Seriously! According to the time command in fish, cldr (my program) takes around 4-5 ms to print files, tldr takes 80-100ms. (PS, on my server, cldr actually takes less than 1 ms to fetch the file, while tldr takes closer to 150 ms. That’s 150 times faster!!!)
This started a couple days ago when a friend of mine alerted me to the fact that the official tldr repo now takes ai contributions for their pages (aka the cheat sheet files that are printed) and for the actual program code. She then showed me a repo called ethical-tldr which does not allow for the use of ai contributions in the pages or program code.
However, their even though their client doesn’t allow vibe coding, it is still written in Python, which itself takes ai contributions.
So, I decided to learn C from scratch and make my own tldr client that’s faster and better than theirs in *every way! Introducing cldr! I am especially proud of the list and search functionality because the interface is just gorgeous.
Note, ABSOLUTELY NO AI was used in the making of this, I learned and wrote everything on my own in just the past 3 days, I don’t think I even copied much or any code from websites.
*There’s still like two things I haven’t implemented that the official client does, but they’re very minor and I’d be willing to bet that 99% of you won’t know the difference.
This looks like a colossally bad idea:
if (!strcmp(platform, "windows") && !isSpecificPlatform) { debugLog("Falling back to rm -rf", 2); debugLog(strfmt("rm -rf exited with code %d",system(strfmt("rm -rf %s", path))), 3); }In addition,
However, their even though their client doesn’t allow vibe coding, it is still written in Python, which itself takes ai contributions.
Per the Makefile, this project is built using GCC. GCC itself accepts AI contributions, provided that the code is “legally insignificant” (meaning less than around 15 lines of code) or is part of test cases: https://gcc.gnu.org/ai-policy.html
For the first part: You’re probably right! I’m not sure of a safer way to do it though. This specific part is an artifact of the original method I used, which I don’t think would have worked on windows (according to some stack overflow post), but the current code might actually work fine, I just don’t have access to anything running windows and would rather not set up a VM.
Second: That’s highly disappointing! I think their policy is fairly well thought out, in a way that allows for accountability and auditability, so I might be okay with continuing to use it, I’ll have to think about it. If you know of an alternative I could use, I will happily take suggestions!
To add to the last comment, I’d also suggest not mixing logging and business code. That is to say that you should put the
systemcall on its own line, so that it is clear that it’s not just a logging call:if (!strcmp(platform, "windows") && !isSpecificPlatform) { debugLog("Falling back to rm -rf", 2); int result = system(strfmt("rm -rf %s", path)); debugLog(strfmt("rm -rf exited with code %d", result), 3); }It may just be a me problem, but I have a tendency to skim over the logging lines unless I’m intentionally reading code carefully, in which case the system call may get overlooked
You can call
rmsafer manner, but you’ll need to fork and use one of theexeclfunctions directly. That way you don’t have to worry about escaping the path. The old implement would potentially delete unrelated data if the path contained whitespace, and could even execute arbitrary commands, since it essentially runsbash -c "rm -rf $path"Those are really good points! I’ve been sanitizing other inputs for shell breakage, but never considered checking this. Either way, I have since removed the rm -rf part of the code, instead opting to only use my “good” method of traversing the directory
Update to my comment: I changed it up a little bit to remove that bit of code. You can optionally uncomment part of the code to re-enable it, in case the current implementation doesn’t work on windows
What your project desperately needs is a description of what it does. I had to follow the links to ethical-tldr back to the original tldr to see what it’s supposed to do.
Ohh! That’s a good point! Thanks! I just updated the README to provide an actual description.
The irony of a tldr missing from tldr…
And yet, you haven’t said it either
It’s basically short-form manpages, so it gives you a one-liner to describe what a CLI utility does, and then just lists a few concrete invocations to show off different use-cases for it.
One other big difference to manpages, though, is that the actual texts aren’t packaged by your distro, but rather downloaded from a server.
I’m not sure why you did this when Tealdeer exists and has for years? https://github.com/tealdeer-rs/tealdeer
My reasoning is simple: I had never heard of it.
But also, this is an incredibly simple program and a really good way to help me learn C. Additionally, it seems like tealdeer accepts ai generated submissions, though generally discourages it. I aim to make my project as free of ai as I possibly can.
https://github.com/tealdeer-rs/tealdeer Rust TLDR
This is an instance of that recurring trope of coding something better out of pure spite
So, I decided to learn C from scratch
…
on my own in just the past 3 daysReally impressive, great job! What resources did you use for learning?
Mostly just w3. Occasionally I had to look up some stuff for more “niche” use cases, but for the most part I just learned from w3.
I already have experience with programming in general, so the hardest part was learning the syntax, and strings… oh god strings are so bad in C, I had to make a new function called strfmt that works like printf but just for returning a formatted string.
Do you know about the sprintf function?
yes! However, I found it very frustrating to use and it kept leading to ludicrous errors beyond my wildest dreams. So I made my own function that works better and easier.
Glad you are enjoying C. As one of my proffesor’s espoused, “It is God’s language”. As a C++ developer, I cannot recommend {fmt} enough.
Are you allergic to vowels?
Those are method names in C, so I’m afraid you’ll have to take it up with a guy named Dennis who isn’t very talkative these days.
Not sure what you mean. You said that you had to make a new function and decided to call it “strfrm”. The standard is actually the opposite afaik. You should be more verbose and not use contractions. Something like “string_format” is far easier to read.
Ah I didn’t see OP’s
strfmtyou’re right I’m wrongAnd I didn’t notice that you weren’t OP. I guess we’re even? Haha
I was just following the naming convention as other string methods, like strlen, strcmp, strcat, (string length, string compare, and string concatenate, respectively), mine is string format, so I did strfmt!
So… C code is faster than python. We knew that. Thanks for the better performing rewrite. Now do it in rust 🤓
Looks like someone did that https://github.com/tldr-pages/tlrc
There’s also https://github.com/tealdeer-rs/tealdeer
In all distros I’ve recently used this is what gets installed when you install the “tldr” package with the package manager. And for good reason.
What is the reason?
I’ll admit I have no idea what makes one of the Rust implementations preferable over the other. Nor can I go looking at the code myself now that GitHub has hosed itself again
Gotta love projects with a properly commented config file that makes it easy to modify its defaults.
Already exists and isn’t vibecoded https://github.com/tealdeer-rs/tealdeer
I was gonna post this as well, this comes with CachyOS by default.
Why in rust? I mean it would be less faster, right? If speed is of the essence ofc.
Because “rewrite it in rust” is its own meme and needs no particular reasoning.
Performance of compiled Rust and compiled C are comparable. Of course, compiling Rust is slower though.
Yeah, but it’s slower because it has compile-time safeguards and guardrails that eliminate some families of really dangerous memory and IPC errors.
Actually not. Rust can be even faster than C, although it’s generally similar in performance to C. The point is that it’s more secure because of the way it handles memory.
The performance metrics are cool (though unscientific) but speed is definitely NOT of the essence lmfao
TLDR is not a high throughput real time application. You run it like once per use, probably at a rate of less than once per day. Then you spend a minute reading the output. We’re talking about minutes-per-year time savings here.
Consider using a different name for the project if you care for searchability. There’s already Unicode CLDR Project.
damn! I guess I should have looked it up before going with the name. I just liked the play on tldr, cldr.
Just go with tlcr!
Ohh, not bad! If I feel like I need to change it, tlcr is probably the best option
You might want to have a glance at the OpenSSF compiler hardening guide, and maybe try out ASAN (
-fsanitize=address).E.g. I get
[nobody@bb14e99a9c8e cldr]$ ./build/cldr --update --language C. caution: filename not matched: ethical-tldr/pages.C./*/*.md ================================================================= ==454==ERROR: LeakSanitizer: detected memory leaks Direct leak of 4 byte(s) in 1 object(s) allocated from: #0 0x7fa48a32c0c1 in malloc (/usr/lib/libasan.so.8+0x12c0c1) (BuildId: b8a4241051a1621937fdc46e867ba7ecb56d96ea) #1 0x7fa489e8be0f (/usr/lib/libc.so.6+0x8be0f) (BuildId: 503200d7fda94a5dc6058d7e0694e5d1dcb2e372) #2 0x7fa48a2f2761 in vasprintf (/usr/lib/libasan.so.8+0xf2761) (BuildId: b8a4241051a1621937fdc46e867ba7ecb56d96ea) #3 0x561d84213d8c in strfmt (/home/nobody/cldr/build/cldr+0x6d8c) (BuildId: 2a0f13efc2c9b12a67e7dd810a2cbdfd1866e53d) SUMMARY: AddressSanitizer: 4 byte(s) leaked in 1 allocation(s).Update to this, ~~I actually just fixed that bug, and ~~I went through the first section of the hardening guide. Changes are now pushed to main and I released v1.2.1
Ah damn! That’s one of the things I had told myself I needed to make sure I checked, but I forgot! Thanks! I’ll look into this tomorrow. Like I said, I literally learned C over the last 3 days to make this, so I’m always happy to learn more!
I wish people still had shame.
Python isn’t for command line utilities that you intend to distribute to others. It’s for personal one-off scripts.
I mean, FFS, if you’re gonna vibecode something, why not pick a fast language? The fact that they can’t even tell the LLM to use the right technology shows that they are a complete imbecile
Python isn’t for command line utilities that you intend to distribute to others
wat
Python isn’t for command line utilities that you intend to distribute to others.
Sorry, but you’re wrong about this. Python can be an excellent choice for many command line utilities intended for distribution to others.
I don’t enjoy making my computer work 20 times harder than it has to
To be honest for the intended purpose of tldr (a script that you run every once in a while), python is quite alright. Python is way more accessible than C.
But cli tools that are intended to be built upon by others with UIs and larger programs, they absolutely need to be built on a low level language so they can be faster and not drag down downstream projects’ potential.
Dependancies
[Optional] wget can be installed for less reliance on other projects using ai, if wget is not found, curl will be used instead
- Small typo: dependencies.
- Is curl using AI in its code now?
Thanks for your work!
Update to my previous comment:
1: Fixed!
2: Yes, it seems like it is! According to the open-slopware repository curl does take ai submissions.
Dang, looks like they went all-in on it.
1: thanks! I’ll fix that soon.
2: this is according to the same friend I talked about, but it would be good to double check her knowledge
Oh yeah this is bound to be maintained
I intend on maintaining this for as long as I’m able to program
excellent. however if you are making a point about AI not being as good s a quality developer, you are missing the point.
CEOs don’t give a shit about quality and effectiveness. it is all about getting rid of labor costs. once all he few remaining providers of service are absorbed, we take what they give and pay through the nose. monthly.
Hell yeah!
