web: add Implant/Other layer categories and per-layer fill patterns

[jorge-ferreira-pii]

Addresses a portion of #10619

[gemini-code-assist]

Code Review

This pull request introduces support for per-layer fill patterns (such as diagonal, cross, and dots) in the web frontend, mirroring the Qt GUI's rendering behavior. It updates both the JavaScript frontend to manage and persist these pattern choices via cookies and the C++ tile generator backend to parse and apply the patterns when rendering polygons and rectangles. The review feedback focuses on performance and code quality improvements, specifically suggesting to refactor a local lambda into a free function, optimize a modulo operation, and add early returns in the rendering functions when the fill pattern is set to none.

[gemini-code-assist]

Helper logic should be defined as a free function in a namespace rather than as a local lambda within a function. Additionally, since period is always positive, we can optimize the modulo operation to avoid a second modulo by using a simple conditional check.

References

1. Helper logic should be defined as a free function in a namespace rather than as a local lambda within a function.

[gemini-code-assist]

If the fill pattern is set to FillPattern::kNone, we can return early immediately. This avoids performing any polygon scanline setup, edge intersections, sorting, or pixel loops entirely, which is a significant performance win.

Suggested change

	{
	// Tile origin in absolute pixel space, anchoring non-solid patterns.
	{
	if (pattern == FillPattern::kNone) {
	return;
	}
	// Tile origin in absolute pixel space, anchoring non-solid patterns.

[gemini-code-assist]

Similarly, if the fill pattern is set to FillPattern::kNone, we can return early from drawFilledRect immediately to avoid any loop overhead.

{
  if (pattern == FillPattern::kNone) {
    return;
  }
  for (int iy = rect.yMin(); iy < rect.yMax(); ++iy) {

[jorge-ferreira-pii]

2.2 Display controls — layers ✅

Feature	GUI	Web	Status
Per-layer visibility (tree)	displayControls	display-controls.js	✅
Per-layer color	color picker	color.h	✅
Selectability column	yes	yes	✅
Per-layer fill pattern (solid/hatch/crosshatch)	yes	FillPattern + patternCovers	❌ → ✅
Implant / Other layer categories	yes	classifyLayer() grouping	🟡 → ✅

The whole 2.2 — Display controls — layers area is now ✅.

[jorge-ferreira-pii]

What was implemented and tested

• Per-layer fill pattern (❌ → ✅) — solid/diagonal/cross/dots/none per layer, chosen from the layer-tree right-click menu and persisted in the or_layer_patterns cookie.

• Implant / Other layer categories (🟡 → ✅) :

asap7/gcd — exercises the Implant category: RVTN/RVTP render real geometry.
sky130hd/gcd — exercises the Other category: nwell/pwell render real geometry.

[gadfort]

I would suggest leaving the Implant and Other categories collapsed when opening.
Also I found it strange that the layer selection was saved between sessions (so when I opened a new design in a new session the fact that only one metal was visible was a little jarring (this behavior differs from the QtGUI where the layer options are not saved between sessions).

[jorge-ferreira-pii]

I would suggest leaving the Implant and Other categories collapsed when opening. Also I found it strange that the layer selection was saved between sessions (so when I opened a new design in a new session the fact that only one metal was visible was a little jarring (this behavior differs from the QtGUI where the layer options are not saved between sessions).

Thanks, Peter! Good catch, those details totally flew under my radar. I've just pushed the fixes for the category collapsing and the session saving, and I also fixed the merge conflict.