Constant product #3771

Chris-Hibbert · 2021-08-27T22:44:36Z

Constant Produce calculations

Takes pool sizes, and requested amount in or amount out, and computes poolFee, protocolFee, amount the user pays and gets, and the changes in the pool balances.

The next steps is to integrate with an autoswap contract.

refs #3340

katelynsills

This is a partial review. I'm still working on understanding the overall design, so these files need more review time:

packages/zoe/src/contracts/constantProduct/README.md
packages/zoe/src/contracts/constantProduct/swap.js
packages/zoe/test/unitTests/contracts/constantProduct/test-compareBondingCurves.js
packages/zoe/test/unitTests/contracts/constantProduct/test-compareNewSwapPrice.js
packages/zoe/test/unitTests/contracts/constantProduct/test-swapScenarios.js

katelynsills · 2021-09-14T21:00:50Z

packages/zoe/package.json

@@ -59,6 +59,7 @@
    "@agoric/install-ses": "^0.5.25",
    "ava": "^3.12.1",
    "c8": "^7.7.2",
+    "jsverify": "^0.8.4",


I wasn't getting great results from jsverify, but it was fun to play with. I think we should move this to Agoric Labs instead.

katelynsills · 2021-09-14T21:43:36Z

packages/zoe/src/contracts/constantProduct/README.md

+the pool and offer to deposit one of the two assets. They will receive an amount
+of the complementary asset that will maintain the invariant that the product of
+the pool values doesn't change. (Except that rounding is done in favor of the
+pool.) The liquidity providers are rewarded by charging a fee. 


Maybe make it clearer that we're talking about a swap, and not providing liquidity by changing the subject of the sentence?

Suggested change

pool.) The liquidity providers are rewarded by charging a fee.

pool.) A fee is charged on the swap to reward the liquidity providers.

katelynsills · 2021-09-21T00:56:21Z

packages/zoe/src/contracts/constantProduct/README.md

+
+* When the user names an input (or output) price, they shouldn't pay more
+  (or receive less) than they said.
+* The pool fee is charged against the computed side of the price.


It's not clear what the "computed side" is. Does this mean, the output if the user specified input, and the input if the user specified output? If so, can we say something like the side not specified by the user?

That's what it means.

"The side not specified by the user" doesn't seem like a phrase I can use more than once as the subject of a sentence. I'll try defining it explicitly once. If that doesn't make the explanation clear enough, I'd appreciate a suggestion.

katelynsills · 2021-09-21T01:01:13Z

packages/zoe/src/contracts/constantProduct/README.md

+|          | In (X) | Out (Y) | PoolFee | Protocol Fee | &Delta;X | &Delta;Y | pool Fee * |
+|---------|-----|-----|--------|-----|-----|-----|-----|
+| **RUN in** | RUN | BLD | BLD | RUN | sGive - PrFee | sGets | sGet - &Delta;Y
+| **RUN out** | BLD | RUN | RUN | BLD | sGive | sGets + PrFee | &Delta;X - sGive
+| **BLD in** | BLD | RUN | BLD | RUN | sGive | sGest + PrFee | &Delta;Y - sGet
+| **BLD out** | RUN | BLD | RUN | BLD | sGive - PrFee | sGets | sGive - &Delta;X


I love the idea of this table.

However, I think some of the values are wrong, or I don't understand how the table is supposed to work:

The protocol fee is always charged in RUN, but in this chart, under the Protocol Fee column, there are both "BLD" and "RUN".

If the pool fee is always charged in whatever denomination isn't specified, then I would expect the first two rows under "Pool Fee" to be "BLD" and the last two rows to be "RUN".

In the last column, "pool Fee *", why is "RUN in" sGet - ΔY, but "BLD in" is ΔY - sGet? Shouldn't these both be the same?

Also, what is sGet? Is that like want, but what the user actually gets? There are some typos so this is sometimes sGets and sGest.

The protocol fee is always charged in RUN, but in this chart, under the Protocol Fee column, there are both "BLD" and "RUN".
fixed.

If the pool fee is always charged in whatever denomination isn't specified, then I would expect the first two rows under "Pool Fee" to be "BLD" and the last two rows to be "RUN".
you're right. fixed.

In the last column, "pool Fee *", why is "RUN in" sGet - ΔY, but "BLD in" is ΔY - sGet? Shouldn't these both be the same?

As we discussed, this got broken up into multiple tables. The current version has rho + deltaX for the first two rows, and rho + deltaX for the last two.

Also, what is sGet? Is that like want, but what the user actually gets? There are some typos so this is sometimes sGets and sGest.

fixed the typo. Yes, this is what the swapper gets.

katelynsills · 2021-09-21T18:12:27Z

packages/zoe/src/contracts/constantProduct/README.md

+specifies that they want to buy BLD and are willing to spend up to 300 RUN, the
+fees will be 1 RUN and 1 BLD because the amounts are low for expository
+purposes. Since the user specified the input price, we calculate the output
+using the constant product formula for &Delta;Y. The protocol fee is always


"Wanting to buy BLD and being willing to spend up to 300 RUN" is ambiguous as to whether it's a swap in or swap out, since the user is always willing to spend up to the amount they gave. But, in the next sentence, it's said that the user specified the "input price" meaning a swap in. Can we make this clearer?

What pool fee and protocol fee configs are being used in this example? It seems like the fee being 1 RUN and 1 BLD assumes a configured fee which isn't explained.

ambiguous as to whether it's a swap in or swap out,

Thanks. I'll fix that.

I'm gonna multiply everything by 1000 so we can get clearer numbers for the fees.

katelynsills · 2021-09-22T01:37:35Z

packages/zoe/src/contracts/constantProduct/swap.js

+  return result;
+}
+
+export const swap = (


I found this function too confusing. Can we add more comments or simplify?

I added more comments. I think your suggestions have reduced some of the noise.

katelynsills · 2021-09-23T00:58:06Z

packages/zoe/src/contracts/constantProduct/swap.js

+  // output of the calculation. When the specified value is in RUN, the protocol
+  // fee will be deducted from amountGiven before adding to the pool. When BLD
+  // was specified, we add the protocol fee to amountWanted. When the specified
+  // value is in RUN, the protocol fee will be deducted from amountGiven before
+  // adding to the pool or added to amountWanted to calculate amoutOut.


I think this has duplicate explanations of how the protocol fee is charged

katelynsills · 2021-09-23T01:07:42Z

packages/zoe/src/contracts/constantProduct/README.md

+actions of arbitrageurs. At any time, a trader can come to the pool and offer to
+deposit one of the two assets. They will receive an amount
+of the complementary asset that will maintain the invariant that the product of
+the balances doesn't change. (Except that rounding is done in favor of the


Suggested change

the balances doesn't change. (Except that rounding is done in favor of the

the balances doesn't decrease. (Rounding is done in favor of the

katelynsills · 2021-09-23T01:08:19Z

packages/zoe/src/contracts/constantProduct/README.md

+
+This algorithm uses the x*y=k formula directly, without fees. Briefly, there are
+two kinds of assets, whose values are kept roughly in balance through the
+actions of arbitrageurs. At any time, a trader can come to the pool and offer to


Hmm, the way this is written doesn't make it clear a trade is intended. It sounds like adding liquidity at first.

I changed it to " At any time a trader can trade with the pool by offering to deposit ..."

katelynsills · 2021-09-23T01:09:55Z

packages/zoe/src/contracts/constantProduct/README.md

+so it charges $20. This matters whenever the values of the smallest unit of the
+currencies are significantly different, which is common in DeFi. (We refer to


I didn't understand this sentence. Do you mean, when the two kinds of tokens have very different valuations?

I think we're saying the same thing. Dollars and Pounds are at the same order of magnitude, but if we were representing them in cents and nano-Pounds, then the smallest units would be very different.

What's a clearer way to say this?

dckc

This looks good.

Since it isn't wired all the way up to the AMM, I wonder if it actually closes #3340, so you might change that to "refs #3340". It's hard to judge, since the description of #3340 is empty.

dckc · 2021-09-29T20:59:30Z

packages/zoe/src/contracts/constantProduct/README.md

+and &delta;Y. The fees are based on &delta;X, and &delta;Y. &rho; is the poolFee
+(e.g. .003).


I found "poolFee (e.g. .003)" somewhat confusing, given that much discussion of fees elsewhere is not ratios but amounts, as in "The Protocol fee is always in RUN." above.

A change here is perhaps not critical; I'm happy to see poolFeeRatio in the code.

The poolFee is sometimes charged on RUN and sometimes on BLD, so we don't construct the ratio until we know which side is which.

dckc · 2021-09-29T21:04:10Z

packages/zoe/src/contracts/constantProduct/swap.js

+  amountGiven,
+  poolAllocation,
+  amountWanted,
+  protocolFeeRatio,


The numerator and denominator of a Ratio must have the same brand, right? When I use "show type definition" here I get to contractSupport/types.js:125, which doesn't state any such constraint.

oh... ratio.js clarifies that the numerator and denominator may have different brands. Glad I checked.

dckc · 2021-09-29T21:09:50Z

packages/zoe/src/contracts/constantProduct/README.md

+
+In these tables BLD represents any collateral. The user can specify how much
+they want or how much they're willing to pay. We'll call the value they
+specified **sGive** or **sGet** and bold it. This table shows which brands the


I asked why sGet and not sWant and we talked about it a little...

FWIW, I see the code uses amountGiven and amountWanted.

packages/zoe/src/contracts/constantProduct/swap.js

packages/zoe/test/unitTests/contracts/constantProduct/test-calcDeltaY.js

dckc · 2021-09-29T22:07:37Z

yarn.lock

@@ -7993,6 +7993,16 @@ jss@10.6.0, jss@^10.5.1:
    is-in-browser "^1.1.3"
    tiny-warning "^1.0.2"

+jsverify@^0.8.4:


I don't see any package.json changes; how did we get a change to yarn.lock?

We added jsverify to do algorithmic testing. It was too expensive for CI, so we've dropped it. Thanks for catching this change I missed.

update to current master type declarations renaming Allow swapOut() to have an empty amountIn representing "no restriction" Drop enforcement in swapIn/swapOut that the quotes are in the forward direction doublePool will need to call them in the reverse direction Improved documentation in the README

Drop jsverify improve README.md add more type declarations extract common code in swapIn & swapOut drop reporting on price improvements reduce duplicated code in swap.js more tests

Don't use q() to publish prices improve internal docs and types in core.js

rename calcSwapInPrices to pricesForStatedInput and calcSwapOutPrices to pricesForStatedOutput.

Chris-Hibbert added enhancement New feature or request Medium Core Economy OBSOLETE in favor of INTER-protocol AMM labels Aug 27, 2021

Chris-Hibbert added this to the Beta Phase 4: Governance milestone Aug 27, 2021

Chris-Hibbert self-assigned this Aug 27, 2021

Chris-Hibbert force-pushed the constant-Product branch from 24ed357 to 02b65e3 Compare August 27, 2021 22:51

Chris-Hibbert requested a review from katelynsills August 27, 2021 23:26

Chris-Hibbert mentioned this pull request Aug 28, 2021

chore: make some old code for updating the constant product calculations accessible #3716

Closed

Chris-Hibbert force-pushed the constant-Product branch from 02b65e3 to ff01c5a Compare September 3, 2021 21:17

This was referenced Sep 3, 2021

Evaluate improvements to protocol fee calculation #3018

Closed

Visual documentation on constant product fee calculations #3794

Closed

This was referenced Sep 21, 2021

(After ConstantProduct AMM is approved) remove newSwap #3860

Closed

(After ConstantProduct AMM) upgrade from MultipoolAutoswap #3861

Closed

(After ConstantProduct AMM) add ParamGovernance to ConstantProduct AMM #3862

Closed

Chris-Hibbert force-pushed the constant-Product branch from 51d0f43 to 57e8f50 Compare September 23, 2021 00:36

katelynsills reviewed Sep 23, 2021

View reviewed changes

dckc approved these changes Sep 29, 2021

View reviewed changes

katelynsills and others added 8 commits September 30, 2021 13:40

chore: make code accessible

59b761d

chore: add old tests

968223e

docs: improve the README, and add some more typescript

2321dcf

chore: cleanup constantProduct calculations

efcb009

Drop jsverify improve README.md add more type declarations extract common code in swapIn & swapOut drop reporting on price improvements reduce duplicated code in swap.js more tests

docs: improve types and documentation

8a04bfb

Don't use q() to publish prices improve internal docs and types in core.js

docs: clarify and correct the README; add tests

ec3fdc4

chore: more tests, improved comments, major rename

cf3f73a

rename calcSwapInPrices to pricesForStatedInput and calcSwapOutPrices to pricesForStatedOutput.

Chris-Hibbert force-pushed the constant-Product branch from d3dd05c to cf3f73a Compare September 30, 2021 20:54

Chris-Hibbert merged commit fc84ee0 into master Sep 30, 2021

Chris-Hibbert deleted the constant-Product branch September 30, 2021 21:22

katelynsills mentioned this pull request Oct 8, 2021

Draft: highlight where constant product differs from v1 and find out why #3942

Closed

Chris-Hibbert mentioned this pull request Dec 21, 2021

use virtual pools in multipoolAutoswap #2259

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Constant product #3771

Constant product #3771

Chris-Hibbert commented Aug 27, 2021 •

edited

Loading

katelynsills left a comment

katelynsills Sep 14, 2021

Chris-Hibbert Sep 25, 2021

katelynsills Sep 14, 2021

Chris-Hibbert Sep 24, 2021

katelynsills Sep 21, 2021

Chris-Hibbert Sep 24, 2021

katelynsills Sep 21, 2021

katelynsills Sep 21, 2021

Chris-Hibbert Sep 30, 2021

katelynsills Sep 21, 2021

katelynsills Sep 21, 2021

Chris-Hibbert Sep 24, 2021

katelynsills Sep 22, 2021

Chris-Hibbert Sep 27, 2021

katelynsills Sep 23, 2021

Chris-Hibbert Sep 27, 2021

katelynsills Sep 23, 2021

Chris-Hibbert Sep 24, 2021

katelynsills Sep 23, 2021

Chris-Hibbert Sep 24, 2021

katelynsills Sep 23, 2021

Chris-Hibbert Sep 24, 2021

dckc left a comment

dckc Sep 29, 2021

Chris-Hibbert Sep 30, 2021

dckc Sep 29, 2021

dckc Sep 29, 2021

dckc Sep 29, 2021

Chris-Hibbert Sep 30, 2021

	pool.) The liquidity providers are rewarded by charging a fee.
	pool.) A fee is charged on the swap to reward the liquidity providers.

	the balances doesn't change. (Except that rounding is done in favor of the
	the balances doesn't decrease. (Rounding is done in favor of the

		so it charges $20. This matters whenever the values of the smallest unit of the
		currencies are significantly different, which is common in DeFi. (We refer to

		and δY. The fees are based on δX, and δY. ρ is the poolFee
		(e.g. .003).

Constant product #3771

Constant product #3771

Conversation

Chris-Hibbert commented Aug 27, 2021 • edited Loading

katelynsills 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

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dckc 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

Chris-Hibbert commented Aug 27, 2021 •

edited

Loading