[p5.js 2.0+ Bug Report]: Floats in strands shaders are always rounded to 4 decimals

[davepagurek]

Most appropriate sub-area of p5.js?

• Accessibility
• Color
• Core/Environment/Rendering
• Data
• DOM
• Events
• Image
• IO
• Math
• Typography
• Utilities
• WebGL
• WebGPU
• p5.strands
• Build process
• Unit testing
• Internationalization
• Friendly errors
• Other (specify if possible)

p5.js version

2.3.0

Web browser and version

All

Operating system

All

Steps to reproduce this

In strands_glslBackend.js, we do this when converting number literals into shader code:

p5.js/src/webgl/strands_glslBackend.js

Lines 273 to 274 in a5c9056

	if (node.baseType === BaseType.FLOAT) {
	return node.value.toFixed(4);

This means that if you include 0.00001 in your strands code, your shader code would be 0.0000. That seems like a problem. Is there any way we can do this in a more context-dependent way where we don't just add more decimals across the board for everything?

[harshiltewari2004]

happy to take this up @davepagurek !

[davepagurek]

Before actually writing code, do you have an idea for how to tackle this one?

[harshiltewari2004]

Took a look — the root cause is that float literals are serialized with node.value.toFixed(4) in both strands backends: strands_glslBackend.js (WebGL) and strands_wgslBackend.js (WebGPU), same line in each. So anything smaller than ~0.0001 collapses to 0.0000, and normal values get padded with trailing zeros.
My thinking for a fix: instead of fixed-precision rounding, serialize the number in a way that preserves its actual value, then make sure the result still reads as a float literal (has a decimal point, or .0 for integer-valued floats). That avoids both the truncation and the padding, and doesn't bake in a fixed decimal count.
The detail I'd want to get right is literal formatting per backend — e.g. whether exponent notation (1e-8) is acceptable in the GLSL version strands targets and in WGSL, or whether those need to be expanded. I'll confirm that against each backend rather than assume.
Does that direction sound right to you, and do you want the fix to cover both backends in one PR, or keep them separate?

[davepagurek]

Right, I think the question is around how to do that serialization. I think if we can keep the method of serialization the same for each backend then that's ideal, less complexity to worry about

[harshiltewari2004]

Sounds good — keeping the serialization identical across both backends makes sense, I'll go that route. I'd like to take this one on if that works;I've got #8823 in review right now, so I'll start on this once that lands. happy to be assigned.