Ed Minnix retweetledi
Devs who love coding don’t wait for the AI limits reset
English
Ed Minnix
718 posts
Ed Minnix
@egregius313
Software engineer focused on performance and security
New Jersey, USA Katılım Ocak 2018
411 Takip Edilen73 Takipçiler
Ed Minnix retweetledi
we have all become little googles
Armin Ronacher ⇌@mitsuhiko
I have such a large graveyard of vibe coded apps and productivity things at this point.
English
It really highlights part of how Emacs making certain things trivial leads to different community dynamics.
In a “normal” editor if something breaks you need to either fork the editor or extension. But Emacs and Vim have historically made it easy enough to just patch it yourself without upstreaming
I’m seeing this become increasingly common with “normal” software as GenAI is changing what is easy/difficult to build
@pidotdev is probably the piece of modern software that’s closest to the “if you want to change it you can” that I’m used to with Emacs
English
@egregius313 I should probably search my init.el for every advice-add that I use to fix upstream package bugs, I’m just too lazy to submit patches when I can continue moving after 10 minutes of monkey patching
English
> parsing a URL shouldn’t require a third party dependency
> use this third party dependency to do it
CPPAlliance@CPPAlliance
Parsing a URL in C++ shouldn't require a third party dependency, a regex, or 200 lines of hand rolled string splitting Boost.URL gives you RFC compliant URL parsing. url_view parses with zero allocations. And parsing is now constexpr 🧵👇
English
@leothrix Loosely reminds me of this post. The trade-offs between enabling ease of human use with the difficulty of making secure systems really show up anytime parsing is involved, it’s crazy
Even more reason to prefer structured data imo
x.com/drmaciver/stat…
David R. MacIver@DRMacIver
PSA: If a function accepts a string then it's a parser. Parsers are hard to get right and dangerous to get wrong. Write fewer of them.
English
Ed Minnix retweetledi
@ClassicGamerTWR @carllerche I understand your point about ORMs worsening performance, but why productivity?
English
If you have solid reasons other than performance, then far be it from me to stop you. I don’t generally recommend it, but only you are in a position to know your exact problem space.
Still don’t use an ORM. That will tank your performance and productivity. No amount of Rust will make up for it.
If you’re not convinced, I’m giving a talk on the underlying performance problems of ORMs at Big Sky Con on July 25th where I’ll lay out the problems in details.
English
1. Stop using ORMs. It’s a great way to make your Rust perform like whatever web technology you’re trying to replace.
2. Don’t use Rust for webapps. The performance is coming from not using ORMs (see #1) which is easier to correct in a GC language than rewriting your webapp in a manually allocated language.
Carl Lerche@carllerche
Sure, Toasty is a cool ORM for Rust, but wouldn't it be nice to be able to easily build a web app in Rust? I think so... maybe one day. (I was told I wasn't supposed to post about it yet...)
English
I frequently include this in my skills for certain tasks lol
“## References
- The source for is at
- ….”
Because otherwise I have issues where the SOTA models decide the best way to figure out a new API is scanning my nuget cache and loading arbitrary assemblies into their PowerShell session 😳
English
@paulthomson83 I just got tired of constantly telling it where my other code is :)
English
Pi doesn't have a good built-in answer for working across repos, but nothing an extension can't fix.
English
@HSVSphere @leothrix I miss the days when Mac firmware came with a Forth interpreter and Emacs
English
Ed Minnix retweetledi
Ed Minnix retweetledi
I don't mean to be flippant when I say this but, yes, people like maximizing money whilst minimizing effort. Other economic factors aside, if software engineers were paid like pure mathematicians, trust me, the people in it would only be in it for the love of the game.
LaurieWired@lauriewired
I’m convinced that a large % of programmers don’t actually like computers. As a side effect, are also perfectly happy to throw away their reasoning to a model as soon as they can. I don’t get it, at ALL. Don’t you *LIKE* understanding the magic of the machine? You do realize hand-programming (I hate that I even have to specify hand now) is fun…right?
English
@TaliaRinger I haven’t tried anything with specs per se. But I have had some success with having a model do property based testing. I’ve done some with hypothesis (Python) and I’m hoping to do some with FsCheck (day job uses C# so gonna try out the F# based one for .NET)
English
chat how are we doing on this
Talia Ringer 🕊🪬@TaliaRinger
The year is 2030. You're a software engineer at a company, writing tests for your program.
English
@egregius313 @mycoliza @rfleury The trick is getting the false positive rate on those checks sufficiently low (ask me how I know).
English
my hot take is that Andrew Zig and Jarred Bun are both annoying blowhards and neither of them understands Rust or how to develop correct software, btw
English
Having worked in static analysis in the past, I frequently tell people: “I want to add these checks to make things run more smoothly for everyone while reducing effort. Why spend the time and money having engineers do the busy work when we can make the machines do it for almost free?”
English
@rfleury the funny thing is, i agree with you! i know i’m stupid, which is why i reach for tools and practices that make it harder to make stupid mistakes. :)
English
(also, you know, i don’t want to learn muscle memory that’s tied to a specific commodity token vendor!)
English
That makes sense. I’ve just been burned by too many people who genuinely have thoughts like this lol (Including but not limited to many Lisps’ love of making most data structures cons-cell based 🫠)
I was mostly highlighting that the Python “well everything’s a dictionary under the hood” excuse is no longer relevant. Similar to how JavaScript objects are optimized in v8 using the Butterfly
English
@egregius313 The advice is sarcastic. A function that returns a dictionary tends to be more mysterious than one that returns an instance of a class. It’s not just about performance.
English
And if you are rewriting a codebase from Python to Rust, go ahead and create a BTreeMap for a map that will only contain two entries for the entire lifetime of the program.
the Rich@Duderichy
pro tip for writing python: use as many dictionaries as possible you don't need classes / objects just pass around weird dictionaries what does it contain? who knows, good luck
English
@theodorvaryag My cat says I’m obligated to answer “cats” lol
Also, nature camp as a kid gave me a fondness for birds of prey tbh. I love seeing hawks and vultures flying overhead, and there used to be a camera nearby that let you see the falcons that lived on top of the local bridge
English
I like dogs, bears, bats, owls, cats, and cows
What animals do you like?
English
The most painful part *for me* seeing Andrew's blogpost being discussed is that it makes it clear that to most people, only Rust and Zig exist as alternatives to C/C++.
No mention of Odin or C3.
Clearly there's A LOT of work left to be done in spreading the word.
English