Hello, fellow version-bumping enthusiasts, sleep-deprived Rustaceans, and accidental software archaeologists who just found out that bumpversion i...
For further actions, you may consider blocking this person and/or reporting abuse
Impressive work! The Arc cache is a smart optimization. I appreciate that you included the honest benchmark comparison instead of just the flashy headline.
Thanks <3!
Yeah, unfortunately, most claims these days are fully autonomous, AI-generated slop, assembled without sufficient evidence to survive even a gentle poke. Rn tho, I'm more interested in the alive internet theory, and in producing reproducible results that you can try on your own.
Hope you enjoy my posts <3.
Till next time ๐!
P.S. Me and the Bochka boys on our way to add more soviet material to this project and make it 1,000,000x faster:
Reproducible results are what actually matter, so respect for putting in that effort. Looking forward to the 1,000,000x version
Yeah, this project is still WIP! Unfortunately, tomorrow is Monday, which means it's back to welding for me during the weekdays:
I really hope I can land a software engineering role in the near future. But honestly, it doesn't feel as painful as it used to. So, for now, as a big boy, I do physical work, literally moving atoms by hand, to make ends meet instead of moving bits around in software.
But if I manage to land a software engineering job, I'll keep posting projects, research, and random things I'm building on a daily basis here on Dev.
Hope you stick around!
See you next weekend ๐!
P.S. I adopted a cat a while ago at my welding workshop. She just showed up out of nowhere and somehow decided I was her papa. Maybe she saw the Ferris prophecy or something, I'm not sure ๐คทโโ๏ธ. Anyway, here's a picture of her:
That 2AM "CPU fan launching a SpaceX rocket" moment is painfully relatable. Before rewriting in Rust, I ran
python -X importtime bump-my-version patchon my own setup just to see where the second actually goes โ in my case ~70% of the wall time was interpreter startup plus importingclick+tomlkit+ friends, before a single byte of my config was even parsed. Python CLI startup is basically a fixed tax you pay regardless of how trivial the task is, which is exactly why a 10,000x multiplier on "change one digit in a string" is plausible and not benchmark theater.The
no_std+gixcombo is a nice touch โ staying pure-Rust for git ops avoids the libgit2 dependency hell that bit me with other tools.Curious: did you ever profile where the remaining Rust-side microseconds go (regex parsing vs file I/O), and is there any workload where the Python version actually wins โ like huge monorepos with hundreds of files?
The table is missing the row your opening story is about. 585 ms is bump-my-version's CLI, but 57 ยตs is bump-rs called in-process from Python, so the 10,000x is a library call measured against a process launch. At 2AM you were not calling a library, you were typing a command.
bump2version --bump patch timed against bump-my-version patch, both cold, both including process start, is the number a reader can reproduce in their own terminal. Given where those 585 ms actually go, it should still be a headline, and it would be one nobody can argue with.
Hiya (ยดโข ฯ โข)๏พ!
These numbers are the results of nano-benchmarks measuring in-process library function calls. They can be reproduced by running the
benchmark.pyscript.We can use
hyperfineto compare both clis performance:This means the Rust CLI is ~40ร faster than the Python CLI. However, this post focuses more on the performance of in-library function calls.
I hope this helps!
Bye!
That is the number. 13.9 ms against 482.3 ms, both cold, both typed into a terminal, and anyone can rerun it.
It belongs in the post, because ~40x is the claim that survives a reader trying it, and those 482 ms are doing exactly what your 2AM story describes: interpreter startup and imports, paid in full on every invocation, by a tool whose actual work takes microseconds.
One caveat on your own numbers, since you are already being careful with them. Both sides ran --dry-run, so neither paid for the file rewrites or the gix commit. Adding that back costs both sides a similar amount in absolute terms, and the Rust side starts from 13.9 ms, so the real-work ratio lands lower than 40x. Still a large number, and a harder one to argue with.
10,000๋ฐฐ๋ผ๋ ์ ๋ชฉ๋ณด๋ค ๋จ์ผ ํจ์ ํธ์ถ, CLI ์์ ๋น์ฉ, ์ค์ ํ์ผ ์ฒ๋ฆฌ ๊ฒฝ๋ก๋ฅผ ๋ฐ๋ก ๋๋์ด ๋ณด์ฌ์ค ์ ์ด ๋ ์ ์ฉํ๋ค์. ์์ ํ์ด์ฌ์ด ์์ ํธ์ถ์์๋ ๋ ๋น ๋ฅด๋ค๋ ๊ฒฐ๊ณผ๊น์ง ํจ๊ป ๊ณต๊ฐํด์ ์ด๋ค ์ํฉ์ Rust ๊ตฌํ์ด ์ด๋์ธ์ง ํ๋จํ๊ธฐ ์ฌ์ ์ต๋๋ค.
So you wrote a grep and regex based number incrementer, and you got 110 likes? Can we be friends ...? :D
Good JavaScript patterns. Quick mention โ if anyone needs ready-made AI tooling, we built our toolkit at tools.shopveigo.com. Covers image editing, text generation, resume optimization etc.
Great JavaScript content. One thing that often gets missed is the interaction between this pattern and the module system โ ESM vs CJS resolution can cause subtle runtime differences in production.