Square Root Without Built-ins | Dataford Interview Questions

Your question is Square Root Without Built-ins. Start with the requirements on the right.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Freddie Mac analytics services may need a predictable square-root calculation without relying on language-provided math functions. Implement an algorithm that computes the square root of a nonnegative number using binary search.

Formal Specification

Write sqrt_approx(x, epsilon):

x is a nonnegative integer or floating-point number.
epsilon is a positive floating-point tolerance.
Return a floating-point value r such that abs(r * r - x) <= epsilon.
Do not use built-in square-root, exponentiation, logarithm, or power functions. Basic arithmetic and comparisons are allowed.
For x = 0, return 0.0.

The evaluator accepts answers within the required squared-error tolerance. Aim for logarithmic convergence rather than scanning candidate values incrementally.

Constraints

0 <= x <= 10^12
10^-12 <= epsilon <= 1
x is an integer or floating-point number
epsilon is positive
Do not use square-root, exponentiation, logarithm, or power functions

Function Signature

def sqrt_approx(x, epsilon):