Regex

Regular Expressions

Regex syntax, patterns, recipes, and cross-language reference.

Core Syntax

Character Classes

.           Any character (except newline, unless s/dotall flag)
\d  \D      Digit [0-9] / Non-digit
\w  \W      Word char [a-zA-Z0-9_] / Non-word
\s  \S      Whitespace [ \t\n\r\f\v] / Non-whitespace
[abc]       Character set (a, b, or c)
[^abc]      Negated set (not a, b, or c)
[a-z]       Range (a through z)
[a-zA-Z]    Multiple ranges
\p{L}       Unicode letter (PCRE/Java/JS with u flag/Rust)
\p{N}       Unicode number

Quantifiers

*    +    ?       0+, 1+, 0-1 (greedy)
{3}  {3,} {3,5}  Exact, min, range
*?   +?   ??      Lazy (non-greedy) versions
*+   ++           Possessive (PCRE/Java) - no backtracking

Greedy: match max, then backtrack. Lazy: match min, then expand. Possessive: match max, never backtrack.

Anchors & Boundaries

^  $        Start/end of string (or line with m flag)
\b  \B      Word boundary / Non-word boundary
\A  \Z      Start/end of string (ignores m flag)

Groups & References

(abc)           Capture group
(?:abc)         Non-capturing group
(?<name>abc)    Named capture group (JS/PCRE/.NET)
(?P<name>abc)   Named capture group (Python)
\1              Backreference to group 1
\k<name>        Backreference to named group (JS/PCRE)
(?P=name)       Backreference to named group (Python)
(?>abc)         Atomic group (PCRE/Java) - no backtracking into group

Lookahead & Lookbehind

Zero-width assertions: match a position without consuming characters.

(?=abc)     Positive lookahead: followed by abc
(?!abc)     Negative lookahead: NOT followed by abc
(?<=abc)    Positive lookbehind: preceded by abc
(?<!abc)    Negative lookbehind: NOT preceded by abc

^(?=.*\d)(?=.*[A-Z]).{8,}$   # Password: 8+ chars, digit + uppercase
\d+(?!%)                      # Number NOT followed by %
(?<=\$)\w+                    # Word preceded by $
(?<!bar)foo                   # foo not preceded by bar

Lookbehind limits: JS requires fixed-length. Python allows fixed-length alternation branches. PCRE allows variable-length with \K.

Flags / Modifiers

g   Global (all matches)       i   Case-insensitive
m   Multiline (^/$ per line)   s   Dotall (. matches \n)
u   Unicode                    x   Extended (whitespace ignored, # comments)

Inline: (?i)pattern or scoped (?i:abc).

Non-Greedy (Lazy) Matching

<.*>      Greedy:  <b>bold</b>  → "<b>bold</b>" (first < to LAST >)
<.*?>     Lazy:    <b>bold</b>  → "<b>" then "</b>" (first < to NEXT >)
<[^>]*>   Negated class: same as lazy, faster (no backtracking)

Common Patterns

Validation

# Email (simplified)
^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$

# URL
^https?://[^\s/$.?#].[^\s]*$

# Phone (US)
^\+?1?[-.\s]?\(?\d{3}\)?[-.\s]?\d{3}[-.\s]?\d{4}$

# IPv4 (strict 0-255)
^(?:(?:25[0-5]|2[0-4]\d|[01]?\d\d?)\.){3}(?:25[0-5]|2[0-4]\d|[01]?\d\d?)$

# Date (YYYY-MM-DD)
^\d{4}-(0[1-9]|1[0-2])-(0[1-9]|[12]\d|3[01])$

# UUID v4
^[0-9a-f]{8}-[0-9a-f]{4}-4[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$

# Semver
^v?(0|[1-9]\d*)\.(0|[1-9]\d*)\.(0|[1-9]\d*)(?:-([\da-zA-Z-]+(?:\.[\da-zA-Z-]+)*))?(?:\+([\da-zA-Z-]+(?:\.[\da-zA-Z-]+)*))?$

# Hex color
^#?([0-9a-fA-F]{3}|[0-9a-fA-F]{6}|[0-9a-fA-F]{8})$

# Strong password (8+ chars, upper, lower, digit, special)
^(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[@$!%*?&])[A-Za-z\d@$!%*?&]{8,}$

# Slug
^[a-z0-9]+(?:-[a-z0-9]+)*$

# JWT structure
^[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+$

Extraction

https?://([^/\s]+)                     # Domain from URL
[^/\\]+$                               # Filename from path
-?\d+\.?\d*                            # Numbers (int or decimal)
(\w+)=([^\s&]+)                        # Key=value pairs
\[([^\]]+)\]\(([^)]+)\)                # Markdown links
^\[(\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})\] \[(\w+)\] (.+)$  # Log line

Substitution Patterns

$0 or \0    Entire match         ${name}     Named group (JS)
$1 or \1    First capture group  \g<name>    Named group (Python)

# Swap first/last name
Find:    (\w+) (\w+)         Replace: $2, $1

# camelCase to snake_case
Find:    ([a-z])([A-Z])      Replace: $1_\l$2

# Remove console.log lines
Find:    ^\s*console\.log\([^)]*\);?\s*\n    Replace: (empty)

Language-Specific Usage

JavaScript

const re = /\d+/g;                              // Literal
const re2 = new RegExp('\\d+', 'g');            // Constructor
/^\d+$/.test('123');                             // true
'abc 123 def 456'.match(/\d+/g);                // ['123', '456']

for (const m of 'a1 b2'.matchAll(/([a-z])(\d)/g)) {
  console.log(m[1], m[2]);                      // a 1, b 2
}

'hello'.replace(/[aeiou]/g, c => c.toUpperCase());  // "hEllO"

// Named groups
const m = '2024-01-15'.match(/(?<y>\d{4})-(?<m>\d{2})-(?<d>\d{2})/);
m.groups.y;                                      // '2024'

'$100 €200'.match(/(?<=\$)\d+/);                 // ['100'] (ES2018+)

Python

import re

re.match(r'^\d+', '123abc')              # Match from start
re.search(r'\d+', 'abc123')              # Search anywhere
re.findall(r'\d+', 'a1 b2 c3')           # ['1', '2', '3']
re.sub(r'\d+', 'X', 'a1b2')              # 'aXbX'

pat = re.compile(r'\b\w{4}\b')            # Compiled for reuse
pat.findall('the quick brown fox')        # ['quick', 'brown']

# Named groups (Python syntax)
m = re.search(r'(?P<year>\d{4})-(?P<month>\d{2})', '2024-01')
m.group('year')                           # '2024'
m.groupdict()                             # {'year': '2024', 'month': '01'}

# Verbose mode
pat = re.compile(r'''
    (?P<year>\d{4})   # year
    -(?P<month>\d{2}) # month
    -(?P<day>\d{2})   # day
''', re.VERBOSE)

Go

import "regexp"

re := regexp.MustCompile(`\d+`)             // Panics on bad pattern
re.MatchString("abc123")                     // true
re.FindString("abc 123 def")                 // "123"
re.FindAllString("a1 b2", -1)               // ["1", "2"]
re.ReplaceAllString("a1b2", "X")             // "aXbX"

// Named groups
re := regexp.MustCompile(`(?P<year>\d{4})-(?P<month>\d{2})`)
match := re.FindStringSubmatch("2024-01")
// match[re.SubexpIndex("year")] == "2024"

Go uses RE2: no backreferences, no lookahead/lookbehind. Guarantees linear-time matching.

Rust

use regex::Regex;

let re = Regex::new(r"\d+").unwrap();
re.is_match("abc123");                       // true
re.find("abc 123").unwrap().as_str();         // "123"

// Named captures
let re = Regex::new(r"(?P<y>\d{4})-(?P<m>\d{2})").unwrap();
let caps = re.captures("2024-01").unwrap();
&caps["y"];                                   // "2024"

Rust regex crate uses RE2-like semantics. Use fancy-regex for PCRE features.

Regex Flavor Differences

CLI Tools

grep / sed / awk

# grep
grep 'pattern' file.txt              # Basic regex (BRE)
grep -E 'extended|regex' file.txt    # Extended regex (ERE)
grep -P '\d{3}' file.txt             # Perl regex (GNU grep only)
grep -oirn 'pattern' src/            # -o only match, -i case, -r recursive, -n line nums

# sed
sed 's/old/new/g' file.txt           # Replace all
sed -E 's/([0-9]+)/[\1]/g' file.txt  # Extended regex with groups
sed '/pattern/d' file.txt            # Delete matching lines

# awk
awk '/pattern/ {print $0}' file.txt            # Print matching lines
awk '$3 ~ /^[0-9]+$/ {print $1}' file.txt      # Field regex match

ripgrep (rg)

rg '\d{3}-\d{4}' src/               # Recursive, fast, .gitignore-aware
rg -t py 'import' .                 # Filter by file type
rg -U 'struct.*\n.*field' .         # Multiline match
rg --pcre2 '(?<=\$)\d+' .           # PCRE2 features
rg -r '$1' '(\w+)@\w+' emails.txt   # Replace (output only)

VS Code Find & Replace

# Enable regex: click .* button or Alt+R
# Capture groups in replace: $1, $2, etc.

Find:    (\w+):\s*(\w+)        Replace: $2: $1
Find:    console\.log\(.*?\)   Replace: (empty)

# Case modifiers in replacement
\u  Next char uppercase    \U  Uppercase until \E
\l  Next char lowercase    \L  Lowercase until \E

Find:    const (\w+)           Replace: const \l$1

Performance & Security

Catastrophic Backtracking

Nested quantifiers cause exponential time on non-matching input:

# DANGEROUS                    # SAFE alternative
(a+)+b                         a+b
(a|a)+b                        [a]+b
(\w+)*@                        \w+@

ReDoS (Regular Expression Denial of Service)

Mitigations:

• Use RE2-based engines (Go, Rust) which guarantee linear time
• Set timeouts on regex execution
• Validate/limit input length before matching
• Never build regex from untrusted user input

Unicode & International Text

\p{L}+              # Any Unicode letter (PCRE, Java, Rust, JS with u flag)
[\u4e00-\u9fff]+     # CJK Unified Ideographs
\p{Emoji}            # Emoji (modern engines)

# JS: always use u flag for Unicode
/\p{L}+/u            # Correct - \w is still ASCII-only even with u

# Python 3: re handles Unicode by default
re.findall(r'\w+', 'cafe\u0301 nai\u0308ve')  # ['cafe\u0301', 'nai\u0308ve']

Multiline & Dotall Modes

/^start/m             # Multiline: ^ matches start of any line
/start.*end/s         # Dotall: . matches \n
/start[\s\S]*?end/    # Cross-line match without s flag

Testing & Debugging

• regex101.com - Test PCRE, JS, Python, Go with live explanation and debugger
• regexr.com - Visual tester with community patterns
• Python: re.compile(pattern, re.DEBUG) prints parse tree
• Build incrementally: start simple, add complexity, test each step

When NOT to Use Regex

• HTML/XML - Use a parser (BeautifulSoup, cheerio, DOMParser)
• JSON - Use JSON.parse() or equivalent
• Balanced nesting - Brackets, recursive grammars need parser generators
• Strict email/URL validation - Full RFC specs are too complex; use a library
• 200+ char patterns - If you can't read it, use code instead

Reference

For more recipes and testing tips: references/recipes.md