Testing a regex once, then shipping it in JavaScript, Python, Java and Go

A regular expression that works in a browser console does not always work in Python, and the one that works in Python may quietly
misbehave in Go. The syntax overlaps enough to feel identical and differs exactly where it hurts: escaping, flags, named groups and
how the replacement string refers to captures.

Test against real text, not the example

Paste the actual log line, the actual CSV row, the actual email header. Sample data hides the cases that break patterns: trailing
whitespace, unicode punctuation, a second delimiter on the same line. The tester highlights every match live, so a pattern that
suddenly matches nothing tells you more than a green checkmark ever would.

Read the groups table before trusting the match

A match count of 12 means little on its own. The groups table shows what each capture actually caught, index by index, so you can
see the difference between a group that captured an empty string and one that never participated in the match. That distinction is
the source of most "why is my variable undefined" bugs downstream.

Replace mode is a different question

Search and replace have different failure modes. In replace mode you can check that $1 lands where you expect, that a greedy
quantifier is not swallowing the delimiter, and that a global flag is doing what you assumed.

Exporting to another language

The escaping rules are where portability breaks:

• JavaScript — literal syntax with flags after the closing slash.
• Python — raw strings (r"...") so backslashes survive; flags are arguments, not suffixes.
• Java — the pattern lives inside a normal string, so every backslash doubles.
• Go — regexp.MustCompile with a backtick string, and no lookahead support in the standard engine.
• PHP — delimiters around the pattern, flags outside them.
• C# and Ruby each have their own flag spelling.

Exporting a tested pattern instead of retyping it removes the entire class of "it worked in the tester" bugs.

A pattern library beats memory

Email, URL, IPv4 and IPv6, ISO dates, UUIDs, hex colours, semver, JWT shape, strong-password rules — these get rewritten badly
thousands of times a day. Starting from a known-good pattern and narrowing it is faster and safer than assembling one from
character classes at 5pm.