1. Our aim is not to match the accepted syntax of all parses. Our goal is not to match Pandoc in this way.

2. The PR as written does not do what is intended either. It may not fail with haskell {.numberLines} is given, but it does not add the numberLines class as specified. Accepting the syntax and parsing the syntax are two separate things.

3. This PR also allows other oddities which we do not want. For instance:

   ```{.haskell} {.numberLines}
   qsort [] = []
   ```

Unfortunately, I cannot accept this PR.

Maybe a bad example. The behavior between this pr and the current case is the same for

```{.haskell} {.numberLines}
qsort [] = []
```

since attributes match is greedy
Both match: attrs=".haskell} {.numberLines"

This might break fewer cases than you think, just allows different ordering for attributes

Both match: attrs=".haskell} {.numberLines"

I'm not sure how that addresses my three statements. Using the case I showed does not add any of these attributes.

In addition to my three earlier statements, it appears this has not yet been tested as it:

1. Breaks at least one test.
2. The cases it claims to allow, while passing through the parser, is not processed properly as attributes or options.

1. Verified that existing tests do not break
2. changed the regex to preserve attributes and options being exclusive
3. Handle explicit lang declaration
4. added additional test

I understand that there does not have to be parity, but it can't hurt (:

I am actually phasing out that old style of testing. SuperFence tests should go here: https://github.com/facelessuser/pymdown-extensions/blob/main/tests/test_extensions/test_superfences.py.

This doesn't mean I've decided to add this functionality, but I will at least consider it if you move the tests and ensure it is all working. Please make sure you've considered edge cases.

I'll think about this and give a response once I've decided.

I do think I'm okay with the idea. It seems like the changes will be minimal.

1. Update to new style tests.
2. Keep option/attribute handling in their respective functions.
3. Ensure lint and tests are working, and I think I'm okay with it.

Hey, @dmadisetti as I've warmed up to this idea, I was digging in and decided to clean up the regex and such, so I've issued a new PR that cleans up my requests and references you and this PR. That PR will supersede this one.

I think when it became apparent that we could do this with very little impact, it kind of became hard to argue against this. Admittedly, I think many will find this new format preferable.

Thanks for getting this rolling and pushing for it, I do think it'll be a good addition.

Much appreciated! Looking forward to the update