" is the same as "{0,1}" (zero or one)

• The vertical bar indicates alternation - either the expression on the left or the right can be matched: "hot|cold" will match "hot" or "cold"

• Elements placed in parenthesis are treated as a group, and can be repeated: "(na)* batman" will match "nananana batman" and "nananananananana batman"
• Grouping may also be used to limit alternation: "(fire|green)house" will match "firehouse" and "greenhouse"

• Anchors match locations, not characters.
• A carat symbol will match the start of a line: "^[[:upper:]]" wil match lines that start with an uppercase letter.
• A dollar sign will match the end of a line: "[[:punct:]]$" will match lines that end with a punctuation mark.
• The two characters may be used together: "cat" will match the word "cat" anywhere on a line, but "^cat$" will only match lines that contain only the word "cat". Likewise, "^[[0-9.]]$" will match lines that are made up of only digits and dot characters.

Regular expressions have evolved over the years, and the various tools that handle regular expressions have different capabilities and slightly different syntax.

In particular, the original Unix search tool grep came in three varieties:

• fgrep, which could search only for fixed text patterns
• grep, which handled basic regular expressions
• egrep, which handled an extended form of regular expressions

The GNU project originally shipped all three commands, but fgrep and egrep were never fully standardized, so they were removed from the Posix standard in 2001. They were recently also removed from the GNU project.

Unlike the original Unix grep, the GNU grep can handle the full extended regular expression syntax, in either of two ways:

• To use the special characters (called “meta-characters”) ?, +, {, |, (, and ) preceed them with a backslash. In other words, while a backslash makes special characters like . or * ordinary, it also makes ordinary characters ? + { | } into special characters.
• Alternately, use the -E option to make grep understand extended regular expressions, which causes ? + { ( | ) } to become special characters.

Other tools, such as sed, similarly require backslashes in front of some of the extended regexp meta-characters (or, if you're using a GNU version of sed, you can use the -E option to enable extended regular expressions, just like GNU grep).

The Perl language introduced one of the most powerful and consistent versions of the regular expression language. There has been increasing consensus around “Perl-Compatible Regular Extensions” (aka PCRE) and that dialect is available in many tools (including GNU grep via the -P option, as well as the PCRE/PCRE2 library for C and C++ programs, which is used in many software packages including Safari and Apache httpd).

Regular expressions can be used in many places:

• Linux
  • GNU grep
  • The bash test command [[ "string" =~ regexp ]]
    • Note that the regular expression is not quoted
    • Example: X="ABC"; if [[ "$X" =~ ^[[:upper:]]{3}$ ]]; then echo "MATCH" ; else echo "NO MATCH" ; fi
  • The less command, using the / and ? keystrokes for searching forward and backward
  • The vi/vim editor, also using the / and ? keystrokes for searching forward and backward
  • The sed and awk utilities

• Windows
  • findstr /R (see notes below)

• Programming Languages (Cross-Platform)
  • Cross-platform Shells (Powershell, zsh, bash)
  • Python
  • JavaScript
  • Perl
  • C / C++ via PCRE/PCRE2 library
  • …and many others!

The Windows findstr command accepts regular expressions or literal expressions. It will guess what you're using, and may guess incorrectly, so it's best to use the /R and /L optons to directly specify if your search pattern is a regexp or literal.

Findstr permits multiple search patterns in a quoted string, separated by a space; this acts like a type of alternation. However, this makes it impossible to use a literal space in a search pattern. If you wish to include a space in your search pattern, prepend /C: to your search string. You can use multiple /C: search strings.

For example, FINDSTR /R /C:“red” /C:“blue” INPUTFILE is roughly equivalent to grep -E “red|blue” INPUTFILE

Findstr is also limited to (approximately) 127 characters in the regular expression.

For information on findstr's regular expression dialect, see help findstr. In particular, the findstr command does not support alternation with the | symbol, repetition other than with the * symbol, named character classes [[:name:]], or grouping ( ).