Merge dev branch into main

[Dherse]

Near complete rewrite of codly:

• performance improvements: a mix of algorithmic improvements, smarter memoization boundary data propagation
• memory improvement: mostly through less boundary propagation and less allocations of temporary arrays
• now using elembic for show/set rule behaviour
• Implemented many feature request (final list TBD) and many requested fixed (final list TBD)

Disclosure: AI was used in the making of this code, mostly around building/improving the tests & to help me do roadmap of issue fixing/tracking, there were so many at this point I was struggling to keep up. Almost none of the actual library code is AI generated.

[MBradbury]

Would it be possible to add the number-format option back in, I think it would be sufficient to do the following for the codly-number arguments:

Add a field for format:

e.field("format", e.types.option(function), doc: __doc("number-format"), default: __default("number-format")),

Call format on the body:

display: it => (it.format)(it.body),

In 1.3.0, I have used this to style the line numbers and also add a link to the line number (e.g., to a website that explains a bash command).

For example, I would do:

#import "@preview/percencode:0.1.0": percent-encode
#let number-format(number) = {
    let rawblock = query(selector(raw).before(here())).last()
    let number-text = text(size: 10pt, fill: dark-grey)[#number]
    if rawblock.lang == "bash" {
        let line-text = rawblock.lines.at(number - 1).text
        link("https://explainshell.com/explain?cmd=" + percent-encode(line-text), number-text)
    } else {
        number-text
    }
}
#show: codly.number-set_(
    format: number-format,
)

[Dherse]

@MBradbury You're right, but I'll do it a bit differently, I'll give codly-number more capabilities, namely taking in a numbering (function or auto or str) and having a line arg. I think this is just an oversight of rewriting the code. I will fix it tonight! Thanks a ton of taking the time to try it out and find the issues!!!

EDIT: this is in fact how I originally intended for it to work, but as I was prototyping I probably forgot 🙈

[MBradbury]

Thanks for taking a look! The accessibility fixes will be nice, as Codly was my blocker for PDF/UA-1.

I think the only other regression I spotted was that language tags now render on top of any code, which is a problem if there is a very long first line of code. I believe in 1.3.0 the language tags would "push" the long line of code onto another line. I have not tried your recent changes, so not sure if it got fixed as part of them.

For example, in the following the -o is covered by the Bash language tag, whereas I think the -o would have been pushed onto the line with rshell-syscall.o in 1.3.0.

\```bash
gcc -fPIC -fcf-protection=none -nolibc -nostdlib -c rshell-syscall.c -o 
rshell-syscall.o
\```

[Dherse]

Ah yes, the old code was "smart" in that regard, I'll fix that too. Thanks! Don't hesitate if you find anythign else.

[MBradbury]

Just tested the latest commit and I was able to adapt my code in #141 (comment) to work with the changes, thanks very much!