Week 10 - Uniswap


These is a written version of Lecture #10.

In this lecture we look at an implementation of Uniswap in Plutus.

This is the last lecture in the Plutus Pioneer Program. However, there will be a special lecture once it is possible to deploy contracts to the testnet.

In this lecture we won’t be introducing any new topics or concepts. Instead we will do an end-to-end walk through of a demo that Lars wrote some months ago that clones the very popular Uniswap contract from Ethereum.

The one new thing we will look at following several requests is how you can query the endpoints created by the PAB with Curl commands just from the console.

What is Uniswap

So for those of you who haven’t heard of Uniswap, what is Uniswap?

Uniswap is a so-called DeFi, or decentralized finance application, that allows swapping of tokens without any central authority. In the case of Ethereum it’s ERC20 tokens.

So you don’t need a centralized exchange, the traditional way to exchange tokens or other crypto assets. Instead everything is governed by smart contracts and works fully automatically on the blockchain.

Another interesting feature of Uniswap is that it doesn’t discover prices the usual way with the so-called order book, but uses a different automatic price discovery system. The idea is that people can create so-called liquidity pools.

If they want other users to be able to swap two different tokens, then somebody can create a liquidity pool and put a certain amount of those two tokens in this liquidity pool, and in return the creator of the pool will receive so-called liquidity tokens that are specific to this one pool.

Other users can use that pool to swap. They take some amount of one of the tokens out in exchange for putting an amount of the other token back in.

Additionally, people can also add liquidity to the pool and receive liquidity tokens, or they can also burn liquidity tokens in exchange for tokens from the pool.

And all these features are also implemented in the version of Uniswap that works on Cardano that we’re going to look at now.


So let’s look at the various operations that are available in turn.

It all starts by somebody setting up the whole system. So some organization or entity that wants to offer this Uniswap service.

It starts with a transaction that creates a UTxO at this script address, here we call that factory for Uniswap factory. It contains an NFT that identifies the factory, the same trick that we have used a couple of times before, and as datum, it will contain the list of all liquidity pools.

So in the beginning, when the factory is just being created, that list will be empty.

Now let’s assume that one user, Alice wants to create a liquidity pool for tokens A and B. A pool that allows others to swap A against B or B against A.


She has to provide some initial liquidity for the pool. So she needs some amount of token A and some amount of token B, let’s say she has 1,000A and 2000B.

It’s important to note here that the ratio between A and B reflects Alice’s belief in the relative value of the tokens. So if she wants to set up a pool with 1000A and 2000B, then she believes that one A has the same value as two Bs.

In order to create the liquidity pool, she will create a transaction with two inputs and three outputs.


The two inputs will be the liquidity she wants to provide; the 1000A and 2000B and the Uniswap factory invoked with the create redeemer. The three outputs will be the newly-created pool.

We call it Pool AB here to indicate that it contains tokens AB, which will contain the liquidity that Alice provided; the 1000A and the 2000B and a freshly-minted token that identifies this pool, an NFT, called AB NFT here.

The datum of the pool, the 1415, will be the amount of liquidity tokens that Alice receives in return for setting up this pool and providing the liquidity. #

If you wonder about the number, that is the square root of the product of 1000 and 2000, so that’s how the initial amount of liquidity tokens is calculated. It doesn’t really matter, you could scale it arbitrarily, but that’s the way Uniswap does it.

The second output is the Uniswap factory again, with the same NFT as before that identifies it. And now the datum has been updated. So in this list that was empty before, the list of all liquidity pools, there is now an entry for the newly-created AB pool.

Finally, there’s a third output for Alice, where she receives the freshly-minted liquidity tokens, called AB here to indicate that they belong to the pool AB.

Now that the liquidity pool has been set up, other users can use it to swap.


So let’s assume that Bob wants to swap 100A against B. What will Bob do?

He will create a transaction that has two inputs and two outputs. The two inputs are the 100A he wants to swap, and the pool with the swap redeemer. The outputs are the Bs he gets in return.

In this example, that would be 181B and the updated pool. So the pool now has the additional 100A that Bob provided. So now it’s 1,100A, and it has 181B fewer than before.

It still, of course, has the NFT that identifies the pool and the datum hasn’t changed because the amount of liquidity tokens that have been minted hasn’t changed.

Now, of course, the question is, where does this 181 come from? This is this ingenious idea, how price discovery works in Uniswap.

So the rule is roughly that the product of the amounts of the two tokens must never decrease. Initially we have 1000 As and 2000 Bs and the product is 2 million.

If you do the calculation, then you will see that after the swap 1100*1819 will be slightly larger than 2 million.

If you think about it or try a couple of examples by yourself, then you will see that in principle, you will always pay this ratio of the As and Bs in the pool, at least if you swap small amounts.

So originally the ratio from A to B was 1:2, 1000:2000. 100 is relatively small in comparison to the 1000 liquidity, so Bob should roughly get 200B, but he does get less and there are two reasons for that.

One is that the amount of tokens in the liquidity pool is never allowed to go to zero. And the more of one sort you take out, the more expensive it gets - the less you get in return. So 100 depletes the pool a bit of As, so Bob doesn’t get the full factor 2 out, he gets a little bit less out. That’s exactly how this product formula works.

This also makes it ingenious, because it automatically accounts for supply and demand. If the next person also wants to swap 100A, they would get even less out.

The idea is if a lot of people want to put A in and want to get B in return, that means the demand for B is high. And that means the price of B in relation to A should rise. And that is exactly what’s happening.

So the more people do a swap in this direction, put A in and get B out, the less of the gap because the price of B rises. If there were swaps in the other direction, you would have the opposite effect.

If there’s an equal amount of swaps from A to B and B to A, then this ratio between the two amounts would stay roughly the same.

There’s an additional reason why Bob doesn’t get the full 200 that he might expect, and that is fees.

We want to incentivize Alice to set up the pool in the first place. She won’t just do that for fun, she wants to profit from it, so she wants to earn on swaps that people make.

The original product formula is modified a bit to insist that the product doesn’t only not decrease, but that it increases by a certain amount, a certain percentage, depending on how much people swap. That’s 3% in this example of the 100A that Bob swaps, and it would be the same if you swap B instead.

This is basically added on top of this product, so anytime somebody swaps, not only does the product not decrease, it actually increases. And the more people swap, the more it increases.

The idea is that if Alice now would close the pool by burning her liquidity tokens, she gets all the remaining tokens in the pool and the product would be higher than what she originally put in.

So that’s her incentive to set up the pool in the first place.

The next operation we look at is the add operation where somebody supplies the pool with additional liquidity.


So let’s say that Charlie also believes that the ratio from A to B should be 1:2 and he wants to contribute 400A and 800B.

He could also have tokens in a different ratio; the ratio reflects his belief in the true relative value of the tokens.

So Charlie wants to add 400 As and 800 Bs, and he creates a transaction with two inputs and two outputs. The inputs are the pool and his contribution, his additional liquidity, and the outputs are the updated pool where now his As and Bs have been added to the pool tokens. Note that now the datum has changed.

So we had 1415 liquidity tokens before, and now we have 1982, and the difference, the 567, go to Charlie. So that’s the second output of this transaction, and that’s the reward to Charlie for providing this liquidity.

And there the formula is a bit complicated, but in principle, it also works with the product. So you check how much the product was before and after the tokens have been added and you take into account the number that have already been minted. That also ensures that now basically Alice profits from the fees that Bob paid with the swap and Charlie doesn’t.

The specific formula doesn’t matter. The idea is just that it’s fair.

So people should receive liquidity tokens proportional to their contribution, but, if they only add liquidity after a couple of swaps have already happened, then they shouldn’t profit from the fees that have accumulated in the meantime.

The next operation we look at is called remove and it allows owners of liquidity tokens for a pool to burn some of them.


So in this example, let’s assume that Alice wants to burn all her liquidity tokens. She could also keep some, she doesn’t have to burn all, but in this example, she wants to burn all her 1415 liquidity tokens.

So for that, she creates another transaction with two inputs and two outputs, the inputs are the liquidity token she wants to burn and, of course, the pool again with the remove redeemer.

The outputs are the tokens from the pool that she receives in return, so in this case, she would get 1078A and 1869B. The second output is the updated pool.

So the 1078A and 1869B have been removed from the pool and the datum has been updated, so the 1415 liquidity tokens that Alice burnt are now subtracted from the 1982 we had before. We see that 567 are remaining which are exactly those that Charlie owns.

The formula for how many tokens Alice gets for burning liquidity tokens is again somewhat complicated, but it’s basically just proportional.

So we know how many liquidity tokens there are in total, 1982, from the datum. And she basically just gets 1415:1982 of the pool. And she gets the tokens in the ratio that they are in now.

So the 1072:1869 should be the same ratio as the 1500:2619 which means that by burning, you don’t change the ratio of the pool.

The last operation is close and it is for completely closing a pool and removing it.


This can only happen when the last remaining liquidity tokens are burnt.

So in our example, Charlie holds all the remaining 567 liquidity tokens and therefore he can close down the pool.

In order to do that, he creates a transaction with three inputs. One is the factory. Note that we only involve the factory when we create the pool and now when we close it again, which also means that the contention on the factory is not very high.

So the factory only gets involved when new pools are created and when pools are closed down, but once they exist and as long as they are not closed, the operations are independent of the factory.

We just need the factory when we want to update the list of existing pools, and by the way, this list is used to ensure that there won’t be duplicate pools. So the create operation that we looked at in the beginning will fail if somebody tries to create a pool that already exists for a pair of tokens that already exist.

Okay, so let’s go back to the close operation.

So the first input is the factory with the close redeemer, second the input is the pool that we want to close. And third input is all the remaining liquidity tokens.

We get two outputs, one is the updated factory. In this case we only had one pool, so the list only contains this one pool, and this is now removed from the list. The second output contains of all the remaining tokens, all the tokens that are still in the pool when it gets closed down.

So the remaining liquidity tokens are burnt and Charlie gets all the remaining tokens from the pool.

Uniswap in Plutus


Code for Uniswap is actually part of the Plutus repository and it is in the plutus-usecases library, split into four modules that are imported by the Plutus.Contracts.Uniswap module - OnChain, OffChain, Types and Pool.

So as the names suggest, OnChain contains the on-chain validation, OffChain contains the off-chain contracts, Types contains common types, and Pool contains the business logic, the calculations, how many liquidity tokens the creator of a pool gets, how many tokens you get when you add liquidity to a pool, how many tokens you get back when you burn liquidity tokens and under which conditions a swap is valid.

We won’t go through all of that in too much detail. It contains nothing we haven’t talked about before, but let’s at least have a brief look.

So let’s look at the Types module first.


U represents the Uniswap coin, the one that identifies the factory.


A and B are used for pool operations where we have these two sorts of tokens inside the pool.


PoolState is the token that identifies a pool, actually in the diagram earlier I said it’s an NFT. By definition, an NFT is something that only exists once. Actually here in the implementation for each pool, an identical coin is created that identifies that pool. So it’s not strictly speaking an NFT.

All the liquidity pools have one coin of that sort.


Liquidity is used for the liquidity tokens that the liquidity providers get.


And all these types are then used in the coin A type. So A is a type parameter, that’s a so-called phantom type. So that means it has no representation at run time. It’s just used to not mix up the various coins to make it easier to see what goes where, so in the datum, a coin is simply an asset class that we have seen before. Recall that AssetClass is a combination of currency symbol and token name.


Then amount is just a wrapper around integer that also contains such a phantom type parameter, so that we don’t confuse amounts for token A and token B, for example.


Then we have some helper functions, for example valueOf for constructing a Value from Coin and Amount. Here, for example, we see the use of this phantom type.

That’s actually a common trick in Haskell because now if you have, for example, pool operations that have two different coins and two different amounts for the different coins. And if the one is tagged with this type capital A and the other with capital B, then normally one could easily confuse them and somehow do operations with the one coin, with the amount for the other, and then make a mistake.

And here the type system enforces that we don’t do that. So we can only use this value of function, for example, if we a coin and an amount with the same tag type tag.

So as I said, that’s a common trick in Haskell, some lightweight type level programming that doesn’t need any fancy GHC extensions.

The unitValue function creates one amount of the given coin and isUnity checks whether this coin is contained in the value exactly once,

Then amountOf checks how often the coin is contained in the value, and finally mkCoin turns a currency symbol into a token name, into a coin.


Then we have the Uniswap type which identifies the instance of the Uniswap system we are running. So of course, nobody can stop anybody from setting up a competing Uniswap system with the competing factory, but the value of this type identifies a specific system.

And all the operations that are specific to pool will be parameterized by a value of this type, but it’s just a wrapper around the coin U. And that is just the NFT that identifies the factory.


Then we have a type for liquidity pools, and that is basically just two coins, the two coins in there.

However, there is one slight complication, only the two types of tokens inside the pool matter, not the order, there is no first or second token.


And in order to achieve that, the Eq instance has a special implementation. If we want to compare two liquidity pools, we don’t just compare the first field with the first field of the other, and the second with the second, but we also try the other way round.

So liquidity pool tokens AB would be the same as liquidity pool with tokens BA. So that’s the only slight complication here.


Then we define the actions, that’s basically the redeemers. So Create with argument LiquidityPool is for creating a new liquidity pool, Close is for closing one, Swap is for swapping, Remove is for removing liquidity and Add is for adding liquidity.

Note that in the diagrams we saw earlier for simplicity, the redeemer was called simply Create. So I didn’t mention this argument of type liquidity pool.


The datum is a bit more complex than we have seen before. It’s not just a simple integer or similarly simple type, it’s the type UniswapDatum.

There are two constructors, one for the factory and one for each pool. The factory will use the Factory constructor and the pool will use the Pool constructor.

And we saw before, the datum contains a list of all liquidity pools that currently exist. And the datum for Pool contains the LiquidityPool that we didn’t see in the diagram. It also contains something we did see in the diagram, the amount of liquidity that has been minted for this pool. Remember that gets updated when somebody adds liquidity or removes liquidity.


Next let’s look at the Pool module, which as I explained before, contains the business logic, the calculations.

So we have calculateInitialLiquidity. It gets the initial amount of token A and B that are put into the pool and returns the liquidity tokens that are returned in exchange for those.

Then calculateAdditionalLiquidity for the case that the pool already exists and somebody provides additional liquidity. So the first two arguments are the amount of token already in there. Then the third one is the liquidity tokens that have already been minted for the pool. And the next two arguments are how many As and Bs are added to the pool. The result is how many liquidity tokens will be minted in exchange for this additional amount.


The calculateRemoval function is for the opposite case. So given how many tokens are in the pool, how many liquidity tokens have been minted, how many liquidity tokens should be removed? It gives how many of tokens A and B remain in the pool.

Now checkSwap is arguably the central function of the whole Uniswap system. It calculates a swap.

This is how many As and Bs are originally in the pool and this says how many As and Bs there are after the swap in the pool. And it just returns whether that’s okay or not.

So in principle, it just checks that the product of the last two arguments is larger than the product of the first two.

And we noted before, it’s a bit more complicated because the fee is taken into account. So in this case, it’s 0.3% and you can see this is taking into account here.


It also makes sure that none of the amounts ever drops to zero. So it’s not allowed to remove all coins of one sort or of both from a pool. That also makes sense because of this product, if one of the factors was zero, then of course it couldn’t be larger than it was before.

Finally, there’s this lpTicker function. It’s just a helper function that given a liquidity pool, computes a token name for the liquidity token. The idea here is that this token name should only depend on the liquidity pool and should be unique. So each pair of tokens should result in a unique token name. In principle it just takes the currency symbols and the token names of the two tokens or coins, concatenates all of them and hashes that, and then uses the hash of the concatenation to get something unique.

A slight complication here is that again we must make sure that the order of coins in the pool doesn’t matter.


Now let’s look at the on-chain part. Only two functions are exported.

First mkUniswapValidator, to make the validator for the Uniswap, both factory and pools, because they share the same script address. They are just distinguished by the datum and by the coins that identify them.

Then validateLiquidityForging which is the monetary policy script for the liquidity tokens, but there is a lot of code in this module and we don’t want to go through it in detail, let’s rather look at the structure.


So this is the mkUniswapValidator function. This function contains all the cases for factories and pools and the various redeemers.

And we have the function validateLiquidityForging, which is the monetary policy for liquidity tokens. The idea here is that it doesn’t contain any logic and simply delegates the logic to the Uniswap validator. The way it does that is it checks the inputs of the forging transaction and checks that it either contains a factory or contains a pool, because if it does, then we know that this validator will run and then the validator can check that the forging is okay.

The way it checks whether either the factory or pool is an input is via the coins that identify a factory or pool. So it checks whether this Uniswap factory coin is in the input or whether one of the pool coins is in the input.

And then we just have helper functions for all the various cases and they look quite long but it’s all straightforward. It’s basically what we saw in the diagram, just spelled out in detail so that all these conditions are satisfied for all the different cases.


Finally, let’s look at the off-chain code.


No surprises here, it’s the usual boiler plate.


We define two different schemas. The idea is that one is for the entity that creates the Uniswap factory, and that only has one endpoint start and no parameters.

Then once that is created a second schema for people that make use of this Uniswap system, and all the contracts in here will be parameterized by the uniswap instance that the first action creates.


We make use of the state mechanism this monad writer mechanism that is accessible via tell, and basically for all the user operations, we have our own state, we call it UserContractState.

So there will be a helper contract that queries for all existing pools. So then the state would be using the Pools constructor and will return a list of pools in a simplified form - it’s just a nested pair of pairs of coin and amount in each pool.

Now the helper function to query the existing funds of a wallet that will just return a value.

Then constructors for all the other operations. So if they have happened, then one of those will be the state. For example, if we did a swap, then afterwards the status will be updated to swapped. If we removed liquidity, it will be updated to removed and so on.


Then some names for the various tokens, so “Uniswap” will be the token name of the NFT in the Uniswap factory, “Pool State” will be the token name for the coins that identify the liquidity pools.

Then our usual boiler plate to actually get a script instance.


And the policy for the liquidity tokens.

Some various helper functions,


Then all the parameters for the endpoints. So, for example, if we want to create a pool we need to know the tokens and the amounts.

If you want to swap, it must know the tokens and how much to swap and the idea in the SwapParams datatype that one of the two last fields should be zero. So if you want to put in A and get out B, we’ve would specify the spAmountA for how many As we want to put in, but we would leave the spAmountB at zero, and the other way round if we want to swap Bs against As.

CloseParams for if you want to close a pool - we just have to specify which pool. So we give the two tokens that are in there.

RemoveParams - you have to specify the pool and how much liquidity we want to burn.

AddParams - again, identify the pool and how many As and how many Bs we want to add.


Now here we have the implementation.

So start, as we saw, sets up the whole system and it again makes use of this other use case we have used before, the Currency.forgeContract to mint the factory NFT that’s then used to identify the Uniswap factory.


The create contract is the contract that creates a liquidity pool. We see all of these will be, as we mentioned before, identified by the Uniswap value, which is the result of this start contract here.

So we have create,


We have close, again parameterized by Uniswap,


remove ,




and swap.

All these functions also make use of the functions from the Pools module, that contain the business logic. So that will be used both in the validator, on the on-chain side, as well as on the off chain side in these contracts here.


The pools contract just queries the existing pools. So it looks for the factory UTxO and checks the datum, and, as we know, the datum of the factory contains the list of all pools.


And finally funds just checks our own funds, the funds in the wallet, and returns them.

So these all return values but we want to write that in the state, and this is now done these endpoint definitions.


So first we have ownerEndpoint for setting up the whole system, which just uses the stop contract.

And then we have userEndpoints, which combine all these operations that a user can do.

Now there is no return value anymore, and instead we make use of the state. So we use the Last monoid again, so only the last told state will be kept.

And we also allow for error, so if there’s an error in one of these contracts, then we will catch that error, but use a Left to write it in the state. If there was no error we write the appropriate user contract state value in the state with the Right constructor of Either.


Finally, we also have a stop endpoint that simply stops, it doesn’t do anything. At any time you can invoke stop or one of the others, and if it was one of the others then recursively userEndpoints is called again, but in the case of stop not, so if the stop endpoint is ever called then the contract stops.

There are also tests for Uniswap contained in this Plutus use-cases library, but we won’t look at them now.

Let’s rather look at the Plutus PAB part and how you can write a front-end for Uniswap.

There is actually one also contained in the Plutus repo. It’s in the plutus-pab library and in the examples folder there’s a Uniswap folder that contains an example on how to do that.


This has been copied it into our Plutus Pioneer Program repo and slightly modified it to make it more suitable for our purposes.

When we look at the Cabal file for this week’s code, there are two executables.

One uniswap-pab, which will run the PAB server, and then uniswap-client, which is a simple console based front-end for the Uniswap application.


You see, in the other-modules field there is a module Uniswap and that’s listed in both. That contains some common definitions that are used by both parts.

So let’s first look at that.


First of all, as we saw with oracle demo, we need some data type that captures the various instances we can run for the wallets. In this case, we have three.

Init hasn’t been mentioned before, that has nothing specifically to do with Uniswap, this is just used to create some example tokens and distribute them in the beginning.

UniswapStart corresponds to the Uniswap start or Uniswap owner schema that we saw just now for setting up the whole system.

UniswapUser corresponds to the other part, to the various endpoints to interact with the system. And this class constructor is parameterized by a value of type Uniswap, which is the result of starting. So after having started the system, the result would be of type Uniswap and this is then needed to parameterize the client.


Then there is some boiler plate, then this initContract function that distributes the initial funds. So it again makes use of forgeContract that we have seen before.

And it now produces tokens with token names A, B, C, D with 1 million of each. Actually it also multiplies that by the number of wallets. So in this case, I want to use four wallets, wallets one to four, so it’s actually 4 million of each of the tokens that will be forged.

And once they have been forged, they are sent from the forging wallet to all the other wallets. So one wallet forges four million of each, and then loops over the other wallets and sends them 1 million each.

So this is just needed to set up example tokens and distribute them amongst the wallets.

The cidFile function is just a helper function because in order to communicate the various contract instance IDs and other things we need, we use helper files and this is function gives the file name for a given wallet.


So now let’s look at the PAB part.

First there is the boiler plate that we saw earlier to actually hook up the PAB mechanism with actual contracts. It uses the Uniswap contracts that we just defined with the three constructors Init, UniswapStart and UniswapUser.

So, UniswapUser user will use the UniswapUser schema that we defined before, UniswapStart will use the UniswapOwner schema that we defined before and Init will use a schema without endpoints.

And we connect these constructors with actual contracts. So UniswapUser with argument us will use the userEndpoints that we looked at earlier, UniswapStart will use the ownerEndpoint, and Init will use the initContract that we just defined, and that’s just for demonstration to create these initial coins.


Now we can look at the main program.

So in the Simulator monad, we execute certain things. First we set up the whole system, we start the server and get the handle to shut it down again.

The first thing is that Wallet 1 activates the Init contract. We know from looking at the code what that will do, it will mint all these example tokens, ABCD, 4 million of each, and then distribute them so that wallets one to four end up with 1 million of each of the four different tokens.

This will concurrently start this contract, but then immediately continue - it won’t block - so we use the waitForState that we saw when we talked about oracles, to wait until Init returns.

What Init will do is that it will write the currency symbol of the forged example tokens into the state. So we wait until we see that and then we remember it and we wait until the Init contract has finished.

And then we write the currency symbol into the file symbol.json. We use the encode function from Data.Aeson, the json standard json library for Haskell.

Then again for Wallet 1, we start the Uniswap system. So we use the UniswapStart constructor and we again use waitForState to wait until we get the result. The result of the UniswapStart as we saw earlier will be a value of type Uniswap, and we need that value in order to parameterize the user contracts.

So we wait until we get this, and call it us, and now Uniswap, the system is running and now we can start the user instances for all the wallets.

So we loop over all wallets and activate the UniswapUser contract which is now parameterized by the us value we got in the previous step here.

Now we have these handles and in order to interact, to communicate, from the front-end with the server, we need these handles. So we write them into a file and this is where we use the helper function cidFile that we saw earlier.

So we will end up with four files w1.cid through to w4.cid, which contain these contract instance IDs for the four contracts.

Then we just wait until the user types a key and then we can shut down the server.


Let’s try this out with cabal run uniswap-pab.

A lot of stuff is happening. Remember, first we forge these example tokens ABCD, and then we need to distribute them to the other wallets.

Then we have to start the Uniswap system. And for that, we again have to first forge the Uniswap NFT that identifies the factory and then create the initial UTxO for the factory that contains an empty list of pools.


Now we see that all the UniswapUser contracts have started for each of the the four wallets.

If we look, we see the various files, so we can look at those.

So symbol.json is the currency symbol of the example tokens and we need that to refer to them.


And then we have these w1.cid - w4.cid files. If you look at one of those, they hold the contract instance IDs for the contract instances for the four wallets.

We need these in order to find the correct HTTP endpoints to communicate with them.


Let’s look at the client next.

As for the oracle, this is also written that in Haskell using the same library for doing HTTP requests.

In the main program we expect one command line parameter, just a number from one to four, so that the main program knows for which wallet it’s running.

Then we read the corresponding CID file to get the contract instance ID for that wallet. We read this symbol.json file to get the currency symbol of the example tokens. We use the readFile function the ByteString library, and decode comes from the Data.Aeson library to decode the json back to Haskell data type.

We check whether there was an error, and if not, we invoke the go function where we pass the contract instance ID and the currency symbol.


And then it’s just a loop. We read a command from the console, and then depending on the command, we involve various helper functions. The commands exactly correspond to the endpoints we have, except for stop, which is not implemented.

So we can query our funds, we can look for existing pools, we can create a pool, we can add liquidity to a pool, we can remove liquidity from a pool, we can close a pool, and we can swap, which is the whole point.

So for each of those, we have a constructor in the Command data type.

Because the currency symbol will always be cs - our example tokens, we don’t need to parameterise the currency symbol. And because the token names are just A, B, C and D we can just use a character for that.

So for example, Create Integer Character Integer Character means create a liquidity pool with a certain amount of a given token and a certain amount of a second token.


The readCommandIO function reads from the keyboard and then tries to pass that as a command. If it fails it will just recursively call the read command again. If it succeeds, it returns the command.

Then there are just various helper functions to convert something of type Command into the corresponding parameter types, like CreateParams or AddParams from the Uniswap module that we saw earlier.


The functions showCoinHeader and showCoin are just to make it look a bit prettier when we query the funds or the pools, and then we have the various endpoints and that all makes use of some helper functions.


There is getStatus, which we need in order to get something back from the contracts, and callEndpoint which uses the Req library, just as last time.

The interesting part is the request. It will be the post request, and we must give the instance ID. This is of type UUID, so we just convert it into a string and then pack it to a text because this HTTP library expects Text.

The request body depends on the third argument in the function.

The response will always be Unit and we just check whether we get a 200 status code or not.

The getStatus is a get request that invokes the status HTTP endpoint, again with the CID.

We have to tell it what we’re dealing with so that’s why we need the UniswapContracts type, and that’s also why this Uniswap client executable also needs access to this Uniswap module.

And then we just check if the state is empty, which happens right in the beginning because that is before anything has told anything to the state. Then we wait a second and recurse and if there’s a state (if it’s Just e), then we know that this is of type Either Text UserContractState.

Recall this UserContractState had one constructor for each of the endpoints, but if there’s an error during contract execution, we get the error message as a Text.

And if something went wrong, then we end in the third case. With these two functions - getStatus and getEndpoint - it’s easy to write all the cases for the endpoints.


So let’s maybe look at one, getFunds. We use the callEndpoint helper function that we just saw. For the endpoint named “funds”, and in this case, the argument, the request body, is just Unit.

We wait for two seconds and then use the getStatus helper function. If we get a Right, then we show the funds, otherwise we recurse.

So we wait until we get the Right, because in this case “funds” should never fail. There’s no way that can fail, therefore we can safely wait forever.


The getPools function is similar. It’s more or less the same, except that instead of “funds”, we have “pools”.

Let’s look at one more example, for creating a pool.


Again we call the endpoint and we wait for two seconds. Here something could actually go wrong. For example, if we try to create a pool where both coins are the same, or if we specify a larger liquidity than exists in the wallet, then we would get an error.

So in this case, if we get an error, we just log it to the console.

The cases for all the other endpoints are very similar.

Trying it Out

Now let’s try it out.

Let’s start three instances for wallets one, two and three, and try to recreate the scenario from the diagrams at the beginning.

We start it simply by using cabal run with a command line parameter for the number of the wallet.

cabal run uniswap-client -- 1

And we do the same for wallets 2 and 3.


Here we see the log messages for the contract instance ID and the symbol for token that we can use.

So what can we do?

We can, for example, query our funds.


And we see that we have A, B, C, D with 1 million each and 100,000 Ada.

We can also look for pools, but right now, there shouldn’t be any, and indeed none are listed.


So let’s switch to Wallet 1, let’s say this is Alice, Bob is 2 and Charlie is 3.

In the diagrams, we started with Alice setting up a liquidity pool for 1000 tokens A and 2000 B tokens.

So to do this here, we can type

Create 1000 'A' 2000 'B'

Remember that was of type Char, so we use single quotes.

We get the created status spec.


So it seems to have worked. We can query for pools again, and indeed there is one now.


We see that it has A and B and with the correct amounts, 1000 and 2000 respectively.

The next step was that Bob swaps 100A for Bs. So, in the console that is running Bob’s wallet, we can write

Swap 100 'A' 'B'

Let’s check how many funds Bob now has. As expected, he has 100 fewer As and 181 as many Bs.


Next Charlie added liquidity.

Add 400 'A' 800 'B'

Now we check the pools again.

It’s 1500 and 2619. We had 1000 at the beginning, then 100 where added by Bob and now 400 by Charlie.


Now, if we go back to Alice, she wants to remove her liquidity. So let’s first query her funds.


So she has less A and Bs now because she provided them as liquidity for the pool, but she has 1415 of the liquidity token.

So for example, she can burn the liquidity tokens and get tokens in exchange. She doesn’t have to burn all, but in the diagram she did. So let’s do this, so

Remove 1415 'A' 'B'

And let’s query her funds again.


So now she doesn’t have liquidity token anymore, but she got As and Bs back.

So she received 1869Bs and 1070 As.

The last step was that Charlie closes the pool. So let’s switch to Charlie

Close 'A' 'B'

And if now we look for pools, then again, we don’t get any. So it all seems to work.

Finally, we will look at how to use the front-end without Haskell and just use something like curl.

Let’s look at the file status.sh which you will find in the code folder.


This script expects one argument, that’s the wallet.

Then we just curl to this URL and interpolate the content of the correct wallet file given by the first parameter.

And because the output is very unwieldy, we pipe it through jq and get the current state and observable state of the resulting json.

So if I try this right now for Wallet 1, for example, we see that wallet one at the moment has these amounts of the tokens ABCD.


At least that was the last status. Maybe it’s not up to date.

Let’s also look at the funds.sh script.


So that again, only takes one parameter, one argument, the wallet, in order to put the correct instance ID in the URL, and then uses the funds endpoint.

And this is a POST request, so we need a request body. This is Unit because the funds endpoint doesn’t require any arguments except the Unit argument.

A bit more interesting is what to do with the post request that do have interesting arguments. For example, if now, Wallet 1 wants to create a pool again with 1000A and 2000 B. In that case we need a request body for the correct parameters for the CreateParams.

So let’s look at this in create.sh


In principle, the curl is simple, so now again, contract instance ID but now with endpoint create. But the question is what to write in this body.

We use similar arguments to in the Haskell implementation. So first the wallet and then the A amount, A token, B amount, and B token.

So maybe we should first check whether it works.

./create.sh 1 1000 A 2000 B

And now if we wait a few seconds and then query the status, it should have updated.


We can now run

./status.sh 1

Now we have this new liquidity pool with A and B.

So the remaining question is, how did we get the JSON body for the curl call? It’s hard to do this by hand, but if we look back at the Haskell output, what we did was here, for example, for create, to always write the URL we are calling and also the request body.

And we can check the code for this. If we look at uniswap-client. It is in the helper function callEndpoint on lines 216 and 217.


So we get the a, that’s just a Haskell value with an instance of ToJSON so that can be encoded to JSON, and we just use encode from the Aeson library. This is now a byte string, but in order to write that to the console, we need two strings, so we use something from the ByteString library


And we unpack this ByteString to a string and then log it. So this is a good way to figure out what requests bodies to use.

You don’t, of course, have to write a whole program, you can also do that in the REPL. You just need a value of the correct type and then use Aeson to encode it and look at the result and then you see the shape of the json that is expected.