SwapMath: One Step of the Swap

Key takeaway: computeSwapStep prices a swap over a single stretch of constant liquidity: given the current price, a target price that must not be exceeded, the liquidity, the remaining amount (positive = exact input, negative = exact output), and the fee, it returns the new price and the step's amountIn, amountOut, and feeAmount. Direction is inferred from the price pair alone (zeroForOne = current >= target), the fee is stripped off the input before the price math, and every rounding decision — input up, output down, fee up — pushes the error toward the pool. The asymmetric fee formula mulDivRoundingUp(amountIn, feePips, 1e6 - feePips) grosses the fee back up from the net input; getting that denominator wrong is a classic fork bug.

What You Are Building

The pure function at the center of the swap loop. Section 11's while loop does nothing but call computeSwapStep repeatedly — once per stretch of constant liquidity — and book the results. All the economics of a single step live here, composed entirely from the SqrtPriceMath functions you built in section 3.