Feat/balancer swap #2290

gztensor · 2025-12-15T16:43:05Z

Description

Implement PalSwap (similar to balancer swap) to replace Uniswap V3.

Type of Change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
Documentation update
Other (please describe):

Breaking Change

This PR changes the following externally accessible entities:

add_liqudiity extrinsic: Remove tick_low and tick_high parameters.
Positions map is replaced with PositionsV2 map
Following state maps removed:
- AlphaSqrtPrice
- CurrentLiquidity
- CurrentTick
- FeeGlobalTao
- FeeGlobalAlpha
- ScrapReservoirTao (applied before removing)
- ScrapReservoirAlpha (applied before removing)
- Ticks
- SwapV3Initialized
- TickIndexBitmapWords

Checklist

I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have run ./scripts/fix_rust.sh to ensure my code is formatted and linted correctly
I have made corresponding changes to the documentation
My changes generate no new warnings
I have added tests that prove my fix is effective or that my feature works
New and existing unit tests pass locally with my changes
Any dependent changes have been merged and published in downstream modules

…ity + tests

…rate_reset_unactive_sn

gztensor · 2026-01-08T21:03:00Z

pallets/subtensor/src/coinbase/run_coinbase.rs

adjust_protocol_liquidity returns the actual adjustment that needs to be made to reserves, which includes previously collected tao and alpha fees.

When tao fees are charged, the tao is coming from user balance to FeesTao map, not accounted for in SubnetTAO, so it needs to be added here.

When alpha fees are charged, the alpha is coming from user's stake, and is also not accounted for in SubnetAlphaIn.

gztensor · 2026-01-08T21:04:47Z

pallets/subtensor/src/staking/move_stake.rs

This is still an approximation.

No formula for this exists, but we could do newtons method. We have an implementation if you want it.

@gztensor see discord.

gztensor · 2026-01-08T21:05:35Z

pallets/subtensor/src/staking/stake_utils.rs

Continue removing storage for provided reserves here

gztensor · 2026-01-08T21:06:33Z

pallets/swap/src/position.rs

Positions are unused atm, but I preserved the struct for later use, when we have user liquidity.

gztensor · 2026-01-08T21:07:09Z

pallets/swap/src/tick.rs

This is part of uniswap v3 logic. Not needed anymore.

…itialization in tests

0xcacti

We will provide one more review, looking at pallets/swap/src/pallet/swap_step.rs and also pallets/swap/src/pallet/impls.rs

it got late and we don't want to lose our progress.

0xcacti · 2026-01-09T02:28:44Z

pallets/swap/src/pallet/balancer.rs

+            Perquintill::from_rational(1u128, 2u128)
+        };
+
+        self.set_quote_weight(new_reserve_weight)


looks like you forgot to set the base weight here? we only set the quote weight.

how could we have possibly gotten through testing? are we missing something?

nevermind, base is always looked up as 1 - quote

Yes, the only weight that is stored and needs to be set is quote. Base weight is always calculated from quote weight.

0xcacti · 2026-01-15T05:36:06Z

pallets/subtensor/src/coinbase/block_emission.rs

We should just delete this whole function in which we have changes. The function isn't used anywhere other than in a test, and the logic for emissions is wrong.

Agree. I did it in a separate PR because it is not related to balancer:
#2346

0xcacti · 2026-01-15T05:36:20Z

pallets/subtensor/src/coinbase/root.rs

I love cleaning.

0xcacti · 2026-01-15T05:51:23Z

pallets/swap-interface/src/lib.rs

Yeah, good point. correc.t

0xcacti · 2026-01-15T05:51:44Z

pallets/swap/src/pallet/impls.rs

-                Self::add_liquidity_at_index(netuid, position.tick_low, liquidity_delta, false);
-                Self::add_liquidity_at_index(netuid, position.tick_high, liquidity_delta, true);
-            }
+    ) -> (TaoCurrency, AlphaCurrency) {


This is high level correct.

of course, you know that I think it's really weird the way that the reserves are in another pallet so we update them in the calling context rather than here, but I suppose my view there doesn't really matter.

As is, it's correct.

0xcacti · 2026-01-15T06:09:05Z

pallets/swap/src/pallet/hooks.rs

I don't entirely understand this file

Hooks in substrate execute without user interaction on certain events. In this case, I am using on_runtime_upgrade to run the migration. It is executed in the beginning of the block following after the runtime upgrade.

0xcacti · 2026-01-15T06:09:59Z

pallets/swap/src/pallet/mod.rs

This is so much nicer.

0xcacti · 2026-01-15T06:27:54Z

pallets/swap/src/pallet/balancer.rs

+    /// Constraints limit balancer weights within certain range of values:
+    ///   - Both weights are above minimum
+    ///   - Sum of weights is equal to 1.0
+    fn check_constraints(quote: Perquintill) -> bool {


We are working on inventing a more advanced theory here for when we reach the MIN_WEIGHT what we should do. All good for now, but these constraints may change in the future.

0xcacti · 2026-01-15T07:20:46Z

pallets/swap/src/pallet/balancer.rs

+    pub fn update_weights_for_added_liquidity(
+        &mut self,
+        tao_reserve: u64,
+        alpha_reserve: u64,
+        tao_delta: u64,
+        alpha_delta: u64,
+    ) -> Result<(), BalancerError> {


Just want to make sure that here we are passing in the actual_delta,
I.e. the quantity injected + fees

I checked and it is because it's only called by adjust_protocol_liqudity which handles this.

0xcacti · 2026-01-15T07:28:17Z

pallets/swap/src/pallet/balancer.rs

+    ///   y - tao reserve
+    ///   w1 - base weight
+    ///   w2 - quote weight
+    pub fn calculate_current_liquidity(&self, tao_reserve: u64, alpha_reserve: u64) -> u64 {


Do we even need this function for anything? could probably get rid of it?

I guess if we want user liquidity at some point in the future.

Yes. I am leaving a few unused things behind. In case we decide to have user liquidity, they will help us have much smoother transition and less or no migrations:

Positions

Current liquidity

Tests for user liquidity (all commented out)

0xcacti · 2026-01-16T05:21:16Z

pallets/swap/src/pallet/impls.rs

-
-        let epsilon = U64F64::saturating_from_num(0.000000000001);
+        // Create balancer based on price
+        let balancer = Balancer::new(if let Some(price) = maybe_price {


nice handles existing subnets just in case. excellent.

just to be clear, when we pass in maybe_price

is it the case that new subnets won't have a price at the time of initialization?

New subnets get their price right here before the first swap. This price is based on whatever reserves were assigned at creation and 0.5 weights.

0xcacti · 2026-01-16T05:32:02Z

pallets/swap/src/pallet/impls.rs

+        let balancer = Balancer::new(if let Some(price) = maybe_price {
+            // Price is given, calculate weights:
+            // w_quote = y / (px + y)
+            let px_high = (price.saturating_to_num::<u64>() as u128)


We should have a test just to make sure this is correct. I mean it makes sense, mult the integer part and the reserve then the fractional part and the reserve and do some fun type casting to get to the end type, but just want to be sure because it's nontrivial.

There is a test for this that involves both integer and fractional part of price:

subtensor/pallets/swap/src/pallet/tests.rs

Line 2723 in a3756d9

// Test that subnet price is still 1.23^2

Explanation: Converting U64F64 to u64 is the same as rounding to the floor integer. I am doing this to avoid both overflow and precision loss of u64 by U64F64 multiplication. This is very hypothetical, but if alpha_reserve is 21M and the price is 1000., it overflows U64F64.

0xcacti · 2026-01-16T05:36:59Z

pallets/swap/src/pallet/impls.rs

        );

-        Self::maybe_initialize_v3(netuid)?;
+        Self::maybe_initialize_palswap(netuid, None)?;


Man should we just pass in a price of None here? what if it's a subnet with a price?

If it's a subnet with a price, it would be handled in the migration right after the balancer upgrade and maybe_initialize_palswap will be a noop in this case. Here it is a non-noop only for new subnets.

gztensor added 30 commits December 15, 2025 11:31

High level architecture for palswap in place, everything compiles

46164e8

Merge branch 'devnet-ready' into feat/balancer_swap

48866a9

Use Sam's bigmath crate for exponentiation

c9aedf6

Adjust fuzzy epsilon

1bc3092

Use rayon for fuzzy testing

81eab8e

Cleanup full vs limited range testing

8bdafa7

Update bigmath, debug current_price, implement adjust_protocol_liquid…

b91ca21

…ity + tests

Implement delta_in and tests

f3daa79

Merge devnet-ready

ef71b3c

Update bigmath, add testcases

0fe9842

Merge branch 'devnet-ready' into feat/balancer_swap

921ace4

Basic swap works

9292044

Swap tests in progress

30be6ea

Fix test_clear_protocol_liquidity_green_path

22019e3

Fix test_claim_root_with_run_coinbase

9c9b646

Add swap initialization to interface

a27f1b5

Fix test_coinbase_subnet_terms_with_alpha_in_gt_alpha_emission

afa0d45

Fix test_mining_emission_distribution_with_no_root_sell

243df1e

Fix test_mining_emission_distribution_with_root_sell

d2f9e2b

Fix test_pending_emission_start_call_not_done

96402fd

Remove non-zero delta-out requirement for zero delta-in. Fix test_mig…

6c55c7b

…rate_reset_unactive_sn

Fix test_add_stake_insufficient_liquidity_one_side_ok

b8d5ffe

Fix dissolve_clears_all_per_subnet_storages

1c663f1

Fix test_large_swap, improve error messaging

25a9e5d

Fix test_add_stake_limit_fill_or_kill

213a984

Fix test_max_amount_add_dynamic

b755f0e

Fix test_max_amount_move_dynamic_dynamic

26cca0b

Re-enable balancer math test cases

7c51712

Fix test_max_amount_move_dynamic_stable

6ce98e2

Fix test_max_amount_move_stable_dynamic

483dcd4

Remove todo() macros from position.rs file

b2744c8

gztensor commented Jan 8, 2026

View reviewed changes

pallets/swap/src/tick.rs

Copy link

Contributor Author

gztensor Jan 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is part of uniswap v3 logic. Not needed anymore.

gztensor added 4 commits January 8, 2026 16:14

Address a todo in test_remove_stake_edge_alpha

1dd3e67

Remove todo macros from liquidity extrinsics, make them return error

56c001e

Fix clippy (all except SafeInt)

c2bbe99

Add more test cases for test_convert_deltas, correct order of swap in…

244fbfc

…itialization in tests

gztensor added the skip-cargo-audit This PR fails cargo audit but needs to be merged anyway label Jan 9, 2026

gztensor added 12 commits January 9, 2026 12:47

Disable add/remove/modify liquidity benchmarks

8d19821

Add reminders to removed benchmarks

80e9cec

Fix custom lints

fbb015c

Fix zepter ci

65bb1c2

Fix benchmarks

87fa1dc

Fix remove_stake_full_limit

b021c56

Add clippy ignore for SafeInt division

6723152

Add tests for palswap initialization with price in migration

8d48806

Fix price initialization

55f9bba

Fix remove_stake_full_limit benchmark

485949d

Merge branch 'devnet-ready' into feat/balancer_swap

9f5ca77

Fix remove_stake_full_limit benchmark

b4ecbc3

0xcacti approved these changes Jan 15, 2026

View reviewed changes

gztensor added 2 commits January 15, 2026 09:02

Merge branch 'devnet-ready' into feat/balancer_swap

3b18a2e

Add R&D tests for exp_scaled

a3756d9

0xcacti reviewed Jan 16, 2026

View reviewed changes

gztensor added 3 commits January 16, 2026 09:43

Fix typo

06c2f71

Merge branch 'devnet-ready' into feat/balancer_swap

3143a79

Merge branch 'devnet-ready' into feat/balancer_swap

903848b

Feat/balancer swap #2290

Are you sure you want to change the base?

Feat/balancer swap #2290

Uh oh!

Conversation

gztensor commented Dec 15, 2025

Description

Type of Change

Breaking Change

Checklist

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!

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!

0xcacti left a comment

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!

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!

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!

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!

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!

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees