fzakaria

#!/usr/bin/env nix
#! nix shell nixpkgs#python3 --command python3
"""Every nixpkgs revision the flake ecosystem pins, and which channel it was.

Reads `pins.jsonl` from a checkout of https://github.com/fzakaria/omniflake --
one row per indexed flake, carrying the nixpkgs its lock file names -- joins it
against every channel bump the nix release archive ever published, and writes
nixpkgs-pins.csv.

    git clone https://github.com/fzakaria/omniflake

fzakaria

SQLite inside the Nix evaluator, via builtins.wasm

A real SQLite, compiled to WebAssembly, running inside Nix evaluation and answering arbitrary SQL against a database it never opens as a file.

# query.nix
builtins.wasm { path = ./sqlite_nix.wasm; } {
  db  = ./index.db;
  sql = "SELECT version, rev FROM versions

fzakaria

#!/usr/bin/env nix-shell
#! nix-shell -i python3 --pure
#! nix-shell -p "python3.withPackages(ps: with ps; [ plotnine pandas numpy scipy ])"
#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/72841a4a8761d1aed92ef6169a636872c986c76d.tar.gz
"""
Generate the figures for the "look at your data" post.

Self-contained: the shebang above pins nixpkgs to an exact commit and pulls in
plotnine + friends, so `./latency_charts.py` reproduces every figure on any
machine with Nix installed -- no virtualenv, no pip, no system packages.

fzakaria

// SPDX-License-Identifier: GPL-2.0
/*
 * shebang_origin.bpf.c - $ORIGIN-relative shebang (#!) interpreter resolution.
 *
 * A binfmt_misc_ops handler for scripts whose interpreter line is
 * "#!$ORIGIN/...": the interpreter is resolved relative to the directory of
 * the script and selected via bpf_binprm_set_interp(). Ordinary scripts are
 * declined so binfmt_script handles them.
 *
 * The whole shebang line is in the prefetched bprm->buf, so no file read is

fzakaria

{
  description = "Linux Kernel Development Environment";

  inputs = {
    nixpkgs.url = "github:NixOS/nixpkgs/nixos-25.11";
  };

  outputs = {
    self,
    nixpkgs,

fzakaria

// SPDX-License-Identifier: GPL-2.0
/*
 * binfmt_misc 'B' (BPF) program: match aarch64 ELF binaries and choose the
 * interpreter (qemu-aarch64) dynamically via bpf_binprm_set_interp().
 *
 * The program is given the BINPRM_BUF_SIZE file header as context. It inspects
 * the ELF header and, on a match, sets the interpreter and returns 1. This
 * makes the handler a superset of a traditional magic rule: the program both
 * decides the match and computes the interpreter path (it could equally derive
 * a path relative to the binary).

fzakaria

package main

import (
	"encoding/json"
	"fmt"
	"os"
	"os/exec"
	"strings"
)

fzakaria

Cache-Aware Struct Benchmarks

Benchmarks demonstrating how struct size and memory layout affect performance due to CPU cache behavior. Companion code for the blog post Every byte matters.

Benchmarks

1. Pointer-Chasing (cache staircase)

fzakaria

ClassPathInfo = provider(
    "Provider for classpath information",
    fields = {
        "compile_jars": "A json fragment containing the mapping of the current jar to its source jar",
    },
)

def _classpath_aspect_impl(target, ctx):
    java_info = target[JavaInfo]
    compile_jars = depset(

fzakaria

"""
FlatBuffer Java rule for Bazel
"""

FlatbufferInfo = provider(
    doc = "A provider for FlatBuffer schema files.",
    fields = {
        "fbs_files": "The list of fbs files that were used as inputs",
    },
)