Goal
Plan (and eventually execute) migrating the examples from p5.js 1.11.13 / p5.sound 1.11.13 / ml5.js 0.12.2 to p5.js 2.x + the split-out p5.sound + ml5.js 1.x.
This is a tracking/planning issue, not scheduled — timing depends on when p5.js 2.0 is adopted in teaching, which is still TBD. Nothing is broken today: every example is pinned to known-good versions, so there is no urgency. The point is to capture scope now so the switch is a deliberate, low-surprise change later.
Current state
- CDN versions are pinned (no
@latest): p5.js & p5.sound 1.11.13, ml5.js 0.12.2. See CLAUDE.md and scripts/audit_libraries.py.
- Zero-toolchain by design (open
index.html; no build). Any migration must preserve that.
Breaking changes that affect this repo
p5.js 2.0
preload() is removed. Examples that load assets in preload() (images, fonts, sounds, mp3s) must move to async loading. Affects e.g. Sound/FrequencyBubbles2-Image, Sound/SoundLevelBubbles2-Image, Sound/SoundVis3/SoundVis4 (mp3), several Games/* sprite/image loaders, Art/* using images/fonts.- p5.sound is split out into a separate library — its script include changes, and some APIs differ.
getMasterVolume() is gone (already removed from the 3 SoundVis debug print()s in 1cc55d8; double-check none remain).- General 2.0 API/behavior deltas to audit (e.g.
loadImage/loadSound return values, color/typography tweaks).
ml5.js 1.x
poseNet(...).on() / handpose(...).on() callback APIs were replaced by bodyPose / handPose + detectStart(...).- Affected examples:
ml5js/PoseNet/* (ElmoGenerator, ElmoGeneratorMultiperson, Skeleton), ml5js/HandPose/*, Games/FlappyArm, WebSerial/ml5js/* (HandWaveDetector, HandWaver, NoseTracker), Node/RemoteHandWave.
Games/SnakeSpeech has a pre-existing runtime error (its ml5 sound-classifier callback does results[0] when the classifier errors headless) — fold the fix into this migration.
Suggested approach (when we decide to switch)
- Decide the trigger: pin to a specific p5.js 2.x patch release once teaching adopts it.
- Migrate in small, verified, committed chunks per category (same workflow used for the accessibility pass): branch off
main, verify each example in-browser, fast-forward merge.
- Update the version pins +
scripts/audit_libraries.py / fix_libraries.py defaults and the _libraries/ offline copies together.
- Start with the
preload() → async conversions (mechanical, broad), then p5.sound include/API, then the ml5 0.12→1.x rewrite (largest, most behavioral).
- Keep the zero-toolchain constraint: CDN includes only, no npm/build.
Open questions
- Pin to which exact p5.js 2.x release?
- Do we want a brief teaching note in affected examples explaining the 1.x→2.x difference (since students may have older references)?
- Any examples worth retiring instead of migrating?
Out of scope
Accessibility (done: canvas describe(), control labels, audio/score captions) and the minor cleanups (duplicated Color CSS/font, implicit globals, per-frame print() debug spam, sparse comments) are tracked separately.
Goal
Plan (and eventually execute) migrating the examples from p5.js 1.11.13 / p5.sound 1.11.13 / ml5.js 0.12.2 to p5.js 2.x + the split-out p5.sound + ml5.js 1.x.
This is a tracking/planning issue, not scheduled — timing depends on when p5.js 2.0 is adopted in teaching, which is still TBD. Nothing is broken today: every example is pinned to known-good versions, so there is no urgency. The point is to capture scope now so the switch is a deliberate, low-surprise change later.
Current state
@latest): p5.js & p5.sound1.11.13, ml5.js0.12.2. SeeCLAUDE.mdandscripts/audit_libraries.py.index.html; no build). Any migration must preserve that.Breaking changes that affect this repo
p5.js 2.0
preload()is removed. Examples that load assets inpreload()(images, fonts, sounds, mp3s) must move to async loading. Affects e.g.Sound/FrequencyBubbles2-Image,Sound/SoundLevelBubbles2-Image,Sound/SoundVis3/SoundVis4(mp3), severalGames/*sprite/image loaders,Art/*using images/fonts.getMasterVolume()is gone (already removed from the 3 SoundVis debugprint()s in1cc55d8; double-check none remain).loadImage/loadSoundreturn values, color/typography tweaks).ml5.js 1.x
poseNet(...).on()/handpose(...).on()callback APIs were replaced bybodyPose/handPose+detectStart(...).ml5js/PoseNet/*(ElmoGenerator, ElmoGeneratorMultiperson, Skeleton),ml5js/HandPose/*,Games/FlappyArm,WebSerial/ml5js/*(HandWaveDetector, HandWaver, NoseTracker),Node/RemoteHandWave.Games/SnakeSpeechhas a pre-existing runtime error (its ml5 sound-classifier callback doesresults[0]when the classifier errors headless) — fold the fix into this migration.Suggested approach (when we decide to switch)
main, verify each example in-browser, fast-forward merge.scripts/audit_libraries.py/fix_libraries.pydefaults and the_libraries/offline copies together.preload()→ async conversions (mechanical, broad), then p5.sound include/API, then the ml5 0.12→1.x rewrite (largest, most behavioral).Open questions
Out of scope
Accessibility (done: canvas
describe(), control labels, audio/score captions) and the minor cleanups (duplicated Color CSS/font, implicit globals, per-frameprint()debug spam, sparse comments) are tracked separately.