float-compare: questions with best practices

[mmorel-35]

In https://github.com/argoproj/argo-cd/pull/18721/files I applyied float-compare rules .
I have two questions here.
With this first case we have situation

assert.Equal(t, float64(1), retryCount)

This suggestion

assert.InEpsilon(t, float64(1), retryCount, 0.0001)

Having to convert an integer to a float to compare with a float then have to be used with assert.InDelta or assert.InEpsilon.

What do you think about using

assert.EqualValue(t, 1, retryCount)

Maybe this situation indicates a mistyping at the origin so the solution is to change the type of retryCount as a counter is usually an integer.

The other question is about

assert.Equal(t, liveReplicas, desiredReplicas)

I was suggested to use assert.InEpsilon but I ended up using

assert.Zero(t, desiredReplicas)
assert.Zero(t, liveReplicas)

The reason behind this is that assert.InEpsilon doesn’t support having 0 as an expected value. The tricky thing in this situation is that desiredReplicas might not be known in certain cases.

[Antonboom]

Maybe this situation indicates a mistyping at the origin so the solution is to change the type of retryCount as a counter is usually an integer.

Nice point, looks like a feature 👍

The reason behind this is that assert.InEpsilon doesn’t support having 0 as an expected value.

Why not assert.InDelta(t, 0., 0., 0.)?

[mmorel-35]

I don't see why InDelta would support 0 as an expected value while InEpsilon doesn't. It didn't occured to me to even try it .

[Antonboom]

@mmorel-35, I want to transfer this issue into discussion.
But looks like I have no right for it, wtf? )

[ccoVeille]

I also have a question with float compared to 0

Something was raised I a recent review when @mmorel-35 used testifylint + manual edit to fix a code using somehow

var f float
assert.Equal(t, 0, f)

[ccoVeille]

Here exactly

prometheus/prometheus#14334 (comment)

[ccoVeille]

I'm wondering why using Zero(the, f) would be valid compared to Equal(t, 0, f)

And why in epsilon or in delta won't accept 0, @mmorel-35 point here.

I mean either in don't get something.

Or we have something to detect with zero used with float

And if you are wondering @Antonboom, I didn't look at the code yet.

I'm raising the point as a note for the discussion, either for myself and for you guys