Make BitZ generic over semirings, rings and fields by frozenspider · Pull Request #140 · worldfnd/BitZ

frozenspider · 2026-09-29T15:29:57Z

A follow-up to #135, another step toward #136 (hopefully second-to-last).

Summary of changes:

Added new blanket trait BitzClaimField.
Generalized concrete Fq to F: BitzClaimField.
- Note that F128 is still hardcoded, and that's intended - we don't plan to generalize over it.
Did the same with u128 and BigUint -> BitzSemiring in a few places.
Relaxed ConstField requirement to Field.
Renamed BitzRing to BitzConstraintRing to avoid confusion with BitzClaimField::Integer.
Added a bridge between BitzConstraintRing and BitzClaimField in the form of ProjectConstraint.
In places where generalization was not feasible, removed direct imports of BigUint, BigInt and Fq, replacing them with type aliases (S, R, F).
Spartan SHA-256 test and benchmark moved to tooling crate in order to get access to ProjectBigIntToFq.

A follow-up PR will replace Fq with dynamic field.

…ences

# Conflicts: # crates/tests/examples/dump_bitz.rs # crates/tests/tests/host.rs # crates/tests/tests/prove.rs # crates/verifier/src/verify.rs # tooling/cli/src/end_to_end.rs # tooling/cli/tests/circuits.rs # tooling/cli/tests/end_to_end.rs

xrvdg · 2026-10-06T05:59:48Z

+type S = num_bigint::BigUint;
+type R = num_bigint::BigInt;


I don't mind single letter types, but here I'm missing the mnemonic for S and R.

It stands for (S)emiring and (R)ing

xrvdg · 2026-10-06T06:07:07Z

-        params: &BitZParams<Q>,
-        row_weights: Vec<Fq<Q>>,
-        column_weights: Vec<Fq<Q>>,
-        target: Fq<Q>,


Both Q and Fq<Q> got replaced by F. Since Q in BitZParams is a phantom data it might not matter that much. But do investigate, these things tend to break later on.

This is by design, as the whole BitZParams was reworked to be generic over any field, not just Fq over some Q.

xrvdg · 2026-10-06T06:13:15Z

 }

-impl<const Q: u128> BitZParams<Q> {
+impl<F: BitzClaimField> BitZParams<F> {


Continuating of previous point. u128 -> BitzClaimField, while somewhere else Fq<Q> -> BitzClaimField.

xrvdg · 2026-10-06T06:16:25Z

+    pub fn fold_bound(&self) -> F::Integer {
+        let rows = u64::try_from(self.shape.rows()).expect("Too many rows");
+        let rows = F::Integer::from(rows);
+        let max_f = F::max_value().lift();
+        rows.checked_mul(&max_f).expect("Multiplication overflow")


This many expects in arithmetic code feels off. Later on the similar calculation is done without expects.

These are just some guardrails - they should probably become Result<_, _> at some point, to avoid DoS in real systems.
How do you want to treat them now?

xrvdg · 2026-10-06T06:17:24Z

+pub struct VirtualStatement<'a, F, M: VirtualMap> {
+    params: VirtualParams<F>,
    map: &'a M,
-    claim: &'a LinearClaim<Fq<Q>>,
+    claim: &'a LinearClaim<F>,


Same as before, Q -> F and Fq<Q> -> F

xrvdg · 2026-10-06T06:21:08Z

        Box::new((0..n).map(|_| F128::from(rand::random::<u128>())).collect());
    let packed: &'static _ = packed.leak();
-    let params: BitZParams<Q114> =
+    let params: BitZParams<Fq<Q114>> =


Continuation of earlier point about BitZParams. Now Fq shows up in BitZParams while it wasn't there before.

Yep, as BitZParams is now generic over a field, Fq<Q114> is the concrete field type used here.

xrvdg · 2026-10-06T06:23:10Z


        for fold in &folds {
-            transcript.prover_message(&fold.to_le_bytes());
+            transcript.prover_message(&fold.to_le_bytes().as_ref());


Would Deref be better than AsRef in this case?

I don't quite follow, how do you want to deref it if we specifically need a reference?

To give some context, I was following the existing convention (partially enforced by spongefish) - the type bound T: Encoding<[u8]> + NargSerialize + ?Sized and argument type &T, so we need to provide a reference so something that is NargSerialize. Given that [u8] is not NargSerialize, we pass in &&[u8].

xrvdg · 2026-10-06T07:08:18Z

+    let modulus = F::modulus();
+    let fold = (shape.rows() as u128) * (modulus - 1);
+    let params = BitZParams::<F>::new(shape, smallest_generator()).unwrap();


Ties back to earlier point of too much exception handling. This piece of code performs the same calculation but doesn't need expect.

This is test code, so we don't care about expects, etc.

frozenspider added 11 commits September 28, 2026 14:52

Placeholder random prime sampling

9c44abe

Placeholder dynamic field

e68c9b8

Added BitzField and changed our crates to work with it

a89dc4e

Reworked the rest of the code to be generic over BitzClaimField

69de99e

Generalize over F

e9b12a3

Removed dynamic field for now

fc0bac5

Extracted ring/field from more places; added missing projection refer…

486cb2b

…ences

Moved Spartan bench and test to cli

5e7ae78

Removed direct references to FqDefault

ca1436f

Removed (most) direct references to Big(U)Int

267dfc5

Renamed BitzRing to BitzConstraintRing

193a937

frozenspider changed the title ~~[WIP] Make project generic over semirings, rings and fields~~ [WIP] Make BitZ generic over semirings, rings and fields Sep 29, 2026

frozenspider added 5 commits September 29, 2026 17:36

Removed BitzBaseField

919d7a1

Fixed BitZParams comparison

8651bcb

Minor tweaks

f36a6a3

More minor tweaks

83e95f2

Cleanup

7b459ea

frozenspider requested review from shreyas-londhe, xrvdg and zkfriendly September 29, 2026 20:13

frozenspider changed the title ~~[WIP] Make BitZ generic over semirings, rings and fields~~ Make BitZ generic over semirings, rings and fields Sep 29, 2026

frozenspider marked this pull request as ready for review September 29, 2026 20:13

Merge branch 'main' into fs/field

c5bc8c4

# Conflicts: # crates/tests/examples/dump_bitz.rs # crates/tests/tests/host.rs # crates/tests/tests/prove.rs # crates/verifier/src/verify.rs # tooling/cli/src/end_to_end.rs # tooling/cli/tests/circuits.rs # tooling/cli/tests/end_to_end.rs

frozenspider added this pull request to stack #142 October 5, 2026 13:19

xrvdg reviewed Oct 6, 2026

View reviewed changes

Code review changes

42eed5f

frozenspider force-pushed the fs/field branch from 7e08fdb to 42eed5f Compare October 6, 2026 16:04

Conversation

frozenspider commented Sep 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xrvdg Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xrvdg Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xrvdg Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xrvdg Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xrvdg Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xrvdg Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

frozenspider commented Sep 29, 2026 •

edited

Loading

xrvdg Oct 6, 2026 •

edited

Loading

xrvdg Oct 6, 2026 •

edited

Loading

xrvdg Oct 6, 2026 •

edited

Loading

xrvdg Oct 6, 2026 •

edited

Loading

xrvdg Oct 6, 2026 •

edited

Loading

xrvdg Oct 6, 2026 •

edited

Loading