frozen-hashbrown

Crates.iofrozen-hashbrown
lib.rsfrozen-hashbrown
version0.1.0
sourcesrc
created_at2023-10-30 13:00:13.751539
updated_at2023-10-30 13:00:13.751539
descriptionFrozen version of Rust standard library's hashbrown
homepage
repositoryhttps://github.com/tyt2y3/frozen-hashbrown
max_upload_size
id1018457
size21,784
Chris Tsang (tyt2y3)

documentation

README

frozen-hashbrown

Frozen version of Rust standard library's hashbrown.

What is this about

  1. Allow you to dump the memory of a std::collections::HashMap into a blob
  2. Load the blob and re-construct the hashmap
  3. Such that we can iterate through the data!

What is this for

  1. It's used in FireDBG to allow us to capture and render HashMap
  2. It could also probably allow us to extract HashMap from coredump

How it works

Online

  1. Construct TableLayout for (K, V)
  2. Extract ctrl and bucket_mask
  3. calculate_layout_for(buckets) and calculate the address NonNull<u8> and Layout
  4. Dump the memory into a blob

Offline

  1. Load the blob into memory
  2. Re-construct hashbrown::map::HashMap for (K, V)
  3. Ready to serve

Why does it work

  1. The HashMap in Rust's standard library is a flat hashmap. Meaning it's only backed by a single contiguous piece of memory.
  2. It's dense for small maps and is very memory efficient
  3. It's more like a glorified Vec<(K, V)> with an index to assist hash key lookup

How to use

use frozen_hashbrown::FrozenHashMap;
use std::collections::HashMap;

let map: HashMap<char, i32> = [('a', 1), ('b', 2), ('c', 3), ('d', 4)]
    .into_iter()
    .collect();
let snapshot = format!("{map:?}");

let frozen = FrozenHashMap::construct(&map);
std::mem::drop(map);
let frozen: Vec<u8> = frozen.store();

let mut unfrozen = FrozenHashMap::load(&frozen).expect("Failed to load");
let unfrozen = unfrozen
    .reconstruct::<char, i32>()
    .expect("Failed to reconstruct");
let unfrozen_snapshot = format!("{unfrozen:?}");

// even the "random" iteration order holds
assert_eq!(snapshot, unfrozen_snapshot);

More examples under https://github.com/tyt2y3/frozen-hashbrown/blob/main/tests/unfreeze.rs

License: MIT OR Apache-2.0

Commit count: 4

cargo fmt