Add float oriented Assert.Equal with precision

[mwpowellhtx]

I am halfway surprised given its maturity that there is not a float oriented Assert.Equal method. There is a double oriented method, there is even a decimal oriented method. I should think that comparison is somewhere along the lines of the double methods.

[bradwilson]

What's the value of adding a float version as opposed to just using the existing double version?

[mwpowellhtx]

@bradwilson Are we guaranteed there would not be precision errors converting float to double?

[mwpowellhtx]

i.e. A Duck search, and at least one SO article.

[bradwilson]

That talked about formatting issues and the imprecision associated with floating points in general, not specifically about any kind of loss of precision.

[mwpowellhtx]

Or in this article, Convert float to double. Along similar lines.

[bradwilson]

I looked at doing this on today's live stream, and since we use Math.Round, it's not as simple as it appears. There are only double versions of Math.Round and not float versions, so it's not clear to me that we're not back in the same problem, only pushing the cast from float to double down a layer into our call to Math.Round.

Interested to hear alternate suggestions, so tagging this as a discussion item for now.

[mwpowellhtx]

Would it make sense to cast float to/from decimal for the Math.Round?

If not, it's cool, appreciate you at least looking into it further. It is not urgent for me at the moment, so if nothing else put it on the shelf in the event that a plausible solution pops up for further consideration.

[Romanx]

Was just catching up on the VOD on twitch and saw this issue. Another option that may be more "correct" would be to use MathF for the round functionality although i'm not sure if it's available for your targets since its netstandard2.1.

[bradwilson]

Would it make sense to cast float to/from decimal for the Math.Round?

No, because the range of valid decimal values is smaller than the range of valid float values.

Was just catching up on the VOD on twitch and saw this issue. Another option that may be more "correct" would be to use MathF for the round functionality although i'm not sure if it's available for your targets since its netstandard2.1.

Correct, we cannot use MathF.

[bradwilson]

Based on the discussion in #1293 I've added an item in the v3 roadmap to enable tolerance-based Assert.Equal for double (and deprecate the precision-based version). This should allow me to add a float based assertion as well, since our reliance on Math.Round will disappear.