serumalt/Source/EffectUnits/AGENTS.md

effects module

Owned files

• Related rack implementation: ../FXProcessor.h, ../FXProcessor.cpp
• Biquad.h
• Hyper.h, Hyper.cpp
• Chorus.h, Chorus.cpp
• Flanger.h, Flanger.cpp
• Phaser.h, Phaser.cpp
• Distortion.h, Distortion.cpp
• EQ.h, EQ.cpp
• Compressor.h, Compressor.cpp
• Delay.h, Delay.cpp
• Reverb.h, Reverb.cpp

Rules

• Implement the FXUnit interface with prepare, reset and process (float* l, float* r, int numSamples, const float p[4]).
• Process full-wet in place; FXProcessor handles the dry/wet mix.
• Document what p[0] through p[3] mean for each unit in the header comment.
• Use std::vector<float> for delay lines and size them in prepare().
• Use juce::MathConstants<T>::pi and twoPi in delay LFOs, not raw literals.
• Never allocate or resize a delay line inside process.
• Guard empty delay lines with an early if (len < 4) return;.

IF-THEN

• IF a unit needs a delay line THEN allocate it in prepare() at the full required length and only read and write it in process.
• Include the existing rack interface via ../FXProcessor.h; no source relocation is implied by this guidance.

Examples

// BAD: reallocates a delay line on the audio thread
void process (float* l, float* r, int n, const float p[4]) override
{
    std::vector<float> delay ((size_t) (sr * 0.05), 0.0f);
    ...
}

// GOOD: sized once at prepare time, reused per block
void prepare (double sampleRate, int) override
{
    sr = sampleRate;
    delayL.assign ((size_t) (sr * 0.05), 0.0f);
    delayR.assign ((size_t) (sr * 0.05), 0.0f);
}

This file overrides /AGENTS.md where they conflict.