ChatGPT for Regular Expressions

I have this project that I’ve been chipping away at for years. I forget about it for long periods of time, and then decide to see where I left off.

It’s a markdown library in Elixir.

I’m at the point now where most things work and I’m throwing chunks of text at it to see if it converts it correctly.

Naturally that exposes some bugs around things I missed. The trick is that because I have these long stretches where I don’t work on the project, I forget things.

I had a bug in my logic that built hyperlinks. I hadn’t considered a hyphen in the link text, so the regex didn’t match and the conversion was ignored.

The regex before the bug fix

~r/\[(?<text>[\w\.?\s]+)\]\((?<url>http[s]?\:\/\/\w+[\.\w+]+\w+)\)/i

I didn’t remember what this did, and didn’t really want to walk through step by step. Decided to see what ChatGPT had to say

I asked ChatGPT explain ~r/\[(?<text>[\w\.?\s]+)\]\((?<url>http[s]?\:\/\/\w+[\.\w+]+\w+)\)/i

This was the response

This is a regular expression written in Ruby (or a similar language) that captures Markdown-style links in a string. Let’s break it down:

Pattern Breakdown

~r/.../i

• ~r/.../ → This is the syntax for a regex literal in Elixir or Ruby.

• /i → Case-insensitive flag (matches uppercase and lowercase letters without distinction).

\[(?<text>[\w\.?\s]+)\]

• \[ → Matches a literal opening square bracket ([).

• (?<text>[\w\.?\s]+) → Named capture group (text):

  • [\w\.?\s]+ → Matches:

    • \w → Any word character (letters, digits, underscores).

    • \. → A literal period (.).

    • ? → An optional question mark.

    • \s → Any whitespace character (spaces, tabs).

  • Example match: "Click here", "My link".

• \] → Matches a literal closing square bracket (]).

\((?<url>http[s]?\:\/\/\w+[\.\w+]+\w+)\)

• \( → Matches a literal opening parenthesis (().

• (?<url>http[s]?\:\/\/\w+[\.\w+]+\w+) → Named capture group (url):

  • http[s]? → Matches “http” or “https”.

  • \:\/\/ → Matches :// (the standard separator in URLs).

  • \w+ → Matches one or more word characters (e.g., example).

  • [\.\w+]+ → Matches:

    • \. → A literal dot (.).

    • \w+ → One or more word characters (e.g., com, org, net).

  • \w+ → Ensures the URL ends in a valid TLD (e.g., .com, .net).

• \) → Matches a literal closing parenthesis ()).

Example Matches

Summary

This regex identifies Markdown-style links and extracts:

• The link text (inside [...]).

• The URL (inside (...)).

Let me know if you need modifications or further explanations! 🚀

The biggest hurdle was that I don’t use named capture groups that often, and frankly, I was lazy and didn’t want to dig into how that worked.

In a matter of probably seconds, I got the info I needed, and knew what the fix was. Simply add \- to the first optional group.

In my opinion, this the perfect use of AI for a developer. I knew the information, I was just looking up what I needed to be reminded of. Sort of the new way to look up docs. In fact, I use it for that all the time.

If I didn’t understand how to build a regex for extracting hyperlinks from markdown, and used ChatGPT, that’s a problem. That level of copy and paste is as bad as it’s always been. Maybe worse now because ChatGPT has return pretty complex solutions with very little work.