X-Git-Url: https://git.ralfj.de/rust-101.git/blobdiff_plain/369875f931d841112dd2b6651fc968bb6c569cdb..26ec22862fa81bba24e3f646cf482439999b0021:/solutions/src/bigint.rs diff --git a/solutions/src/bigint.rs b/solutions/src/bigint.rs index 80ac63d..5869ba8 100644 --- a/solutions/src/bigint.rs +++ b/solutions/src/bigint.rs @@ -37,6 +37,20 @@ fn overflowing_add(a: u64, b: u64, carry: bool) -> (u64, bool) { } } +// Subtract with carry, returning the difference and the carry +fn overflowing_sub(a: u64, b: u64, carry: bool) -> (u64, bool) { + let diff = u64::wrapping_sub(a, b); + let carry_n = if carry { 1 } else { 0 }; + if diff <= a { // the first diff did not wrap + let diff_total = u64::wrapping_sub(diff, carry_n); + let had_wrap = diff_total > diff; + (diff_total, had_wrap) + } else { // the first diff did wrap + // it is impossible for this to wrap again, as we are just substracting 0 or 1 + (diff - carry_n, true) + } +} + impl BigInt { /// Construct a BigInt from a "small" one. pub fn new(x: u64) -> Self { @@ -55,16 +69,16 @@ impl BigInt { } } - /// Construct a BigInt from a vector of 64-bit "digits", with the last significant digit being first + /// Construct a BigInt from a vector of 64-bit "digits", with the last significant digit being first. Solution to 05.1. pub fn from_vec(mut v: Vec) -> Self { - // remove trailing zeroes + // remove trailing zeros while v.len() > 0 && v[v.len()-1] == 0 { v.pop(); } BigInt { data: v } } - /// Increments the number by 1. Solution to 05.1. + /// Increments the number by 1. pub fn inc1(&mut self) { let mut idx = 0; // This loop adds "(1 << idx)". If there is no more carry, we leave. @@ -131,6 +145,7 @@ impl PartialEq for BigInt { } impl Minimum for BigInt { + // This is essentially the solution to 06.1. fn min<'a>(&'a self, other: &'a Self) -> &'a Self { debug_assert!(self.test_invariant() && other.test_invariant()); if self.data.len() < other.data.len() { @@ -139,16 +154,14 @@ impl Minimum for BigInt { other } else { // compare back-to-front, i.e., most significant digit first - let mut idx = self.data.len()-1; + let mut idx = self.data.len(); while idx > 0 { + idx = idx-1; if self.data[idx] < other.data[idx] { return self; } else if self.data[idx] > other.data[idx] { return other; } - else { - idx = idx-1; - } } // the two are equal return self; @@ -165,12 +178,14 @@ impl fmt::Debug for BigInt { impl<'a, 'b> ops::Add<&'a BigInt> for &'b BigInt { type Output = BigInt; fn add(self, rhs: &'a BigInt) -> Self::Output { - let mut result_vec:Vec = Vec::with_capacity(cmp::max(self.data.len(), rhs.data.len())); + let max_len = cmp::max(self.data.len(), rhs.data.len()); + let mut result_vec:Vec = Vec::with_capacity(max_len); let mut carry:bool = false; // the carry bit - for (i, val) in (&self.data).into_iter().enumerate() { + for i in 0..max_len { // compute next digit and carry + let lhs_val = if i < self.data.len() { self.data[i] } else { 0 }; let rhs_val = if i < rhs.data.len() { rhs.data[i] } else { 0 }; - let (sum, new_carry) = overflowing_add(*val, rhs_val, carry); + let (sum, new_carry) = overflowing_add(lhs_val, rhs_val, carry); // store them result_vec.push(sum); carry = new_carry; @@ -179,8 +194,8 @@ impl<'a, 'b> ops::Add<&'a BigInt> for &'b BigInt { result_vec.push(1); } // We know that the invariant holds: overflowing_add would only return (0, false) if - // the arguments are (0, 0, false), but we know that in the last iteration, `val` is the - // last digit of `self` and hence not 0. + // the arguments are (0, 0, false), but we know that in the last iteration, one od the two digits + // is the last of its number and hence not 0. BigInt { data: result_vec } } } @@ -209,11 +224,81 @@ impl ops::Add for BigInt { } } +impl<'a, 'b> ops::Sub<&'a BigInt> for &'b BigInt { + type Output = BigInt; + fn sub(self, rhs: &'a BigInt) -> Self::Output { + let max_len = cmp::max(self.data.len(), rhs.data.len()); + let mut result_vec:Vec = Vec::with_capacity(max_len); + let mut carry:bool = false; // the carry bit + for i in 0..max_len { + // compute next digit and carry + let lhs_val = if i < self.data.len() { self.data[i] } else { 0 }; + let rhs_val = if i < rhs.data.len() { rhs.data[i] } else { 0 }; + let (sum, new_carry) = overflowing_sub(lhs_val, rhs_val, carry); + // store them + result_vec.push(sum); + carry = new_carry; + } + if carry { + panic!("Wrapping subtraction of BigInt"); + } + // We may have trailing zeroes, so get rid of them + BigInt::from_vec(result_vec) + } +} + +impl<'a> ops::Sub for &'a BigInt { + type Output = BigInt; + #[inline] + fn sub(self, rhs: BigInt) -> Self::Output { + self - &rhs + } +} + +impl<'a> ops::Sub<&'a BigInt> for BigInt { + type Output = BigInt; + #[inline] + fn sub(self, rhs: &'a BigInt) -> Self::Output { + &self - rhs + } +} + +impl ops::Sub for BigInt { + type Output = BigInt; + #[inline] + fn sub(self, rhs: BigInt) -> Self::Output { + &self - &rhs + } +} + #[cfg(test)] mod tests { use std::u64; - use super::overflowing_add; - use super::BigInt; + use super::{overflowing_add,overflowing_sub,BigInt,Minimum,vec_min}; + + #[test] + fn test_min() { + let b1 = BigInt::new(1); + let b2 = BigInt::new(42); + let b3 = BigInt::from_vec(vec![0, 1]); + + assert_eq!(b1.min(&b2), &b1); + assert_eq!(b2.min(&b1), &b1); + assert_eq!(b3.min(&b2), &b2); + assert_eq!(b2.min(&b3), &b2); + } + + #[test] + fn test_vec_min() { + let b1 = BigInt::new(1); + let b2 = BigInt::new(42); + let b3 = BigInt::from_vec(vec![0, 1]); + + assert_eq!(vec_min(&vec![b2.clone(), b1.clone(), b3.clone()]), Some(&b1)); + assert_eq!(vec_min(&vec![b2.clone(), b3.clone()]), Some(&b2)); + assert_eq!(vec_min(&vec![b3.clone()]), Some(&b3)); + assert_eq!(vec_min::(&vec![]), None); + } #[test] fn test_overflowing_add() { @@ -224,6 +309,57 @@ mod tests { assert_eq!(overflowing_add(1 << 63, (1 << 63) -1 , true), (0, true)); } + #[test] + fn test_overflowing_sub() { + assert_eq!(overflowing_sub(100, 10, false), (90, false)); + assert_eq!(overflowing_sub(100, 10, true), (89, false)); + assert_eq!(overflowing_sub(10, 1 << 63, false), ((1 << 63) + 10, true)); + assert_eq!(overflowing_sub(10, 1 << 63, true), ((1 << 63) + 9, true)); + assert_eq!(overflowing_sub(42, 42 , true), (u64::max_value(), true)); + } + + #[test] + fn test_add() { + let b1 = BigInt::new(1 << 32); + let b2 = BigInt::from_vec(vec![0, 1]); + let b3 = BigInt::from_vec(vec![0, 0, 1]); + let b4 = BigInt::new(1 << 63); + + assert_eq!(&b1 + &b2, BigInt::from_vec(vec![1 << 32, 1])); + assert_eq!(&b2 + &b1, BigInt::from_vec(vec![1 << 32, 1])); + assert_eq!(&b2 + &b3, BigInt::from_vec(vec![0, 1, 1])); + assert_eq!(&b2 + &b3 + &b4 + &b4, BigInt::from_vec(vec![0, 2, 1])); + assert_eq!(&b2 + &b4 + &b3 + &b4, BigInt::from_vec(vec![0, 2, 1])); + assert_eq!(&b4 + &b2 + &b3 + &b4, BigInt::from_vec(vec![0, 2, 1])); + } + + #[test] + fn test_sub() { + let b1 = BigInt::new(1 << 32); + let b2 = BigInt::from_vec(vec![0, 1]); + let b3 = BigInt::from_vec(vec![0, 0, 1]); + let b4 = BigInt::new(1 << 63); + + assert_eq!(&b2 - &b1, BigInt::from_vec(vec![u64::max_value() - (1 << 32) + 1])); + assert_eq!(&b3 - &b2, BigInt::from_vec(vec![0, u64::max_value(), 0])); + assert_eq!(&b2 - &b4 - &b4, BigInt::new(0)); + assert_eq!(&b3 - &b2 - &b4 - &b4, BigInt::from_vec(vec![0, u64::max_value() - 1])); + assert_eq!(&b3 - &b4 - &b2 - &b4, BigInt::from_vec(vec![0, u64::max_value() - 1])); + assert_eq!(&b3 - &b4 - &b4 - &b2, BigInt::from_vec(vec![0, u64::max_value() - 1])); + } + + #[test] + #[should_panic(expected = "Wrapping subtraction of BigInt")] + fn test_sub_panic1() { + let _ = BigInt::new(1) - BigInt::new(5); + } + + #[test] + #[should_panic(expected = "Wrapping subtraction of BigInt")] + fn test_sub_panic2() { + let _ = BigInt::from_vec(vec![5,8,3,33,1<<13,46,1<<49, 1, 583,1<<60,2533]) - BigInt::from_vec(vec![5,8,3,33,1<<13,46,1<<49, 5, 583,1<<60,2533]); + } + #[test] fn test_inc1() { let mut b = BigInt::new(0);