Regular expressions

Regular expression syntax

A regular expression, or regex, describes a set of strings by their structure. The re module is part of the standard library: https://docs.python.org/3.14/library/re.html. First, describe valid examples in words, then build the expression and test it on valid and invalid strings.

Ordinary letters match themselves. A dot matches any character except a newline by default; \. matches a literal dot. Square brackets specify one character from a set: [0-9], [ABC]. The expression [^0-9] matches one character outside the set of ASCII digits. A hyphen inside a character class may specify a range, so escape a literal hyphen or place it at an edge.

Classes, repetition, and alternatives

\d matches a Unicode decimal digit, \w a letter, digit, or underscore, and \s a whitespace character. These sets are broader than ASCII digits or an ordinary space. For a phone number written with digits 0–9, explicitly use [0-9]. For the Ukrainian alphabet, [а-я] is incomplete: “і”, “ї”, “є”, and “ґ” are also needed; the range also includes letters outside the Ukrainian alphabet. It is better to specify the exact alphabet explicitly or check against a set of allowed letters.

The quantifier * means zero or more repetitions, + one or more, ? zero or one, and {m,n} from m to n inclusive. They apply to the preceding atom: a character, class, or group. ab+ matches the letter a followed by one or more b characters; (ab)+ repeats the pair. The alternative cat|dog selects one of two options.

The anchor ^ marks the start of the string, while $ marks the end or the position before a final newline. For strict validation of the entire input, fullmatch is therefore more convenient than combining match with $. The word boundary \b is defined by the transition between \w and \W, not by a Ukrainian dictionary: an apostrophe may create a boundary within a word.

Greedy and lazy repetition

Repetition is greedy by default: it tries to consume as much as possible while still allowing the rest of the expression to match. An additional ? makes repetition lazy. Neither mode, however, turns regex into a complete parser for nested markup.

import re

text = "<b>Olia</b><i>Ivan</i>"
print(re.findall(r"<.*>", text))
print(re.findall(r"<.*?>", text))

The first result is a list containing the entire string; the second is ['<b>', '</b>', '<i>', '</i>']. Tags with quotes, nested elements, and malformed HTML require a specialized parser. This learning demonstration of quantifier differences is not an HTML converter.

Groups, flags, and the Match object

Ordinary parentheses capture a fragment. A named group (?P<name>...) lets you access it by name; a noncapturing group (?:...) groups an expression without a separate result. Reuse previously matched text with a backreference such as \1 or (?P=name).

import re

pattern = re.compile(r"(?P<code>\+380) ?(?P<num>[0-9]{9})")
match = pattern.fullmatch("+380 671234567")
if match is not None:
    print(match.group("code"), match.group("num"))
    print(match.span("num"))
print(pattern.fullmatch("+380 671234567\n") is None)

Output: +380 671234567, then (5, 14) and True. The span method returns a half-open interval: the start is included and the end excluded. group(0) contains the entire match, groups() a tuple of ordinary groups, and groupdict() a dictionary of named groups. Before accessing groups, check that the result is not None.

Figure 6.5. Components of an expression for a learning phone-number format

Lookahead and lookbehind

The assertions (?=...) and (?!...) require a fragment to be present or absent ahead without consuming it. For example, (?=.*[0-9]) checks for a digit later on the same line. Similarly, (?<=...) and (?<!...) check preceding text. In the re module, a lookbehind pattern must have a fixed length: an arbitrary (?<=a+) is not allowed. For several validation rules, performing several clear checks and explaining each error is often more convenient than constructing one long expression.

re.IGNORECASE disables case sensitivity; re.MULTILINE changes ^ and $ to mean line starts and ends. It does not let a dot match a newline; that requires re.DOTALL. re.VERBOSE lets you split an expression across lines and add comments; ordinary spaces outside character classes are then ignored. To require a space, use [ ] or escaping.

Screenshot

Cursor in re.compile pattern, Alt+Enter, Check RegExp.

Figure 6.6. Checking a learning regular expression in PyCharm

Searching, extracting, and transforming data

re.match checks the beginning of the text, re.search finds the first match anywhere, and re.fullmatch checks the entire text. finditer returns an iterator of Match objects, convenient for groups and positions. findall returns a list whose element shape depends on the number of capturing groups: the entire match, one group, or a tuple of groups.

import re

text = "A12 B7 C305"
print(re.findall(r"[A-Z][0-9]+", text))
print(re.findall(r"([A-Z])([0-9]+)", text))
print(re.split(r"[;,]\s*", "one, two;three"))

Output:

['A12', 'B7', 'C305']
[('A', '12'), ('B', '7'), ('C', '305')]
['one', 'two', 'three']

If parentheses are needed only for an alternative, use (?:...) to avoid changing the findall result. In re.split, capturing groups add separators to the result, which is useful for a parser but may be surprising during simple text splitting.

Replacing while preserving meaning

re.sub accepts a replacement string or a function that receives a Match and returns new text. A group reference in a Python replacement string takes the form \g<name> or \g<1>. Do not confuse it with the $1 syntax in an editor's replacement dialog. A replacement function lets you convert a type, calculate a value, and check semantics.

import re


def double(match: re.Match[str]) -> str:
    return str(int(match.group()) * 2)


print(re.sub(r"[0-9]+", double, "Cabinet 12, desk 3"))
print(re.sub(r"([A-Z])([0-9]+)", r"\g<2>-\g<1>", "A12"))

Output: Cabinet 24, desk 6 and 12-A. A date such as 31.02.2026 matches a simple digit pattern but does not exist in the calendar. After checking the form, call datetime.date and catch ValueError. Similarly, regex does not confirm that a mailbox exists or a phone number is available.

Screenshot

Ctrl+R; enable Regex; search date groups; replacement $3-$2-$1.

Figure 6.7. Previewing a date-format replacement in the editor

In PyCharm, check one match first, then review all replacements, and only then apply them. The editor dialog may use a different regex engine from Python's re; the final program check is always performed by its interpreter. Help: https://www.jetbrains.com/help/pycharm/regular-expressions.html.