Welcome to Tony's Notebook

How to make Python regexes a little less greedy

There are these little things that once you learn about them you wonder how you ever did without them. The Python non-greedy modifier definitely falls into that category. I spent far too much time and consumed far too much coffee figuring out how to get a certain matching operation to work in some recent code. It worked though. Mostly. Until it didn't. There's always that "edge case" where the regex fails and you are back to the wonderful world of convoluted character sets and weird modifiers that is regex. When my carefully thought out (really blood, sweat and tears hacked out) regex failed yet again I decided enough was enough.

Here was the problem:

title: This is some title
description: This is the description

Some content...

This is a simplified version of the metadata that each piece of content on the site has. What the code needs to do is extract the metadata and the content.

This seems straightforward. You might come up with:


We can simplify that but getting rid of the extra new lines in our captured text by using the .strip() function in Python so you end up with:


The metadata drops into the first () and the content into the second () and there are rainbows and unicorns and all is good in the world. Until this happens...

title: This is some title
description: This is the description

Some content...

Item | Description
--- | ---
A | A thing
B | Another thing

Some more content...

And now there are tears because it all goes horribly wrong. You see Python regexes are downright greedy. They try to match as much text as possible. Which means your regex now matches right down to the first --- in the Markdown table. This is where you probably start trying all kinds of variations on your regex to restrict the match to only the metadata. But there's an easy little fix...


The secret is that addition of the ? operator. Like many operators it has many functions but when it's next to * it means "don't be so darn greedy".

Here's the actual code where I use it:

def extract_parts(source):
    m = re.search(r'---([\s\S]*?)---([\s\S]*)', source, re.MULTILINE)
    metadata = m.group(1)
    markdown = m.group(2)
    return metadata.strip(), markdown.strip()

This little ? turns out to be hellishly useful. For example:

<p>Para 1</p><p>Para 2></p>

If you only want the first para you could use <p>.*?</p>, and you'd only match the first para.

You can test this out with the following code:

import re

s = "<p>para 1</p><p>para 2</p>"

m = re.search(r'<p>.*</p>', s)

m = re.search(r'<p>.*?</p>', s)

Yes. Useful indeed. Once you know about the non-greedy operator you'll wonder how you ever did without it!