tekknolagi

Do it yourself: https://github.com/tekknolagi/db.py

I made up some stuff for tests and copied some tables from their demos

Fenster is written in C and has an optional C++ wrapper (it's all in the file)

Or PocketPy Or fe

What do you mean same data?

When is it getting open sourced?

Anything using codecs, especially CJK. The amount of scary stuff we found when reimplementing it was... Well, scary

There's also Cabal.chat which works based on similar tech

Not that I know of :/ I think some stuff in the article applies broadly to content addressing and some is pretty implementation specific. I haven't kept up with the field.

Yesterday I found this and ported it to Python: https://github.com/tekknolagi/db.py

It's ~150LOC but doesn't have aggregates (yet)

Hi! Author here. Funny to see my term paper posted. A lot has happened in the space since then and I haven't kept up as well as I would have liked.

Author here! Let me know if you have other similar implementations!

It's probably written in a programming language with the same terms, too. (Bolin)

Not if you had LCD video games like on a Gameboy.

I built a compiler (https://bernsteinbear.com/blog/compiling-ml-models/) for micrograd (https://github.com/karpathy/micrograd/) to better understand how it worked

I have a rather tricky time reading the guest language syntax. Is there a good reason you didn't use "normal" dot+parens method call syntax? Can you do more with yours?

Yes and it's called grumpy

I'm not sure what you mean that they're the same thing. They're different domains with different transfer functions. There's even a recent ish paper exploring how they can be used to inform one another. Can you say more?

In particular PyPy's abstract values have multiple domains at once: type, known bits, known range, etc. They do it all in one pass

You can also compile a scrap to C (and then probably from there to native, or apparently with pnut you can now compile to shell...)

I'm not 100% sure what you're asking. The package manager equivalent is in development but currently looks like a tiny wrapper around a hash->serialized_program database.

Hi! It's me, I'm one of the interpreter and compiler authors. Happy to answer questions

Author here! Happy to answer questions or hear new ideas

This is pretty cool. Thanks for the pointer.

Also, hi Matthias!

This is very cool and seems similar to what we do/did in Cinder: https://bernsteinbear.com/blog/cinder-jit-bisect/

EDIT: Oops, this tool is mentioned in the post already (but the post is not, so here it is if you want to read about it). Neat!

I don't think so... It was part of the download menu for me

I thought copy as MD existed years ago and then went to look for it recently and it was gone. Now it's back. Am I losing it?

For the (slow) interpreter, it's just a string. For the compiler, we may end up just using a string in the end too, and interning, and then doing identity equality on the pointer. This is especially cheap because we have small/in-pointer/immediate strings.

That's correct. No hashes; just string equality on the tag name. Everything gets collapsed to an integer. I'll have to read about why OCaml hashes things. Maybe separate compilation?

Edit: oh, it's unique per compilation and isn't meant to be shared around. The variants are shared with string names in flat scraps (small serialized ASTs)

Right now they're all like OCaml polymorphic variants. There are no static types. That's a future thing we'd like to add to the interpreter/compiler

Your example is legal (modulo the constraint that every tag must have an associated value) and works because we store the tag either in the pointer or the tag field.