""" Calculates the Fibonacci sequence using iteration, recursion, memoization, and a simplified form of Binet's formula NOTE 1: the iterative, recursive, memoization functions are more accurate than the Binet's formula function because the Binet formula function uses floats NOTE 2: the Binet's formula function is much more limited in the size of inputs that it can handle due to the size limitations of Python floats NOTE 3: the matrix function is the fastest and most memory efficient for large n See benchmark numbers in __main__ for performance comparisons/ https://en.wikipedia.org/wiki/Fibonacci_number for more information """ import functools from collections.abc import Iterator from math import sqrt from time import time import numpy as np from numpy import ndarray def time_func(func, *args, **kwargs): """ Times the execution of a function with parameters """ start = time() output = func(*args, **kwargs) end = time() if int(end - start) > 0: print(f"{func.__name__} runtime: {(end - start):0.4f} s") else: print(f"{func.__name__} runtime: {(end - start) * 1000:0.4f} ms") return output def fib_iterative_yield(n: int) -> Iterator[int]: """ Calculates the first n (1-indexed) Fibonacci numbers using iteration with yield >>> list(fib_iterative_yield(0)) [0] >>> tuple(fib_iterative_yield(1)) (0, 1) >>> tuple(fib_iterative_yield(5)) (0, 1, 1, 2, 3, 5) >>> tuple(fib_iterative_yield(10)) (0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55) >>> tuple(fib_iterative_yield(-1)) Traceback (most recent call last): ... ValueError: n is negative """ if n < 0: raise ValueError("n is negative") a, b = 0, 1 yield a for _ in range(n): yield b a, b = b, a + b def fib_iterative(n: int) -> list[int]: """ Calculates the first n (0-indexed) Fibonacci numbers using iteration >>> fib_iterative(0) [0] >>> fib_iterative(1) [0, 1] >>> fib_iterative(5) [0, 1, 1, 2, 3, 5] >>> fib_iterative(10) [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55] >>> fib_iterative(-1) Traceback (most recent call last): ... ValueError: n is negative """ if n < 0: raise ValueError("n is negative") if n == 0: return [0] fib = [0, 1] for _ in range(n - 1): fib.append(fib[-1] + fib[-2]) return fib def fib_recursive(n: int) -> list[int]: """ Calculates the first n (0-indexed) Fibonacci numbers using recursion >>> fib_iterative(0) [0] >>> fib_iterative(1) [0, 1] >>> fib_iterative(5) [0, 1, 1, 2, 3, 5] >>> fib_iterative(10) [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55] >>> fib_iterative(-1) Traceback (most recent call last): ... ValueError: n is negative """ def fib_recursive_term(i: int) -> int: """ Calculates the i-th (0-indexed) Fibonacci number using recursion >>> fib_recursive_term(0) 0 >>> fib_recursive_term(1) 1 >>> fib_recursive_term(5) 5 >>> fib_recursive_term(10) 55 >>> fib_recursive_term(-1) Traceback (most recent call last): ... Exception: n is negative """ if i < 0: raise ValueError("n is negative") if i < 2: return i return fib_recursive_term(i - 1) + fib_recursive_term(i - 2) if n < 0: raise ValueError("n is negative") return [fib_recursive_term(i) for i in range(n + 1)] def fib_recursive_cached(n: int) -> list[int]: """ Calculates the first n (0-indexed) Fibonacci numbers using recursion >>> fib_iterative(0) [0] >>> fib_iterative(1) [0, 1] >>> fib_iterative(5) [0, 1, 1, 2, 3, 5] >>> fib_iterative(10) [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55] >>> fib_iterative(-1) Traceback (most recent call last): ... ValueError: n is negative """ @functools.cache def fib_recursive_term(i: int) -> int: """ Calculates the i-th (0-indexed) Fibonacci number using recursion """ if i < 0: raise ValueError("n is negative") if i < 2: return i return fib_recursive_term(i - 1) + fib_recursive_term(i - 2) if n < 0: raise ValueError("n is negative") return [fib_recursive_term(i) for i in range(n + 1)] def fib_memoization(n: int) -> list[int]: """ Calculates the first n (0-indexed) Fibonacci numbers using memoization >>> fib_memoization(0) [0] >>> fib_memoization(1) [0, 1] >>> fib_memoization(5) [0, 1, 1, 2, 3, 5] >>> fib_memoization(10) [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55] >>> fib_iterative(-1) Traceback (most recent call last): ... ValueError: n is negative """ if n < 0: raise ValueError("n is negative") # Cache must be outside recursuive function # other it will reset every time it calls itself. cache: dict[int, int] = {0: 0, 1: 1, 2: 1} # Prefilled cache def rec_fn_memoized(num: int) -> int: if num in cache: return cache[num] value = rec_fn_memoized(num - 1) + rec_fn_memoized(num - 2) cache[num] = value return value return [rec_fn_memoized(i) for i in range(n + 1)] def fib_binet(n: int) -> list[int]: """ Calculates the first n (0-indexed) Fibonacci numbers using a simplified form of Binet's formula: https://en.m.wikipedia.org/wiki/Fibonacci_number#Computation_by_rounding NOTE 1: this function diverges from fib_iterative at around n = 71, likely due to compounding floating-point arithmetic errors NOTE 2: this function doesn't accept n >= 1475 because it overflows thereafter due to the size limitations of Python floats >>> fib_binet(0) [0] >>> fib_binet(1) [0, 1] >>> fib_binet(5) [0, 1, 1, 2, 3, 5] >>> fib_binet(10) [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55] >>> fib_binet(-1) Traceback (most recent call last): ... ValueError: n is negative >>> fib_binet(1475) Traceback (most recent call last): ... ValueError: n is too large """ if n < 0: raise ValueError("n is negative") if n >= 1475: raise ValueError("n is too large") sqrt_5 = sqrt(5) phi = (1 + sqrt_5) / 2 return [round(phi**i / sqrt_5) for i in range(n + 1)] def matrix_pow_np(m: ndarray, power: int) -> ndarray: """ Raises a matrix to the power of 'power' using binary exponentiation. Args: m: Matrix as a numpy array. power: The power to which the matrix is to be raised. Returns: The matrix raised to the power. Raises: ValueError: If power is negative. >>> m = np.array([[1, 1], [1, 0]], dtype=int) >>> matrix_pow_np(m, 0) # Identity matrix when raised to the power of 0 array([[1, 0], [0, 1]]) >>> matrix_pow_np(m, 1) # Same matrix when raised to the power of 1 array([[1, 1], [1, 0]]) >>> matrix_pow_np(m, 5) array([[8, 5], [5, 3]]) >>> matrix_pow_np(m, -1) Traceback (most recent call last): ... ValueError: power is negative """ result = np.array([[1, 0], [0, 1]], dtype=int) # Identity Matrix base = m if power < 0: # Negative power is not allowed raise ValueError("power is negative") while power: if power % 2 == 1: result = np.dot(result, base) base = np.dot(base, base) power //= 2 return result def fib_matrix_np(n: int) -> int: """ Calculates the n-th Fibonacci number using matrix exponentiation. https://www.nayuki.io/page/fast-fibonacci-algorithms#:~:text= Summary:%20The%20two%20fast%20Fibonacci%20algorithms%20are%20matrix Args: n: Fibonacci sequence index Returns: The n-th Fibonacci number. Raises: ValueError: If n is negative. >>> fib_matrix_np(0) 0 >>> fib_matrix_np(1) 1 >>> fib_matrix_np(5) 5 >>> fib_matrix_np(10) 55 >>> fib_matrix_np(-1) Traceback (most recent call last): ... ValueError: n is negative """ if n < 0: raise ValueError("n is negative") if n == 0: return 0 m = np.array([[1, 1], [1, 0]], dtype=int) result = matrix_pow_np(m, n - 1) return int(result[0, 0]) if __name__ == "__main__": from doctest import testmod testmod() # Time on an M1 MacBook Pro -- Fastest to slowest num = 30 time_func(fib_iterative_yield, num) # 0.0012 ms time_func(fib_iterative, num) # 0.0031 ms time_func(fib_binet, num) # 0.0062 ms time_func(fib_memoization, num) # 0.0100 ms time_func(fib_recursive_cached, num) # 0.0153 ms time_func(fib_recursive, num) # 257.0910 ms time_func(fib_matrix_np, num) # 0.0000 ms