Floating Point Imprecision

[Larry_Tweed1]

Hello. I was aware of the issue with floating point imprecision prior to this lesson, but I’m glad Sam covered it. However, we’ve used tests for a float == 0 before, without mention. Specifically, in the Health.cs file, we have…

        public void TakeDamage(GameObject instigator, float damage)
        {

            healthPoints.value = Mathf.Max(healthPoints.value - damage, 0);
            if (healthPoints.value == 0)
            {

                Die();
                AwardExperience(instigator);
            }
            else
            {
                takeDamage.Invoke(damage);
            }

        }

healthPoints is a LazyValue, but it was previously a normal float. Why is it safe to check the float healthPoints or healthPoints.value for equality with 0?

[bixarrio]

It may be because we will set it to 0. When we take damage, we subtract the damage from the health. eventually it will go below 0 and we will then set it to exactly 0

[Larry_Tweed1]

So if you set it to 0, you can safely test for equality with 0?

[Brian_Trotter]

If you explicitly set it 0, you can test for zero, yes. If it’s the result of a floating point operation (for example: 10f -5f -1f-4f; you should not trustthe results. To be honest, best practice is to always use Mathf.Approximately. With some code editors, healthPoints.value==0 will be flagged with a warning about floating point comparison.

In my personal Health scripts, I always use ints for the values (meaning I have to convert BaseStats.GetStat(Stat.Health) to an int with Mathf.FloorToInt. I still prefer the stats and stat modifiers to be floats until they get into health, where they are converted.

[Larry_Tweed1]

Thanks Brian. Yeah, I prefer health to be integers also (can’t really think of an occasional where I would need decimals), but I also like to keep my code in sync with Sam’s while I’m taking the course. I’ll probably revisit once I finish off this course.