@@ -39,7 +43,7 @@ on your system: it is pretty _[fast](#benchmarks)_...
- Fast
- Really fast
- Minimal dependencies
-- Tiny binary (~370kb [^1])
+- Tiny binary (~370kb)
- Actually really fast
- Cool NixOS logo (other, inferior, distros are not supported)
- Reliable detection of following info:
@@ -56,47 +60,36 @@ on your system: it is pretty _[fast](#benchmarks)_...
- Did I mention fast?
- Respects [`NO_COLOR` spec](https://no-color.org/)
-[^1]: With the Mold linker, which is enabled by default in the Flake package,
- the binary size is roughly 350kb. That's nearly 20kb reduction in size :)
-
## Motivation
-[Rube-Goldmark Machine]: https://en.wikipedia.org/wiki/Rube_Goldberg_machine
+Fastfetch, as its name probably hinted, is a very fast fetch tool written in C.
+However, I am not interested in _any_ of its additional features, and I'm not
+interested in its configuration options. Sure I can _configure_ it when I
+dislike the defaults, but how often would I really change the configuration...
-Fastfetch, as its name _probably_ already hinted, is a very fast fetch tool
-written in C. I used to use Fastfetch on my systems, but I eventually came to
-the realization that I am _not interested in any of its additional features_. I
-don't use Sixel, I don't change my configuration more than maybe once a year and
-I don't even display most of the fields that it does. Sure the configurability
-is nice and I can configure the defaults that I do not like but how often do I
-really do that?
+Microfetch is my response to this problem. It is an _even faster_ fetch tool
+that I would've written in Bash and put in my `~/.bashrc` but is _actually_
+incredibly fast because it opts out of all the customization options provided by
+tools such as Fastfetch. Ultimately, it's a small, opinionated binary with a
+nice size that doesn't bother me, and incredible speed. Customization? No thank
+you. I cannot re-iterate it enough, Microfetch is _annoyingly fast_.
-Since I already enjoy programming challenges, and don't use a fetch program that
-often, I eventually came to try and answer the question _how fast can I make my
-fetch script?_ It is an _even faster_ fetch tool that I would've written in Bash
-and put in my `~/.bashrc` but is _actually_ incredibly fast because it opts out
-of all the customization options provided by tools such as Fastfetch. Since
-Fetch scripts are kind of a coming-of-age ritual for most Linux users, I've
-decided to use it on my system. You also might be interested if you like the
-defaults and like speed.
+The project is written in Rust, which comes at the cost of "bloated" dependency
+trees and the increased build times, but we make an extended effort to keep the
+dependencies minimal and build times managable. The latter is also very easily
+mitigated with Nix's binary cache systems. Since Microfetch is already in
+Nixpkgs, you are recommended to use it to utilize the binary cache properly. The
+usage of Rust _is_ nice, however, since it provides us with incredible tooling
+and a very powerful language that allows for Microfetch to be as fast as
+possible. Sure C could've been used here as well, but do you think I hate
+myself?
-Ultimately, it's a small, opinionated binary with a nice size that doesn't
-bother me, and incredible speed. Customization? No thank you. I cannot
-re-iterate it enough, Microfetch is _annoyingly fast_. It does not, however,
-solve a technical problem. The "problem" Microfetch solves is entirely
-self-imposed. On the matter of _size_, the project is written in Rust, which
-comes at the cost of "bloated" dependency trees and the increased build times,
-but we make an extended effort to keep the dependencies minimal and build times
-managable. The latter is also very easily mitigated with Nix's binary cache
-systems. Since Microfetch is already in Nixpkgs, you are recommended to use it
-to utilize the binary cache properly. The usage of Rust _is_ nice, however,
-since it provides us with incredible tooling and a very powerful language that
-allows for Microfetch to be as fast as possible. ~~Sure C could've been used
-here as well, but do you think I hate myself?~~ Microfetch now features
-handwritten assembly to unsafely optimize some areas. In hindsight you all
-should have seen this coming. Is it faster? Yes.
-
-Also see: [Rube-Goldmark Machine]
+> [!IMPORTANT]
+> **Update as of November 30th, 2025**:
+>
+> Microfetch now inlines handwritten assembly for even better performance. I
+> know I previously said I do not hate myself but I'm beginning to suspect this
+> is no longer the case. Enjoy the performance benefits!
## Benchmarks
@@ -204,31 +197,17 @@ You can't.
### Why?
-Customization, of most kinds, are expensive: I could try reading environment
-variables, parse command-line arguments or read a configuration file to allow
-configuring various fields but those inflate execution time and the resource
-consumption by a lot. Since Microfetch is closer to a code golf challenge than a
-program that attempts to fill a gap, I have elected not to make this trade.
+Customization, of any kind, is expensive: I could try reading environment
+variables, parse command-line arguments or read a configuration file but all of
+those increment execution time and resource consumption by a lot.
### Really?
-[main module]: ./src/main.rs
-[discussions tab]: https://github.com/NotAShelf/microfetch/discussions
-
-To be fair, you _can_ customize Microfetch by, well, patching it. It is
-certainly not the easiest way of doing so but if you are planning to change
-something in Microfetch, patching is the best way to go. It will also the only
-way that does not compromise on speed, unless you patch in bad code. Various
-users have adapted Microfetch to their distribution by patching the
-[main module] and inserting the logo of their choice. This is also the best way
-to go if you plan to make small changes. If your changes are not small, you
-might want to look for a program that is designed to be customizable; Microfetch
-is built for maximum performance.
+To be fair, you _can_ customize Microfetch by, well, patching it. It's not the
+best way per se, but it will be the only way that does not compromise on speed.
The Nix package allows passing patches in a streamlined manner by passing
-`.overrideAttrs` to the derivation. You can apply your patches in `patches` and
-share your derivations with people. Feel free to use the [discussions tab] to
-share your own variants of Microfetch!
+`.overrideAttrs` to the derivation.
## Contributing
@@ -240,22 +219,13 @@ Contributions that help improve performance in specific areas of Microfetch are
welcome. Though, prepare to be bombarded with questions if your changes are
large.
-### Hacking
+## Hacking
-A Nix flake is provided. You may use `nix develop` to get started. Direnv users
-may instead run `direnv allow` to get a complete environment with shell
-integration.
+A Nix flake is provided. `nix develop` to get started. Direnv users may simply
+run `direnv allow` to get started.
-Non-Nix user will need `cargo`, `clang` and `mold` installed on their system to
-build Microfetch. As Mold seems to yield _slightly_ better results than the
-default linker, it has been set as the default in `.cargo/config.toml` for
-x86-64 Linux. You may override those defaults using the `RUSTFLAGS` environment
-variable. For example:
-
-```sh
-# Use ld instead of Mold
-$ RUSTFLAGS="-C linker=/path/to/ld.lld" cargo build
-```
+Non-nix users will need `cargo` and `gcc` installed on their system, see
+`Cargo.toml` for available release profiles.
## Thanks
@@ -272,7 +242,6 @@ person about current issues. To list a few, special thanks to:
- [@sioodmy](https://github.com/sioodmy) - Being cute
- [@mewoocat](https://github.com/mewoocat) - The awesome NixOS logo ASCII used
in Microfetch
-- [@uzaaft](https://github.com/uzaaft) - Helping me going faster
Additionally a big thank you to everyone who used, talked about or criticized
Microfetch. I might have missed your name here, but you have my thanks.
diff --git a/flake.nix b/flake.nix
index 90978a2..ed36872 100644
--- a/flake.nix
+++ b/flake.nix
@@ -10,12 +10,9 @@
forEachSystem = nixpkgs.lib.genAttrs systems;
pkgsForEach = nixpkgs.legacyPackages;
in {
- packages = forEachSystem (system: let
- pkgs = pkgsForEach.${system};
- in {
+ packages = forEachSystem (system: {
default = self.packages.${system}.microfetch;
- microfetch = pkgs.callPackage ./nix/package.nix {};
- microfetch-mold = pkgs.callPackage ./nix/package.nix {useMold = true;};
+ microfetch = pkgsForEach.${system}.callPackage ./nix/package.nix {};
});
devShells = forEachSystem (system: {
diff --git a/nix/package.nix b/nix/package.nix
index 052f576..dca3cc2 100644
--- a/nix/package.nix
+++ b/nix/package.nix
@@ -1,22 +1,14 @@
{
lib,
- stdenv,
- stdenvAdapters,
rustPlatform,
+ stdenvAdapters,
llvm,
- useMold ? stdenv.isLinux && !stdenv.hostPlatform.isAarch,
}: let
toml = (lib.importTOML ../Cargo.toml).package;
pname = toml.name;
inherit (toml) version;
-
- # Select stdenv based on useMold flag
- stdenv =
- if useMold
- then stdenvAdapters.useMoldLinker llvm.stdenv
- else llvm.stdenv;
in
- rustPlatform.buildRustPackage.override {inherit stdenv;} {
+ rustPlatform.buildRustPackage.override {stdenv = stdenvAdapters.useMoldLinker llvm.stdenv;} {
inherit pname version;
src = let
fs = lib.fileset;
@@ -34,14 +26,7 @@ in
cargoLock.lockFile = ../Cargo.lock;
enableParallelBuilding = true;
- buildNoDefaultFeatures = true;
- doCheck = false;
-
- # Only set RUSTFLAGS for mold if useMold is enabled
- env = lib.optionalAttrs useMold {
- CARGO_LINKER = "clang";
- RUSTFLAGS = "-C link-arg=-fuse-ld=mold";
- };
+ env.RUSTFLAGS = "-C link-arg=-fuse-ld=mold";
meta = {
description = "Microscopic fetch script in Rust, for NixOS systems";
diff --git a/nix/shell.nix b/nix/shell.nix
index ae65e10..111b803 100644
--- a/nix/shell.nix
+++ b/nix/shell.nix
@@ -1,31 +1,27 @@
{
mkShell,
- cargo,
- rustc,
- mold,
- clang,
rust-analyzer-unwrapped,
rustfmt,
clippy,
+ cargo,
taplo,
+ rustc,
rustPlatform,
gnuplot,
}:
mkShell {
- name = "microfetch";
strictDeps = true;
+
nativeBuildInputs = [
cargo
rustc
- mold
- clang
rust-analyzer-unwrapped
(rustfmt.override {asNightly = true;})
clippy
taplo
- gnuplot # for Criterion.rs plots
+ gnuplot # For Criterion.rs plots
];
env.RUST_SRC_PATH = "${rustPlatform.rustLibSrc}";
diff --git a/src/colors.rs b/src/colors.rs
index 7c65944..07ab9bd 100644
--- a/src/colors.rs
+++ b/src/colors.rs
@@ -1,4 +1,4 @@
-use std::sync::LazyLock;
+use std::{env, sync::LazyLock};
pub struct Colors {
pub reset: &'static str,
@@ -37,8 +37,8 @@ impl Colors {
}
pub static COLORS: LazyLock