First up, cleaning up syntax tree so I can finally publish the version based on prism and extract out a bunch of functionality into other gems.

I've taken a half-year break from open-source to focus on the home front, celebrating my third kid being born earlier this year. I'm slowly working my way back to the internet here. If you've been waiting for me to respond, thanks for your patience!.

It makes me unreasonably frustrated that x86-64 uses AH for the remainder in 8-bit unsigned division. I'm sure there's a good hardware reason, but I wish it were DH to be consistent.

This was the dream the whole time, and it’s really nice that when it finally happens it’s relatively quiet. That means it’s actually working.

If you are a Ruby developer, then any time Ilya publishes something you need to pay attention. Very impressive work went into this post.

RT @_byroot: Here we go. That one I really needed to get of my chest for the better part of a year, so I feel relieved now. It's really lo….

Ruby 3.4.2 is out, which includes fixes for just about everything I know of related to Prism. We’re into the truly esoteric code now, and it appears that everything is working!.

This is what I'm talking about. Looks like the TLS area is cleaned up before the pthread_key_create callback is called, but only on macOS. Works as expected on Windows and Linux. Annoying. I guess I can't rely on automatic TLS within pthread callbacks.

This is incredibly niche, but maybe someone knows. I'm seeing pthread_key_create destruction callbacks fired _after_ the thread has cleaned up thread_local variables in C11, but only on macOS. On all other platforms, TLS is unchanged. Wat?.

Truly and completely heartbroken for this loss. Noah was inspirational in his desire to make Ruby and Rails accessible. He was always willing to chat, even before we got to work together. This community will deeply miss him and his myriad contributions. I will miss him a lot.

Just released Prism v1.3.0. * Prism::StringQuery.* Prism::Relocation.* A simpler build system.* Much more documentation.* More invalid code rejected.* Better error messages.* A bunch more stuff. Last release before Ruby 3.4!.

It's certainly much more verbose. Occasionally it reads like go, because you're constantly checking for errors. But maybe that's okay?.

The first thing I've noticed in CPython is that error handling is soooooo clean. Basically every function can return NULL, and it's expected that you will have set an error in that case. Compared to setjump/longjump, this is much easier to understand.

Claiming evidence of a multiverse is certainly one way to get your blog post out there.

Maybe I'll do this at another point in the year so that I can keep the cadence going either way. As a brief teaser to get you interested, check out this super interesting video about Brzozowski derivatives:.

So I was more than 50% of the way through writing "Advent of Regexp" and realized I wanted to have a full implementation of it working as part of the series. I got distracted and well. I didn't finish it yet. As such, I apologize but I'm going to delay this whole series.

That being said — I spent an hour this morning working on Prism issues and PRs, so I'll be around.

After a wonderful and educational 4.5 years, I have moved on from my position at Shopify. This week, I've started a new job working on the Python runtime team at Meta. I'm looking forward to learning about a new JIT, removing the GIL, and helping to make CPython better.

The Ruby section of this blog post is giving me stress flashbacks to when we were just trying to match the Ripper lexer in Prism.

RT @eregontp: "Benchmarking Ruby parsers is trickier than one might expect, but overall the result is clear: . " (new blog post). https://t….