When to Abandon Regex

Signs you should stop

Some patterns are fighting the wrong battle. Signs your regex has crossed the line:

Pattern is over 100 characters and growing. You're parsing, not matching shapes.
You need to track context across the input. "Inside a quoted string, ignore X" — that's lexer work, not regex.
You need to count or balance. Nested parens, balanced quotes — non-regular by definition.
You need recursion. If (?R) is on the table, you've outgrown regex.
You're chaining 3+ regex passes. Probably want a parser.
The pattern is fragile to formatting changes. One stray space breaks it.
You can't read it out loud. Production-ready regex should be readable.

What to reach for instead

Parsers for nested structures:

HTML / XML → BeautifulSoup, lxml, DOMParser
JSON → json.loads, JSON.parse
YAML → PyYAML, js-yaml
Source code → AST (Python ast, JS Babel parser)
Custom DSLs → write a real lexer/parser (PLY, Lark, nom, ANTLR)

Specialized libraries for known formats:

URLs → urllib.parse, new URL()
Dates → datetime, dateutil, date-fns
Phone → libphonenumber
IPs → ipaddress, net.IP
Versions → packaging.version, semver

The reframing

Reaching for the right tool isn't admitting defeat — it's senior judgment. Beginners reach for regex for everything because it's the hammer they know. Experienced engineers know which problems regex actually solves and which it just postpones.

Code

Right tool comparisons·python

# WRONG — parsing JSON with regex
import re
# trying to extract 'name' from JSON-like text
re.findall(r'"name":\s*"([^"]+)"', '{"name": "pippa", "data": {"name": "x"}}')
# ['pippa', 'x']  — caught both 'name' fields, but breaks on escaped quotes

# RIGHT — use json
import json
data = json.loads('{"name": "pippa", "data": {"name": "x"}}')
print(data['name'])  # 'pippa'
print(data['data']['name'])  # 'x'

# WRONG — parsing dates by hand
re.match(r'(\d{4})-(\d{2})-(\d{2})', '2026-05-04')
# Then parse, then validate, then handle leap years...

# RIGHT — use datetime
from datetime import datetime
dt = datetime.strptime('2026-05-04', '%Y-%m-%d')
# Validates, parses, handles edge cases

# WRONG — parsing source code
re.findall(r'def\s+(\w+)\(', code)
# Fragile to comments, strings, decorators

# RIGHT — use AST
import ast
tree = ast.parse(code)
for node in ast.walk(tree):
    if isinstance(node, ast.FunctionDef):
        print(node.name)

Signs you should stop

What to reach for instead

The reframing

Code

External links

Exercise

Progress

Comments 0