For the record, version 1.0 is _supposed_ to support POSIX-style regular expressions with character classes, ranges, and limited backtracking, providing a drop-in API compatible with the Rust [regex](https://github.com/rust-lang/regex) library without the PCRE context-sensitive expressions such as '(x?)' and ''. I maintain that there's something wrong with a regular expression library that can't use _itself_ to parse out regular expressions, which is why "self-hosting" is on the post-1.0 list.

I started the Barre project ten weeks ago. I expected it to take about four to six weeks, and yet here I am, ten weeks in, and I still don't have a full implementation of the core, much less any of the extensions.

I'm disappointed by this slow progress, but not surprised. Half of it has been learning Rust. I mean, I thought I knew Rust when I finished all of the exercises in the Programming Rust textbook, but that was only half the story. I'd only done exercises and built off the work in the book. I hadn't written a program in Rust on my own, "editing a blank page." Rust does not support graph structures comfortably; the options for doing so are limited and verbose, and my choice to use a memory arena means that I had a different set of difficulties with convincing Rust's correctness operators to give me access to nodes as I needed them.

The other half has been that, no really, implementing laziness in a language that doesn't support it natively is really tricky. Tommy McGuire's implementation is blunt and obvious, and while his code was excellent inspiration for what I'm trying to acheive, it's not written in a way that makes it easily portable to Rust. On the other hand, Michael Adams's implementation, while much more complete and tractable in Rust terms, is also written in a highly advanced Racket that re-uses fields, treats types as run-time signals, and does deep exploitation of Lisp's list-like structures to perform post-hoc transformations. Trying to synthesize out of both of those, as well as Matt Might's original implementation, has proven more challenging than I imagined.

As always, there's too much to learn, and not a whole lot of time in which to do it all.