Skip to main content

Use `math.isclose` Instead of Direct Equality for Floats


ImportanceReview GuidanceRequires Scanning Tool
LowMerge After ReviewNo

In most programming languages, floating point arithmetic is imprecise due to the way floating point numbers are stored as binary representations. Moreover, the result of calculations with floats can vary based on when rounding happens. Using equality or inequality to compare floats or their operations will almost always be imprecise and lead to bugs.

For these reasons, this codemod changes any operations involving equality or inequality with floats to the recommended math.isclose function. This codemod uses the default parameter values rel_tol=1e-09 and abs_tol=0.0 but makes them explicit as a starting point for you to consider depending on your calculation needs.

Our changes look like the following:

+import math
def foo(a, b):
- return a == b - 0.1
+ return math.isclose(a, b - 0.1, rel_tol=1e-09, abs_tol=0.0)

If you have feedback on this codemod, please let us know!


Why is this codemod marked as Merge After Review?

This change makes your code more accurate but in some cases it may be necessary to adjust the abs_tol and rel_tol parameter values for your particular calculations.

Codemod Settings