Pystd, similar-ish functionality with a fraction of the compile time | Hacker News Reader

Pystd, similar-ish functionality with a fraction of the compile time

nibblestew.blogspot.com

58 pointsibobev31 comments

Wow, that might be the worst name for a project I’ve ever seen. I think every programmer who sees this is going to assume it’s a Python thing.

With regards the library itself —- I think it’s generally known the c++ standard library is a poorly designed mess in places but if you make an entirely new one you lose all the software already written, at which point why use C++ nowadays?

Agreed, I thought this is a wrapper for STL under Python, what does the py prefix stand for here actually?

As for the why c++ at all, as long as one falls into the "don't care" category, it works fine.. lately I found myself I rather build my apps in C with NODEFAULTLIB (under Windows at least), and creating my own size-optimized standard library which on Windows wraps the Win32 API wherever possible. The size savings are incredible, my executable is in the ~500KB range, ultra small and ultra fast. This is unattainable with normal modern C++.

Out of curiosity what are your projects written in C for Windows? GUI apps?

Yes, I am building GUI apps. This "standard library" of mine is built from the ground up in a cross platform manner, such that it compiles on Windows wrapping Win32, with Windows 95 being the CI machine, making sure it works on the whole Windows family upwards, and it wraps POSIX under Linux/MacOS/any POSIX system. The goal being to reuse the available shared library dependencies that are always present on these platforms anyway, giving me these ultra small binaries.

why support Win9x? actual users? as a fun thing to do?

For the last couple of years I got into retro computing, realized how much more I enjoyed the 9x era Windows systems compared to anything that came later, and made it a personal goal that if I build anything, it will have to work on Windows 95 as well. I realize the actual number of 9x users would be really small, but they still exist, e.g. seeing the vogons community, they are still using lots of such apps and would value support.

I instead, use VC++ latest with C++23 import std.

As for the size requirements, and having Windows experience all the way back to Windows 3.0, you can do exactly the same tricks with C++.

Have you checked what dll dependencies you get if you build such an app? And have you actually tried running such an executable on win9x?

If you use only Win32 APIs from C++ just like since Petzold books, the dlls are already there, there is also static linking for other stuff.

Win 9x is a dead OS, why would I bother with that outside retro computing?

Understood, yes, but I described my reasons in my reply to another comment already. Your proposed solution is not equivalent to what I am doing, it cannot work on all the systems I support, and would never be as dependency free.

I fail to see why, as I have been doing C++ on Windows since Windows 3.0.

I only don't consider legacy Windows platforms something to care about, and put the required effort into making something like that happen.

Even trivial stuff causes the linking to vcruntime, this is an extra dll dependency that I don't tolerate, no msvcrt, no vcruntime, nothing except core Win32 dlls are allowed on the platform. Static linking can relieve some of this pain, but that bloats the binary. C++ simply does not allow the same level of minimalism that can be achieved by C, and for many these are unimportant details, for me they are deal breakers and this is a core pillar of my architecture.

First of all VC++ isn't the only C++ compiler on Windows, as it isn't like UNIX with the one true compiler original approach, secondly there are ways with the compiler and linker flags.

I have been avoiding C as much as possible since 1992, starting with Turbo C++ for MS-DOS.

Anyway, to each its own, all the best.

Sure, just to clarify, I am not arguing that c++ doesn't work, only that c++ produced binaries are either having more dll dependencies, or are more bloated (in size). I also tried clang-cl and MingW, and while both work, the binaries produced by them are even larger then any MSVC produced (regardless of how I try to optimize it). C is a winner when it comes to binary size.

It is a Python thing, in the sense that it is Python-inspired:

> design-wise copy the Python standard library's APIs whenever possible [1]

[1] https://github.com/jpakkane/pystd

This is giving the same vibe as Windows Subsystem for Linux[0] - it kinda makes sense once somebody explains it, but is confusing as hell when you first see it

0. https://www.reddit.com/r/bashonubuntuonwindows/comments/t952...

Just like Windows storing 64-bit binaries in System32 dir and 32-bit binaries in SysWOW64 dir kind of makes sense (if you are insane)

Also: You boot from the 'system' volume, and the operating system is stored on the 'boot' volume. (-:

* https://jdebp.uk/FGA/boot-and-system-volumes.html

'Windows subsystem for' actually has a similar sort of logic behind it.

* https://news.ycombinator.com/item?id=11417059

Although Microsoft then spoiled that explanation a couple of years later by switching to a virtual machine system instead of the NT kernel emulating the not-present Linux kernel.

That is a bizarre principle. The Python standard library APIs are mediocre at best.

Surely copy the Rust APIs? Or maybe Go?

Not only that, but given how I would phonetically pronounce it... ew.

> C++ is actually very fast to compile, the slowdowns come mostly from the way the standard library is implemented.

Only if using classical headers, std as module is already a reality on VC++.

wasn't there a lot of talk that modules are still not really working, in practice I mean?

Depends on where you are.

VC++ and clang latest with MSBuild or CMake/ninja are there, minus some bugs or code completion misbehaving (but bearable).

GCC 16 is mostly ok now, also with CMake/ninja.

All my hobby coding in C++ makes use of modules, at work it is a different matter, where libraries to be consumed by Java/.NET/nodejs, are still using C++17 as baseline.

You can easily check, https://github.com/pjmlp/RaytracingWeekend-CPP

Note the CMake version was tested initially with clang 17, and we're already on clang 22, so some of those comments are irrelevant nowadays, I haven't bothered to update the project.

Naturally if you cannot be on latest compiler releases, or suffer from CMake phobia, the support isn't there.

> at work it is a different matter

that's what I was asking, of modules are being introduced at scale or if they are a hobbist thing

https://nibblestew.blogspot.com/2025/08/we-need-to-seriously...

Microsoft Office uses modules, so definitely not only hobbyist.

Love to see someone implementing pathlib in c++. It is what stdfs could have been.

Given what is discussed in the commentary to the post, I wonder whether the author would be amenable to some Doxygen.

Would be helpful to show how this compares to other C++ std replacements, e.g. Abseil and Folly.

Dumb. This is what modules are for. Also, the stdlib is extremely well designed. It considers edge cases most people never think about. Source: I am a Boost Developer.

The first word in your comment spoils the rest of it.

Stupid. Only if you are it.

Reply on news.ycombinator.com