Alter fract() or add new fraction() function - addressing a negative float's decimal

Topic

fract() returns the fractional part of a number. However, what THAT means is a matter of discussion, as a fractional part of a negative number has 3 potential outputs.

1. fraction of -1.3 -> -0.3
2. fraction of -1.3 -> 0.3
3. fraction of -1.3 -> 0.7

In p5.js, fract(-1.3) produces 0.7. This output surprised me!

But this is in keeping with p5strands (which aliases glsl's fract) and WGSL's implementation: x - floor(x).

This all came up became I am trying to have feature parity between L5 creative coding library and p5.js. In Lua, I round toward 0 and remove the integer part, so fract(-1.3) produces 0.3.

Other implementations of fract

• C's modf, works the way i'm describing (rounding toward 0).
• python copies the same
• Rust, i think does the same?

Discussion so far

We are moving discussion from discord. @limzykenneth said that GSLS / shader langs return within the range [0.0, 1.0].

@davepagurek proposed creating a new function name like fractional(n) to disambiguate from fract()'s current output. @limzykenneth said he was leaning toward a signed value output such as e.g. fraction(-1.3) === -0.3.

My own instinct is to separate the decimal part from the sign so that fraction(-1.3) returns -> 0.3.

That's where we are at right now! So we need to decide to do this, what to name it, and which implementation for an added function, if we are commited to creating a new function for different output than fract's current implementation.

References

• Current Implementation of fract()
• Original feature request and discussion of implementing fract()

Start with the current fract() implementation in src/math/calculation.js and review issue #3487 for prior context. Read the linked discussion and references to compare the possible negative-number behaviors. This is complete only after the API name and semantics are agreed, then the selected behavior is implemented and covered by the project’s relevant checks.