feat: poseidon hash #83

lonerapier · 2024-06-01T11:38:56Z

closes #75

It changes the following:

add Poseidon hash function
add sage script modified from original constant generator script for testing purposes

Autoparallel

This is looking super good! I left comments, but you don't have to take them or work on them, they were just suggestions.

On a final note, I did get the SHA-256 merged, so you may want to rebase into that. I'm fine with calling the module hash or hashes, it makes no difference to me. I really like your README that was specific to Poseidon, I probably should have done that for SHA-256.

Autoparallel · 2024-06-05T15:57:42Z

src/field/prime/mod.rs

@@ -211,6 +213,15 @@ impl From<PlutoPrime> for usize {
  }
 }

+impl<const P: usize> FromStr for PrimeField<P> {


Autoparallel · 2024-06-05T16:07:21Z

src/hash/poseidon/mod.rs

+#[derive(Clone, Copy, PartialEq)]
+pub enum SpongeState {
+  /// Sponge is initialised with state marked as zero
+  Init,
+  /// Sponge is absorbing elements
+  Absorbing,
+  /// Sponge is squeezing elements.
+  Squeezing,
+}
+
+/// Sponge config containing sponge rate, and state
+/// * `rate` describes number of elements absorbed by sponge before any permutation round
+/// * `sponge_state` describes current [`SpongeState`]. Initialised in [`SpongeState::Init`] state
+///   and, if sponge is squeezing, then it can't absorb new elements.
+/// * current new elements absorbed into hash state based on the sponge `rate`.
+pub struct SpongeConfig {
+  rate:          usize,
+  capacity:      usize,
+  sponge_state:  SpongeState,
+  absorb_index:  usize,
+  squeeze_index: usize,
+}


One thing I would suggest here based on some checks you do later is use the typestate pattern

The reason why is that you can enforce that a sponge is actually a completely different type when it is in a different state. For instance, remove the enum and consider:

pub trait SpongeState {} pub struct Init; pub struct Absorbing; pub struct Squeezing; impl SpongeState for Init {} impl SpongeState for Absorbing {} impl SpongeState for Squeezing {}

Now, for instance, you define

pub struct PoseidonSponge<F: FiniteField, S: SpongeState> {...}

you definitely don't have to do this, but it would mean that you enforce that, at compile time, a sponge construction is being used properly as only specific states via a compile-time type can do specific actions.

impl<F: FiniteField> Posei

Likewise, you can return a type that is now in a specific state too.

TIL

this is great advice, will check

I like this pattern a lot. I find it really useful for most things that have some kind of "state" to them (hence typestate as a name lol). Anything that moves an error up to compile time is always preferred imo

Autoparallel · 2024-06-05T16:08:58Z

src/hash/poseidon/tests/constants.rs

+pub fn constants() -> (Vec<usize>, Vec<Vec<usize>>) {
+  let rc16: Vec<usize> = vec![
+    10, 69, 45, 20, 69, 36, 74, 82, 77, 24, 23, 26, 19, 54, 69, 88, 27, 58, 58, 33, 13, 58, 56, 2,
+    74, 72, 33, 83, 2, 49, 29, 8, 41, 30, 47, 82, 83, 30, 73, 94, 68, 81, 16, 51, 33, 93, 42, 96,
+    3, 67, 14, 84, 13, 73, 35, 57, 40, 56, 66, 2, 60, 87, 63, 64, 42, 10, 43, 60, 65, 34, 74, 74,
+    99, 3, 77, 71, 70, 96, 10, 9, 18, 14, 24, 93, 97, 54, 16, 70, 81, 35, 35, 75, 100, 81, 88, 26,
+    22, 29, 90, 12, 50, 65, 82, 89, 48, 23, 22, 8, 42, 62, 5, 52, 33, 17, 13, 40, 37, 2, 27, 85,
+    25, 58, 29, 77, 53, 37, 76, 54, 75, 1, 79, 86, 13, 64, 71, 100, 74, 48, 50, 94, 19, 29, 91, 33,
+    9, 0, 16, 25, 19, 76, 38, 50, 43, 52, 85, 84, 4, 7, 90, 0, 24, 42, 37, 53, 41, 71, 34, 7, 13,
+    65, 3, 79, 46, 38, 66, 12, 14, 81, 79, 31, 51, 83, 33, 24, 75, 47, 77, 9, 65, 37, 30, 90, 60,
+    78, 56, 98, 92, 47, 67, 5, 74, 71, 73, 14, 16, 5, 84, 32, 82, 41, 11, 5, 99, 22, 86, 70, 46,
+    36, 78, 14, 66, 19, 45, 77, 65, 97, 35, 46, 10, 28, 6, 10, 64, 95, 35, 97, 66, 65, 84, 52, 30,
+    40, 69, 32, 5, 29, 70, 77, 24, 18, 97, 75, 24, 53, 45, 94, 17, 78, 9, 19, 67, 53, 66, 69, 93,
+    61, 42, 78, 53, 36, 93, 99, 67, 5, 22, 43, 13, 65, 97, 11, 70, 61, 65, 84, 54, 25, 55, 71, 32,
+    70, 83, 9, 61, 36, 1, 69, 40, 41, 47, 82, 33, 48, 71, 84,
+  ];
+
+  let mds16: Vec<Vec<usize>> = vec![
+    vec![72, 34, 99, 22, 24, 11, 70, 84, 77, 1, 59, 58, 26, 81, 80, 55],
+    vec![86, 95, 64, 34, 32, 35, 72, 66, 1, 85, 63, 3, 27, 74, 76, 13],
+    vec![60, 89, 42, 44, 73, 82, 68, 49, 32, 48, 2, 63, 86, 62, 93, 83],
+    vec![90, 14, 46, 82, 69, 29, 66, 92, 28, 93, 86, 27, 25, 98, 74, 2],
+    vec![75, 4, 23, 20, 68, 73, 74, 98, 72, 86, 82, 35, 29, 79, 25, 44],
+    vec![80, 3, 5, 58, 91, 21, 62, 97, 68, 60, 47, 82, 75, 87, 90, 96],
+    vec![43, 39, 34, 21, 49, 100, 98, 40, 66, 90, 75, 29, 42, 12, 79, 47],
+    vec![58, 30, 91, 95, 48, 17, 86, 90, 85, 44, 18, 65, 20, 75, 82, 99],
+    vec![4, 74, 60, 81, 44, 83, 20, 21, 34, 95, 84, 87, 6, 31, 17, 94],
+    vec![70, 22, 13, 47, 100, 75, 49, 65, 69, 77, 60, 86, 90, 97, 62, 45],
+    vec![67, 29, 58, 15, 95, 99, 6, 50, 97, 81, 19, 54, 57, 45, 83, 72],
+    vec![27, 97, 94, 9, 1, 52, 26, 19, 47, 22, 3, 4, 39, 15, 11, 46],
+    vec![74, 20, 89, 27, 94, 8, 81, 36, 70, 72, 76, 11, 15, 67, 19, 85],
+    vec![38, 73, 14, 8, 84, 53, 83, 45, 87, 19, 15, 41, 99, 96, 57, 76],
+    vec![9, 90, 1, 66, 88, 67, 14, 11, 18, 61, 30, 81, 36, 39, 4, 69],
+    vec![95, 72, 48, 70, 13, 87, 34, 82, 46, 56, 51, 62, 97, 20, 65, 24],
+  ];
+  (rc16, mds16)
+}


Do you want to add any code to generate these constants? That would be pretty sweet to do in a later PR!

yup, i've mentioned this in the readme as next steps. will tackle in a future PR for sure.

ronkathon/src/hash/poseidon/README.md

Line 121 in eb6a1de

- [ ] Grain LFSR for round constants generation

I had modified original sage script to generate this rust file for testing purposes for now

Nice! Sorry i missed that in the README. Was too focused on the beautiful implementation :)

0xJepsen

Over all i think this is really good. I think Colins feedback on the typestate pattern is good too.

lonerapier · 2024-06-06T04:59:28Z

@Autoparallel done with the changes, it looks really cool now, and most of the previous state transition errors are now detected at compile time which is so cool :)

thanks for this suggestion.

lonerapier added 6 commits June 1, 2024 17:04

add script to create constants

f640df7

add poseidon hash

87af53c

add sponge construction

624a29c

add tests

31c1e1d

complete docs

da5c48b

Merge branch 'main' into feat/poseidon

54600a5

lonerapier marked this pull request as ready for review June 5, 2024 12:40

lonerapier requested review from Autoparallel, devloper and 0xJepsen June 5, 2024 12:40

silence clippy

eb6a1de

Autoparallel approved these changes Jun 5, 2024

View reviewed changes

Merge branch 'main' into feat/poseidon

576b686

0xJepsen approved these changes Jun 5, 2024

View reviewed changes

lonerapier added 3 commits June 6, 2024 10:16

sponge: use typestate pattern

af96b73

Merge branch 'main' into feat/poseidon

34f507e

add docs for generating constants

31ddf46

0xJepsen merged commit 192c7db into pluto:main Jun 6, 2024
4 checks passed

lonerapier deleted the feat/poseidon branch June 6, 2024 15:14

github-actions bot mentioned this pull request Jul 1, 2024

chore: release v0.1.0 #111

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: poseidon hash #83

feat: poseidon hash #83

lonerapier commented Jun 1, 2024

Autoparallel left a comment

Autoparallel Jun 5, 2024

Autoparallel Jun 5, 2024

Autoparallel Jun 5, 2024

lonerapier Jun 5, 2024

Autoparallel Jun 5, 2024

Autoparallel Jun 5, 2024

lonerapier Jun 5, 2024

lonerapier Jun 5, 2024

Autoparallel Jun 5, 2024

0xJepsen left a comment

lonerapier commented Jun 6, 2024

feat: poseidon hash #83

feat: poseidon hash #83

Conversation

lonerapier commented Jun 1, 2024

Autoparallel left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

0xJepsen left a comment

Choose a reason for hiding this comment

lonerapier commented Jun 6, 2024