Flash swaps are an integral feature of Pegasys. In fact, under the hood, all swaps are actually flash swaps! This simply means that pair contracts send output tokens to the recipient before enforcing that enough input tokens have been received. This is slightly atypical, as one might expect a pair to ensure it's received payment before delivery. However, because Syscoin transactions are atomic, we can roll back the entire swap if it turns out that the contract hasn't received enough tokens to make itself whole by the end of the transaction.
To see how this all works, let's start by examining the interface of the
For the sake of example, let's assume that we're dealing with a DAI/WSYS pair, where DAI is
token0 and WSYS is
amount1Out specify the amount of DAI and WSYS that the
msg.sender wants the pair to send to the
to address (one of these amounts may be 0). At this point you may be wondering how the contract receives tokens. For a typical (non-flash) swap, it's actually the responsibility of
msg.sender to ensure that enough WSYS or DAI has already been sent to the pair before
swap is called (in the context of trading, this is all handled neatly by a router contract). But when executing a flash swap, tokens do not need to be sent to the contract before calling
swap. Instead, they must be sent from within a callback function that the pair triggers on the
To differentiate between the "typical" trading case and the flash swap case, pairs use the
data parameter. Specifically, if
data.length equals 0, the contract assumes that payment has already been received, and simply transfers the tokens to the
to address. But, if
data.length is greater than 0, the contract transfers the tokens and then calls the following function on the
The logic behind this identification strategy is simple: the vast majority of valid flash swap use cases involve interactions with external protocols. The best way to pass information dictating how these interactions happen (function arguments, safety parameters, addresses, etc.) is via the
data parameter. It's expected that
data will be
abi.decoded from within
pegasysCall. In the rare case where no data is required, callers should ensure that
data.length equals 1 (i.e. encode a single junk byte as
bytes), and then ignore this argument in
pegasysCall with the
sender argument set to the
msg.sender of the
amount1 are simply
There are several conditions that should be checked in all
The first 2 lines simply fetch the token addresses from the pair, and the 3rd ensures that the
msg.sender is an actual Pegasys pair address.
At the end of
pegasysCall, contracts must return enough tokens to the pair to make it whole. Specifically, this means that the product of the pair reserves after the swap, discounting all token amounts sent by 0.3% LP fee, must be greater than before.
In the case where the token withdrawn is not the token returned (i.e. DAI was requested in the flash swap, and WSYS was returned, or vice versa), the fee simplifies to the simple swap case. This means that the standard
getAmountIn pricing function should be used to calculate e.g., the amount of WSYS that must be returned in exchange for the amount of DAI that was requested out.
This type of fee calculation gives a slight advantage to the caller, as the fee derived from repayment in a corresponding token will always be slightly less than the fee derived from a direct token repayment, as a result of the difference between the amount required to pay back a swap, versus the amount withdrawn and then directly returned. The approximate comparison of fees is ~ 30 bps for a swap fee vs. 30.09 bps for a direct repayment.
In the case where the token withdrawn is the same as the token returned (i.e. DAI was requested in the flash swap, used, then returned, or vice versa with WSYS), the following condition must be satisfied:
DAIReservePre - DAIWithdrawn + (DAIReturned * .997) >= DAIReservePre
It may be more intuitive to rewrite this formula in terms of a "fee" levied on the withdrawn amount (despite the fact that Pegasys always levies fees on input amounts, in this case the returned amount, here we can simplify to an effective fee on the withdrawn amount). If we rearrange, the formula looks like:
(DAIReturned * .997) - DAIWithdrawn >= 0
DAIReturned >= DAIWithdrawn / .997
So, the effective fee on the withdrawn amount is
.003 / .997 ≈ 0.3009027%.
For further exploration of flash swaps, see the whitepaper.