vireuikit 0.2.0 → 0.4.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -2,8 +2,8 @@
2
2
 
3
3
  A UI kit built on top of [VireGlass](https://github.com/pruffit/vireglass)'s physically-derived
4
4
  glass material: atom state and press/hover/selection behavior for controls, a capsule-to-target
5
- growth morph, and a live GPU backdrop — colored vapor, condensation and light tracks — that plugs
6
- into VireGlass's WebGL2 renderer as a backdrop pass.
5
+ growth morph, and a live GPU backdrop — a lit cloud-chamber: dark mist, static lights, and
6
+ condensation tracks — that plugs into VireGlass's WebGL2 renderer as a backdrop pass.
7
7
 
8
8
  **Status: 0.x.** The API can still change between minor versions.
9
9
 
@@ -44,11 +44,30 @@ const frame = growthFrame(sourceShape, targetShape, anchorPoint, growthPhase(sta
44
44
  // frame.body: the shape to draw as glass this frame.
45
45
  ```
46
46
 
47
- ## GPU backdrop: colored vapor and condensation
47
+ ## GPU backdrop: a lit chamber, not colored smoke
48
48
 
49
49
  `createMediumBackdrop()` returns a `VireGlassBackdropPass` factory: a self-contained fluid
50
- simulation (curl-noise vapor, condensing droplets, decaying light tracks) that composites straight
51
- into the texture VireGlass's lens samples — no offscreen canvas, no readback.
50
+ simulation (curl-noise vapor, condensing droplets, decaying condensation tracks) that composites
51
+ straight into the texture VireGlass's lens samples — no offscreen canvas, no readback.
52
+
53
+ The chamber itself is dark; what's visible is light scattered by vapor, droplets and tracks, not
54
+ the gas's own color. `lights` (`VireUIKitMediumParams`, up to `MEDIUM_MAX_LIGHTS`) are static
55
+ sources — position, direction, cone half-angle, falloff, color, intensity. The defaults are an edge
56
+ strip along the chamber floor and two spotlights placed above the frame, aimed down across it. A
57
+ spotlight has no drawn edge: its beam is a soft lobe with static shafts fanning out from the lamp,
58
+ dimmed by the gas between the lamp and each point (a grid-resolution shadow pass,
59
+ `shadowExtinction`), so beams show where there is vapor to scatter them. `lightRigForCover`/
60
+ `lightRigForCharacter` tint the spotlights toward a cover's hue like colored gels and leave the
61
+ strip near neutral. Vapor species (`channelColors`) fold into a near-neutral ambient base;
62
+ `channelScatter` is how strongly each species scatters light.
63
+
64
+ Tracks come in three kinds (`MEDIUM_TRACK_PRESETS`): `alpha` (short, thick, thickening toward its
65
+ end — the playing source sprays a burst of them from the cover's position, a few on the beat and
66
+ the rest over the next 1.6 s), `electron` (thin, wiggly) and `muon` (long, straight,
67
+ thin). The calm state schedules a lone electron or muon every 8–20 s from no source at all. Each track is drawn twice: as a soft residue stamped into the simulation grid, and
68
+ as a crisp droplet chain in content pixels (`trackLayerAmount`, `MEDIUM_TRACK_LAYER_LOOK`) that
69
+ appears at full length, then broadens, sags and fades over one to two seconds. Tracks are lit by
70
+ the same lights, plus `trackLightFloor` so one is never fully dark.
52
71
 
53
72
  ```ts
54
73
  import { createMediumBackdrop } from 'vireuikit/web';
@@ -80,16 +99,29 @@ The backdrop reads as a chamber seen from the side, not a flat field, without a
80
99
  the composite samples the same vapor/condensate grid twice, once at identity and once at a larger
81
100
  scale and a fixed offset (`MEDIUM_DEFAULTS.depthFarScale`/`depthFarOffsetFrac`, the offset as a
82
101
  fraction of the grid so it stays clear of the wrap at any grid size) — the far read is both slower
83
- on screen and finer-grained, since the field's own motion maps to screen motion as `v / scale`. A
84
- cheap 4-tap blur on that far read (`depthFarBlurRadius`) supplies aerial perspective's softness and
85
- lower contrast in the same pass, weighted down (`depthFarWeight`) so it never competes with the near
86
- plane. Gravity is two separate properties: a small downward drift on vapor's own velocity
87
- (`gravityVaporDrift`, a motion cue only — a uniform drift on this periodic grid cannot itself
88
- accumulate density) and a compositing-only density boost near the bottom of the frame
89
- (`gravityBottomBoost`/`gravityBottomBoostStart`, a steady-state property that cannot affect the
90
- conserved water total) that actually produces the "denser at the bottom" reading.
91
- Each emitted track also carries a `depth` (0 far … 1 near) that scales its own width and intensity,
92
- so a track reads thinner, dimmer and softer the farther back it lives.
102
+ on screen and finer-grained, since the field's own motion maps to screen motion as `v / scale`. An
103
+ 8-tap rotated ring blur on that far read (`depthFarBlurRadius`) supplies aerial perspective's
104
+ softness and lower contrast in the same pass, weighted down (`depthFarWeight`) so it never competes
105
+ with the near plane; the ring (not an axis-aligned box) keeps that blur from turning a residual
106
+ seam or resize artifact into parallel straight lines. Gravity is two separate properties: a small
107
+ downward drift on vapor's own velocity (`gravityVaporDrift`, a motion cue only — a uniform drift on
108
+ this periodic grid cannot itself accumulate density) and a compositing-only density boost near the
109
+ bottom of the frame (`gravityBottomBoost`/`gravityBottomBoostStart`, a steady-state property that
110
+ cannot affect the conserved water total) that actually produces the "denser at the bottom" reading.
111
+ Condensate settling adds a second, always-on share of the ambient curl field's own sideways push
112
+ (`condensateSettleWiggle`) so a droplet falling at a constant speed doesn't trace a dead-straight
113
+ vertical line when turbulence is otherwise low. Each emitted track also carries a `depth` (0 far …
114
+ 1 near) that scales its own width and intensity, so a far track reads thinner, dimmer and finer,
115
+ while the nearest tracks read a little LARGER than the base preset, not merely full-size — a
116
+ Gaussian stamp's edge softness scales with its own width, so the same knob gives depth of field too.
117
+
118
+ The curl-noise potential itself is periodic over the simulation grid in space, the same way it
119
+ already was in time: each octave's spatial frequency is adjusted, per grid size, to the nearest
120
+ value that makes a full grid width (or height) an exact integer number of lattice cells
121
+ (`computeMediumSpatialPeriods`) — so velocity never jumps at the domain's own wrap, and the far
122
+ plane's tile boundary reads as an ordinary point in the field rather than a seam. The near plane's
123
+ vapor and condensate reads use cubic B-spline reconstruction (4 bilinear taps) instead of plain
124
+ bilinear, so the simulation grid's own cells don't read as squares at typical (coarse) grid sizes.
93
125
 
94
126
  ## License
95
127