From docs.python: re:

A regular expression (or RE) specifies a set of strings that matches it; the functions in this module let you check if a particular string matches a given regular expression

This post gives an overview and examples of regular expression syntax as implemented by the re built-in module (Python 3.7+). Assume ASCII character set unless otherwise specified. This post is an excerpt from the book Python re(gex)? by Sundeep Agarwal.

(The above diagram was created using Regulex.)

Elements that define a regular expression

Anchors Description
\A restricts the match to start of string
\Z restricts the match to end of string
^ restricts the match to start of line
$ restricts the match to end of line
\n newline character is used as line separator
\b restricts the match to start/end of words
word characters: alphabets, digits, underscore
\B matches wherever \b doesn’t match

^, $ and \ are metacharacters in the above table, as these characters have special meaning. Prefix a \ character to remove the special meaning and match such characters literally. For example, \^ will match a ^ character instead of acting as an anchor.

Feature Description
| multiple RE combined as conditional OR
  each alternative can have independent anchors
(pat) group pattern(s), also a capturing group, i.e.: a(b|c)d is same as abd|acd
(?:pat) non-capturing group
(?P<name>pat) named capture group
. Match any character except the newline character \n
[] Character class, matches one character among many
Greedy Quantifiers Description
* Match zero or more times
+ Match one or more times
? Match zero or one times
{m,n} Match m to n times (inclusive)
{m,} Match at least m times
{,n} Match up to n times (including 0 times)
{n} Match exactly n times
pat1.*pat2 any number of characters between pat1 and pat2
pat1.*pat2|pat2.*pat1 match both pat1 and pat2 in any order

Greedy here means that the above quantifiers will match as much as possible that’ll also honor the overall RE. Appending a ? to greedy quantifiers makes them non-greedy, i.e. match as minimally as possible. Quantifiers can be applied to literal characters, groups, backreferences and character classes.

Character class Description
[aeiou] Match any vowel
[^aeiou] ^ inverts selection, so this matches any consonant
[a-f] - defines a range, so this matches any of abcdef characters
\d Match a digit, same as [0-9]
\D Match non-digit, same as [^0-9] or [^\d]
\w Match word character, same as [a-zA-Z0-9_]
\W Match non-word character, same as [^a-zA-Z0-9_] or [^\w]
\s Match whitespace character, same as [\ \t\n\r\f\v]
\S Match non-whitespace character, same as [^\ \t\n\r\f\v] or [^\s]
Lookarounds Description
lookarounds custom assertions, zero-width like anchors
(?!pat) negative lookahead assertion
(?<!pat) negative lookbehind assertion
(?=pat) positive lookahead assertion
(?<=pat) positive lookbehind assertion
(?!pat1)(?=pat2) multiple assertions can be specified in any order
  as they mark a matching location without consuming characters
((?!pat).)* Negate a grouping, similar to negated character class
Flags Description
re.IGNORECASE or re.I flag to ignore case
re.DOTALL or re.S allow . metacharacter to match newline character
flags=re.S|re.I multiple flags can be combined using | operator
re.MULTILINE or re.M allow ^ and $ anchors to match line wise
re.VERBOSE or re.X allows to use literal whitespaces for aligning purposes
  and to add comments after the # character
  escape spaces and # if needed as part of actual RE
re.ASCII or re.A match only ASCII characters for \b, \w, \d, \s
  and their opposites, applicable only for Unicode patterns
re.LOCALE or re.L use locale settings for byte patterns and 8-bit locales
(?#comment) another way to add comments, not a flag
(?flags:pat) inline flags only for this pat, overrides flags argument
  flags is i for re.I, s for re.S, etc, except L for re.L
(?-flags:pat) negate flags only for this pat
(?flags-flags:pat) apply and negate particular flags only for this pat
(?flags) apply flags for whole RE, can be used only at start of RE
  anchors if any, should be specified after these flags
Matched portion Description
re.Match object details like matched portions, location, etc
m[0] or entire matched portion of re.Match object m
m[n] or matched portion of nth capture group
m.groups() tuple of all the capture groups’ matched portions
m.span() start and end+1 index of entire matched portion
  pass a number to get span of that particular capture group
\N backreference, gives matched portion of Nth capture group
  applies to both search and replacement sections
  possible values: \1, \2 up to \99 provided no more digits
\g<N> backreference, gives matched portion of Nth capture group
  possible values: \g<0>, \g<1>, etc (not limited to 99)
  \g<0> refers to entire matched portion
(?P<name>pat) named capture group
  refer as 'name' in re.Match object
  refer as (?P=name) in search section
  refer as \g<name> in replacement section

\0 and \100 onwards are considered as octal values, hence cannot be used as backreferences.

Functions to match/extract patterns

Function Description Check if given pattern is present anywhere in input string
  Output is a re.Match object, usable in conditional expressions
  r-strings preferred to define RE
  Use byte pattern for byte input
  Python also maintains a small cache of recent RE
re.compile Compile a pattern for reuse, outputs re.Pattern object
re.sub search and replace
re.sub(r'pat', f, s) function f with re.Match object as argument
re.escape automatically escape all metacharacters
re.split split a string based on RE
re.findall returns all the matches as a list
  if 1 capture group is used, only its matches are returned
  1+, each element will be tuple of capture groups
re.finditer iterator with re.Match object for each match
re.subn gives tuple of modified string and number of substitutions

The function definitions are given below:, string, flags=0)
re.compile(pattern, flags=0)
re.sub(pattern, repl, string, count=0, flags=0)
re.split(pattern, string, maxsplit=0, flags=0)
re.findall(pattern, string, flags=0)
re.finditer(pattern, string, flags=0)
re.subn(pattern, repl, string, count=0, flags=0)

Regular expression examples

As a good practice, always use raw strings to construct RE, unless other formats are required. This will avoid clash of special meaning of backslash character between RE and normal quoted strings.

  • examples for
>>> sentence = 'This is a sample string'

# need to load the re module before use
>>> import re
# check if 'sentence' contains the pattern described by RE argument
>>> bool('is', sentence))
# ignore case while searching for a match
>>> bool('this', sentence, flags=re.I))
>>> bool('xyz', sentence))

# output can be directly used in conditional expressions
>>> if'ring', sentence):
...     print('mission success')
mission success

# use raw byte strings if input is of byte data type
>>> bool('is', b'This is a sample string'))
  • difference between string and line anchors
# string anchors
>>> bool('\Ahi', 'hi hello\ntop spot'))
>>> words = ['surrender', 'unicorn', 'newer', 'door', 'empty', 'eel', 'pest']
>>> [w for w in words if'er\Z', w)]
['surrender', 'newer']

# line anchors
>>> bool('^par$', 'spare\npar\ndare', flags=re.M))

* examples for `re.findall`

# match whole word par with optional s at start and optional e at end
>>> re.findall(r'\bs?pare?\b', 'par spar apparent spare part pare')
['par', 'spar', 'spare', 'pare']

# numbers >= 100 with optional leading zeros
>>> re.findall(r'\b0*[1-9]\d{2,}\b', '0501 035 154 12 26 98234')
['0501', '154', '98234']

# if multiple capturing groups are used, each element of output
# will be a tuple of strings of all the capture groups
>>> re.findall(r'(x*):(y*)', 'xx:yyy x: x:yy :y')
[('xx', 'yyy'), ('x', ''), ('x', 'yy'), ('', 'y')]

# normal capture group will hinder ability to get whole match
# non-capturing group to the rescue
>>> re.findall(r'\b\w*(?:st|in)\b', 'cost akin more east run against')
['cost', 'akin', 'east', 'against']

# useful for debugging purposes as well before applying substitution
>>> re.findall(r't.*?a', 'that is quite a fabricated tale')
['tha', 't is quite a', 'ted ta']
  • examples for re.split
# split based on one or more digit characters
>>> re.split(r'\d+', 'Sample123string42with777numbers')
['Sample', 'string', 'with', 'numbers']

# split based on digit or whitespace characters
>>> re.split(r'[\d\s]+', '**1\f2\n3star\t7 77\r**')
['**', 'star', '**']

# to include the matching delimiter strings as well in the output
>>> re.split(r'(\d+)', 'Sample123string42with777numbers')
['Sample', '123', 'string', '42', 'with', '777', 'numbers']

# use non-capturing group if capturing is not needed
>>> re.split(r'hand(?:y|ful)', '123handed42handy777handful500')
['123handed42', '777', '500']
  • backreferencing within search pattern
# whole words that have at least one consecutive repeated character
>>> words = ['effort', 'flee', 'facade', 'oddball', 'rat', 'tool']

>>> [w for w in words if'\b\w*(\w)\1\w*\b', w)]
['effort', 'flee', 'oddball', 'tool']
working with matched portions
>>>'b.*d', 'abc ac adc abbbc')
<re.Match object; span=(1, 9), match='bc ac ad'>
# retrieving entire matched portion, note the use of [0]
>>>'b.*d', 'abc ac adc abbbc')[0]
'bc ac ad'

# capture group example
>>> m ='a(.*)d(.*a)', 'abc ac adc abbbc')
# to get matched portion of second capture group
>>> m[2]
'c a'
# to get a tuple of all the capture groups
>>> m.groups()
('bc ac a', 'c a')
  • examples for re.finditer
# numbers < 350
>>> m_iter = re.finditer(r'\d+', '45 349 651 593 4 204')
>>> [m[0] for m in m_iter if int(m[0]) < 350]
['45', '349', '4', '204']

# start and end+1 index of each matching portion
>>> m_iter = re.finditer(r'ab+c', 'abc ac adc abbbc')
>>> for m in m_iter:
...     print(m.span())
(0, 3)
(11, 16)
examples for re.sub
>>> ip_lines = "catapults\nconcatenate\ncat"
>>> print(re.sub(r'^', r'* ', ip_lines, flags=re.M))
* catapults
* concatenate
* cat

# replace 'par' only at start of word
>>> re.sub(r'\bpar', r'X', 'par spar apparent spare part')
'X spar apparent spare Xt'

# same as: r'part|parrot|parent'
>>> re.sub(r'par(en|ro)?t', r'X', 'par part parrot parent')
'par X X X'

# remove first two columns where : is delimiter
>>> re.sub(r'\A([^:]+:){2}', r'', 'foo:123:bar:baz', count=1)
  • backreferencing in replacement section
# remove any number of consecutive duplicate words separated by space
>>> re.sub(r'\b(\w+)( \1)+\b', r'\1', 'aa a a a 42 f_1 f_1 f_13.14')
'aa a 42 f_1 f_13.14'

# add something around the matched strings
>>> re.sub(r'\d+', r'(\g<0>0)', '52 apples and 31 mangoes')
'(520) apples and (310) mangoes'

# swap words that are separated by a comma
>>> re.sub(r'(\w+),(\w+)', r'\2,\1', 'good,bad 42,24')
'bad,good 24,42'
  • using functions in replacement section of re.sub
>>> from math import factorial
>>> numbers = '1 2 3 4 5'
>>> def fact_num(n):
...     return str(factorial(int(n[0])))
>>> re.sub(r'\d+', fact_num, numbers)
'1 2 6 24 120'

# using lambda
>>> re.sub(r'\d+', lambda m: str(factorial(int(m[0]))), numbers)
'1 2 6 24 120'
  • examples for lookarounds
# change 'foo' only if it is not followed by a digit character
# note that end of string satisfies the given assertion
# 'foofoo' has two matches as the assertion doesn't consume characters
>>> re.sub(r'foo(?!\d)', r'baz', 'hey food! foo42 foot5 foofoo')
'hey bazd! foo42 bazt5 bazbaz'

# change whole word only if it is not preceded by : or -
>>> re.sub(r'(?<![:-])\b\w+\b', r'X', ':cart <apple -rest ;tea')
':cart <X -rest ;X'

# extract digits only if it is preceded by - and followed by ; or :
>>> re.findall(r'(?<=-)\d+(?=[:;])', '42 foo-5, baz3; x-83, y-20: f12')

# words containing all lowercase vowels in any order
>>> words = ['sequoia', 'subtle', 'questionable', 'exhibit', 'equation']
>>> [w for w in words if'(?=.*a)(?=.*e)(?=.*i)(?=.*o).*u', w)]
['sequoia', 'questionable', 'equation']

# match if 'do' is not there between 'at' and 'par'
>>> bool('at((?!do).)*par', 'fox,cat,dog,parrot'))
# match if 'go' is not there between 'at' and 'par'
>>> bool('at((?!go).)*par', 'fox,cat,dog,parrot'))
  • examples for re.compile

Regular expressions can be compiled using re.compile function, which gives back a re.Pattern object. The top level re module functions are all available as methods for this object. Compiling a regular expression helps if the RE has to be used in multiple places or called upon multiple times inside a loop (speed benefit). By default, Python maintains a small list of recently used RE, so the speed benefit doesn’t apply for trivial use cases.

>>> pet = re.compile(r'dog')
>>> bool('They bought a dog'))
>>> bool('A cat crossed their path'))

>>> remove_parentheses = re.compile(r'\([^)]*\)')
>>> remove_parentheses.sub('', 'a+b(addition) - foo() + c%d(#modulo)')
'a+b - foo + c%d'
>>> remove_parentheses.sub('', 'Hi there(greeting). Nice day(a(b)')
'Hi there. Nice day'

Note: Here are two great web applications to explore regular expressions: - regex101 - regexr