Wip FixedPoint refactor by ryangoree · Pull Request #184 · delvtech/hyperdrive-rs

ryangoree · 2024-07-29T08:34:25Z

Resolved Issues

closes #127

Description

dpaiton · 2024-07-31T22:28:17Z

-        // Calculate exp(y * ln(x)) to get x^y
-        Self::exp(ylnx)?.try_into()
-    }
+            // FIXME: What's the correct return value here?


python float handles this by including special values inf and -inf and nan.

We mimic that behavior in the python fixedpointmath lib:
https://github.com/delvtech/fixedpointmath/blob/main/fixedpointmath/fixed_point.py#L164

then we check that member variable whenever people want to perform an operation.

You can see all of the rules here: https://github.com/delvtech/fixedpointmath/blob/main/tests/test_fp_nonfinite_checks.py

dpaiton · 2024-08-01T03:19:21Z

-            true => self.share_reserves() - long_exposure - self.minimum_share_reserves(),
-            false => fixed!(0),
-        }
+    // FIXME: This function is redundant with `calculate_solvency`.


I agree that we should only have one function, but I think calculate_idle_share_reserves is more clear than calculate_solvency. The former is more explicit in what is being computed.

If we allowed negative idle share reserves, then we could also implement a check_solvency function that returns a bool based on the sign of calculate_idle_share_reserves. Or we do not allow negative idle share reserves and leave it to the developer to know that this is the function to run to check & calculate solvency.

…ypes.

ryangoree · 2024-08-09T17:34:59Z

@@ -1,859 +1,35 @@
+//! A generic fixed point wrapper around the `U256` type from ethers-rs.


Since ethers-rs is being deprecated, I checked out alloy to see what they use for ints now and they're using ruint which looks nice. Also found uint. It might be nice to refactor to something similar soon.

ryangoree · 2024-08-12T11:48:26Z

closed in favor of #186

Related to: #127 (replaces wip pr: #184) # Description This refactors the `FixedPoint` type to be generic with implementations for `i128`, `u128`, `I256`, and `U256`. For the most part, the library can be used the same, but with a type param, eg: ```diff - fn foo() -> FixedPoint {} + fn foo() -> FixedPoint<U256> {} ``` However, the API has been expanded and there's plenty code reduction we can do in future PRs. It's probably also worth rethinking argument types now that the library can support negative numbers and be bound by different limits. ## Expanded API I wrote doc comments throughout to aid in adoption, but below is a skeleton containing most of the new API to get a quick sense: ```rs pub struct FixedPoint<T: FixedPointValue> { raw: T, decimals: u8, } impl<T: FixedPointValue> FixedPoint<T> { pub const MIN: Self; pub const MAX: Self; // Constructors // pub fn new<V: Into<T>>(value: V) -> Self; pub fn try_from<V: TryInto<T> + Debug>(value: V) -> Result<Self>; pub fn from_sign_and_abs(sign: FixedPointSign, abs: U256) -> Result<Self>; pub fn from_dec_str(s: &str) -> Result<Self>; pub fn saturate_sign(sign: FixedPointSign) -> Self; pub fn zero() -> Self; pub fn one(&self) -> Self; // Getters // pub fn raw(&self) -> T; pub fn decimals(&self) -> u8; pub fn sign(&self) -> FixedPointSign; // Predicates // pub fn is_negative(&self) -> bool; pub fn is_positive(&self) -> bool; pub fn is_zero(&self) -> bool; // Conversion to other FixedPoint types // pub fn change_type<U: FixedPointValue + TryFrom<T>>(self) -> Result<FixedPoint<U>>; // Conversion to unsigned & signed ethers types // pub fn to_u256(self) -> Result<U256>; pub fn to_i256(self) -> Result<I256>; // Conversion to unsigned and signed std types // pub fn to_u128(self) -> Result<u128>; pub fn to_i128(self) -> Result<i128>; // Formatting // pub fn to_scaled_string(&self) -> String; // Math // pub fn abs(&self) -> Self; pub fn unsigned_abs(&self) -> FixedPoint<U256>; // `U256` to avoid overflow. pub fn abs_diff(&self, other: Self) -> FixedPoint<U256>; pub fn mul_div_down(self, other: Self, divisor: Self) -> Self; pub fn mul_div_up(self, other: Self, divisor: Self) -> Self; pub fn mul_down(self, other: Self) -> Self; pub fn mul_up(self, other: Self) -> Self; pub fn div_down(self, other: Self) -> Self; pub fn div_up(self, other: Self) -> Self; pub fn pow(self, y: Self) -> Result<Self>; } // Direct conversions between primitive types and FixedPoint for any // `FixedPointValue` that can be converted to and from the primitive type. conversion_impls!(i8, u8, i16, u16, i32, u32, i64, u64, isize, usize, [u8; 32], bool); // Value // /// A value that can be used to perform fixed-point math. pub trait FixedPointValue: /* ... */ { const MIN: Self; const MAX: Self; const MAX_DECIMALS: u8 = 18; fn is_signed() -> bool; fn is_negative(&self) -> bool; fn is_positive(&self) -> bool; fn is_zero(&self) -> bool; fn flip_sign(self) -> Self; fn flip_sign_if(self, condition: bool) -> Self; fn abs(self) -> Self; fn unsigned_abs(self) -> U256; pub fn from_u256(u: U256) -> Result<Self>; pub fn from_u128(u: u128) -> Result<Self>; pub fn to_u256(self) -> Result<U256>; pub fn to_u128(self) -> Result<u128>; } // Conversions traits // pub trait Fixed: FixedPointValue { fn fixed(self) -> FixedPoint<Self>; } // Add `.fixed()` to all types that implement `FixedPointValue`. impl<T: FixedPointValue> Fixed for T {} pub trait ToFixed: Sized + Debug { fn to_fixed<T: FixedPointValue + From<Self>>(self) -> FixedPoint<T>; fn try_to_fixed<T: FixedPointValue + TryFrom<Self>>(self) -> Result<FixedPoint<T>>; } // Add `.to_fixed()` & `.try_to_fixed()` to all sized types that implement // `Debug`. impl<T: Sized + Debug> ToFixed for T {} // Macros // macro_rules! uint256 {} // -> `U256` macro_rules! int256 {} // -> `I256` macro_rules! fixed {} // -> `FixedPoint<T>` (where `T` is inferred) macro_rules! fixed_u256 {} // -> `FixedPoint<U256>` macro_rules! fixed_i256 {} // -> `FixedPoint<I256>` macro_rules! fixed_u128 {} // -> `FixedPoint<u128>` macro_rules! fixed_i128 {} // -> `FixedPoint<i128>` ```

ryangoree mentioned this pull request Jul 29, 2024

[WIP] fixed-point formatting, refactor from dnum, more (to be broken up) delvtech/hyperdrive-frontend#1299

Closed

jalextowle reviewed Jul 29, 2024

View reviewed changes

Comment thread crates/fixedpointmath/src/value.rs Outdated

jalextowle reviewed Jul 29, 2024

View reviewed changes

Comment thread crates/fixedpointmath/src/lib.rs Outdated

jalextowle reviewed Jul 29, 2024

View reviewed changes

Comment thread crates/fixedpointmath/src/lib.rs Outdated