Regular expressions

In Python we access regular expressions through the "re" library. We call methods like re.match() to test for patterns.

With match and search we evaluate regular expressions. More advanced methods like groupdict can process groups. Findall handles multiple matches—it returns a list.

`Match` example

This program uses a regular expression in a loop. It applies a for-loop over the elements in a list. In the loop body, we call re.match().

Then We test this call for success. If it was successful, groups() returns a tuple containing the text content that matches the pattern.

import re

# Sample strings.
list = ["dog dot", "data day", "no match"]

# Loop.
for element in list:
    # Match if 2 words starting with letter "d."
    m = re.match(r"(d\w+)\W(d\w+)", element)

    # See if success.
    if m:
        print(m.groups())('dog', 'dot')
('data', 'day')

Search

This method is different from match. Both apply a pattern. But search attempts this at all possible starting points in the string. Match just tries the first starting point.

So Search scans through the input string and tries to match at any location. In this example, search succeeds but match fails.

import re

# Input.
value = "voorheesville"

m = re.search(r"(vi.*)", value)
if m:
    # This is reached.
    print("search:", m.group(1))

m = re.match(r"(vi.*)", value)
if m:
    # This is not reached.
    print("match:", m.group(1))
search: ville

`Split`

The re.split() method accepts a pattern argument. This pattern specifies the delimiter. With it, we can use any text that matches a pattern as the delimiter to separate text data.

Here We split the string on one or more non-digit characters. The regular expression is described after the script output.

import re

# Input string.
value = "one 1 two 2 three 3"

# Separate on one or more non-digit characters.
result = re.split(r"\D+", value)

# Print results.
for element in result:
    if element != "":
        print(element)1
2
3

Findall

This is similar to split(). Findall accepts a pattern that indicates which strings to return in a list. It is like split() but we specify matching parts, not delimiters.

Here We scan a string for all words starting with the individual letters "dp," and with one or more following word characters.

import re

# Input.
value = "abc 123 def 456 dot map pat"

# Find all words starting with d or p.
list = re.findall(r"[dp]\w+", value)

# Print result.
print(list)
['def', 'dot', 'pat']

Finditer

Unlike re.findall, which returns strings, finditer returns matches. For each match, we call methods like start() or end(). And we can access the value of the match with group().

import re

value = "123 456 7890"

# Loop over all matches found.
for m in re.finditer(r"\d+", value):
    print(m.group(0))
    print("start index:", m.start())123
start index: 0
456
start index: 4
7890
start index: 8

Start, end

We can use special characters in an expression to match the start and end of a string. For the start, we use the character "^" and for the end, we use the "$" sign.

Here We loop over a list of strings and call re.match. We detect all the strings that start or end with a digit character "\d."

Tip The match method tests from the leftmost part of the string. So to test the end, we use ".*" to handle these initial characters.

import re

list = ["123", "4cat", "dog5", "6mouse"]
for element in list:

    # See if string starts in digit.
    m = re.match(r"^\d", element)
    if m:
        print("START:", element)

    # See if string ends in digit.
    m = re.match(r".*\d$", element)
    if m:
        print("  END:", element)START: 123
  END: 123
START: 4cat
  END: dog5
START: 6mouse

Or, repeats

Here we match strings with three letters or three dashes at their starts. And the final three characters must be digits. We use non-capturing groups with the "?:" syntax.

And We use the "3" codes to require three repetitions of word characters or hyphens.

Finally We specify digit characters with the code "\d" and the metacharacter "$" to require the end of the string.

import re

values = ["cat100", "---200", "xxxyyy", "jjj", "box4000", "tent500"]
for v in values:

    # Require 3 letters OR 3 dashes.
    # ... Also require 3 digits.
    m = re.match(r"(?:(?:\w{3})|(?:\-{3}))\d\d\d$", v)
    if m:
        print("  OK:", v)
    else:
        print("FAIL:", v)  OK: cat100
  OK: ---200
FAIL: xxxyyy
FAIL: jjj
FAIL: box4000
FAIL: tent500

Named groups

A regular expression can have named groups. This makes it easier to retrieve those groups after calling match(). But it makes the pattern more complex.

Here We can get the first name with the string "first" and the groups() method. We use "last" for the last name.

import re

# A string.
name = "Roberta Alden"

# Match with named groups.
m = re.match(r"(?P<first>\w+)\W+(?P<last>\w+)", name)

# Print groups using names as id.
if m:
    print(m.group("first"))
    print(m.group("last"))Roberta
Alden

Groupdict

A regular expression with named groups can fill a dictionary. This is done with the groupdict() method. In the dictionary, each group name is a key.

And Each value is the data matched by the regular expression. So we receive a key-value store based on groups.

Here With groupdict, we eliminate all references to the original regular expression. We can change the data to dictionary format.

import re

name = "Roberta Alden"

# Match names.
m = re.match(r"(?P<first>\w+)\W+(?P<last>\w+)", name)

if m:
    # Get dict.
    d = m.groupdict()

    # Loop over dictionary with for-loop.
    for t in d:
        print("  key:", t)
        print("value:", d[t])  key: last
value: Alden
  key: first
value: Roberta

Comment

Sometimes a regular expression is confusing. A comment can be used to explain a complex part. One problem is the comment syntax may be confusing too—this should be considered.

Here We see that a Regex comment starts with a "#" character (just like in Python itself).

import re

data = "bird frog"

# Use comments inside a regular expression.
m = re.match(r"(?#Before part).+?(?#Separator)\W(?#End part)(.+)", data)

if m:
    print(m.group(1))
frog

Not-followed-by

We use a negative match pattern to ensure a value does not match. In this example, we match all the 3-digit strings except ones that are followed by a "dog" string.

Tip This is called a "negative lookahead assertion." It may be clearer to filter out results in Python code after matching.

import re

data = "100cat 200cat 300dog 400cat 500car"

# Find all 3-digit strings except those followed by "dog" string.
# ... Dogs are not allowed.
m = re.findall(r"(?!\d\d\ddog)(\d\d\d)", data)

print(m)
['100', '200', '400', '500']

Benchmark, search

Regular expressions often hinder performance in programs. I tested the in-operator on a string against the re.search method.

Version 1 This version of the code uses the in-operator to find the letter "x" in the string.

Version 2 Here we use re.search (a regular expression method) to find the same letter.

Result I found that the in-operator was much faster than the re.search method. For searching with no pattern, prefer the in-operator.

import time
import re

input = "max"

if "x" in input:
    print(1)

if re.search("x", input):
    print(2)

print(time.time())

# Version 1: in.
c = 0
i = 0
while i < 1000000:
    if "x" in input:
        c += 1
    i += 1

print(time.time())

# Version 2: re.search.
i = 0
while i < 1000000:
    if re.search("x", input):
        c += 1
    i += 1

print(time.time())1
2
1381081435.177
1381081435.615 [in        = 0.438 s]
1381081437.224 [re.search = 1.609 s]

Benchmark, match

In another test I rewrote a method that uses re.match to use if-statements and a for-loop. It became much faster.

Version 1 In this experiment, stringmatch returns after finding an invalid length or an invalid start character.

Version 2 This version of the code uses a regular expression instead of custom Python looping logic.

Result In 2024, it is faster to use a for-loop and test characters with Python statements.

import re
import time

def stringmatch(s):
    # Check for "CA+T" with if-statements and loop.
    if len(s) >= 3 and s[0] == 'C' and s[len(s) - 1] == 'T':
        for v in range(1, len(s) - 2):
            if s[v] != 'A':
                return False
        return True
    return False

def stringmatch_re(s):
    # Check for "CA+T" with re.
    m = re.match(r"CA+T", s)
    if m:
        return True
    return False

print(time.time())

# Version 1: use string loop with if-statement.
for i in range(0, 10000000):
    result = stringmatch("CT")
    result = stringmatch("CAAT")
    result = stringmatch("DOOOG")

print(time.time())

# Version 2: use re.match.
for i in range(0, 10000000):
    result = stringmatch_re("CT")
    result = stringmatch_re("CAAT")
    result = stringmatch_re("DOOOG")

print(time.time())1726672414.6579642
1726672417.3540194     stringmatch    =  2.69 s
1726672428.6217597     stringmatch_re = 11.26 s

Summary

A regular expression is often hard to correctly write. But when finished, it is shorter and overall simpler to maintain. It describes a specific type of logic.