Fix numbered group binding

[max-sixty]

Ref #69 should re.replace("ab","$1_$2") equal "a_b" or "ab"?

In this test $42a binds as a single group reference; which would suggest $1_ did the same (I think?)

Here it looks like any non-number character following the number does not bind, and so $42a would only bind the 42. I'm no regex expert but a cursory look suggests there's are various approaches to valid number references: https://www.regular-expressions.info/replacebackref.html

As an initial attempt, I changed the test case referenced above.

Let me know if this is reasonable

[max-sixty]

Reading the docs, I actually think the current implementation is correct:

The longest possible name is used. e.g., $1a looks up the capture group named 1a and not the capture group at index 1. To exert more precise control over the name, use braces, e.g., ${1}a.

In the motivating example, re.replace("ab","$1_$2") should just be b, because 1_ is captured and doesn't match any names.

If that's correct: should we keep the existing behavior or make the change? I have a patch that adds a couple tests in the former case

[BurntSushi]

Unfortunately, I think we're going to want to keep the existing behavior. Making this change now would be a breaking change. If the docs and implementation differed (where the docs stated the ideal behavior), then I might be willing to call this an implementation bug and fix it. But since the docs and implementation match, I don't think we can get away with fixing this.

So I think perhaps some more test on the current behavior would be great. I will mark #69 as a regex2 issue.

[max-sixty]

OK no problem, I'll update with a couple of tests.

Any other issues where a beginner can contribute? This was the easiest I could find where you pointed to a function and specified the change...

[max-sixty]

@BurntSushi this ended up not being much - but still worth merging? Anything else needed?

[max-sixty]

Gentle ping to push a notification!

[BurntSushi]

Great, thanks!