Retrofitting any non trivial code that wasn't using type hinting or thinking about type hinting friendly code is a generally unfun experience.

Like writing testable code there is a design you have to follow for type hintable code.

There is lots of good code you can write outside these designs.

Taking off nearly well over 3 hours late, hopefully I get to catch a bit of today's programming.

Not used them but heard good things about docling and sycamore

Having an AI produce an MRE one isn't given, could be great if done right.

Also I find LLM search useful for finding a GitHub issue I remember but GitHub and Google search fail me. So better duplicate suggestions seems like a good feature too.

This seems pretty reasonable to me, they are giving you a way to report a bug, but they are tracking confirmed bugs separately.

Look at pip for an alternative situation, issues will be reported that no one will ever reply to, or possibly even read.

I've never been a fan of TDD, but in this context it made a lot of sense, because I don't trust the LLM (I probably shouldn't trust myself either but I usually do)

It would have taken me longer with this approach doing it without an LLM, but I wouldn't have used this approach

I've not read the PEP in detail and in the long discourse thread how these template strings should work changed over the course of the discussion

I was surprised at the acceptance notice as I didn't realize consensus has been achieved. I tune out after 100 replies if I'm not invested in the outcome

You can use style='{': docs.python.org/3/library/lo...

To get str.format() style formatting.

Isn't that normal??