https://github.com/ilai-deutel/kibi
the repo is here btw. i appreciate that the dependency tree is appropriately minimal. also impressive how many comments are included in the source. maybe they’re not counted towards the LOC?
Maybe the comments were written by another LLM. It certainly has that smell.
In case the reference is not clear, the name is likely a callback to antirez’s Kilo[1], an editor in <1000 lines of C.
/tmp/kibi % cloc .
98 text files.
89 unique files.
11 files ignored.
github.com/AlDanial/cloc v 1.96 T=0.05 s (1936.6 files/s, 82405.1 lines/s)
-------------------------------------------------------------------------------
Language files blank comment code
-------------------------------------------------------------------------------
Rust 15 241 379 1480
Markdown 5 179 22 502
INI 50 3 20 388
YAML 10 36 21 281
XML 1 0 1 62
TOML 3 10 1 57
Bourne Shell 1 13 8 46
JSON 2 0 0 31
Dockerfile 1 2 0 3
SVG 1 0 0 1
-------------------------------------------------------------------------------
SUM: 89 484 452 2851
-------------------------------------------------------------------------------
I'm not sure if this counts as less than 1024.Even though that's 400-ish over, the readme states:
1.: Counted per platform, excluding tests and Clippy directives, see count_loc.sh
However, I personally feel like it's a stupid metric anyway because that doesn't count dependencies (although kibi specifically seems to have minimal dependencies), and at some point you could simply abstract your code away into a library and use it as a dependency to prune some lines from the codebase.It also excludes tests, which is how they reach such a low number. It seems they are trying to count the number of lines actually used in the executable. I can't see the value in this metric. If they want to talk about executable size, they should just use that. If they want to talk about the complexity of maintaining their project, then surely you must count all lines because those still make the project harder to maintain. I guess they're trying to measure the logical complexity of the final result, but I can't for the life of me think why a user or developer would care about this.
There are some OS specific files (unix.rx, windows.rs, etc) that you can discount (imo).
If you really wanted to codegolf the repo, I'm sure you can make it literally <1024 lines.
[flagged]
hehe, for some reason "written in rust" is becoming a trademark