[patch] Record the Hexa vs ktsu widget verdict

[matt-edmondson]

Refs #338.

CLAUDE.md said seven widget pairs "both survive on purpose until the comparison tab settles which to keep". The comparison tab shipped in #337; the verdict never got written down. This writes it down, and changes no widget code.

Nothing is obsoleted here. Retiring public API is a breaking change and that call is yours — this records what the decision is actually between, so it can be made on facts rather than on the pair list.

Five of the seven are not duplicates

Pair	Why not
DividerContainer vs HorizontalSplitter/VerticalSplitter	A retained layout container owning a list of DividerZones with per-zone draw callbacks, persisted sizes (GetSizes/SetSizesFromList), an onResized callback and nesting — versus one stateless drag handle mutating a ref float. The comparison tab can't even show them side by side; its splitter row points at the Advanced Demos tab for the container.
Tree vs IconTreeNode	Tree is a ScopedAction drawing connector lines around nested content: no label, no collapsing, no ID-stack interaction. IconTreeNode is a real collapsible node with an icon glyph and flags, and the caller must TreePop(). They compose rather than compete — which is why TreeTests already drives both in one class.
RadialProgressBar/SkeletonLoader vs BufferingBar/Spinner	Four different widgets, not two pairs: determinate radial (plus RadialCountdown/RadialCountUp), determinate linear, indeterminate spinner, shimmering placeholder. The only overlap is "determinate progress", where radial vs linear is a visual choice.
TextCentered vs TextCenteredV/H/VH	Alignment.Center builds a container of (ContentRegionAvail.X, contentSize.Y) — its vertical extent is the content's own height, so vertical centring is a no-op and TextCentered is horizontal-only. It equals TextCenteredH and nothing else. TextCenteredV/VH have no ktsu counterpart; TextCenteredWithin(text, size, clip) with ellipsis clipping has no Hexa counterpart.
ImageCentered vs ImageCenteredV/H/VH	Same shape, plus a hard difference the demo already comments on: ImageCentered returns bool for a click, the Hexa three return void. ImageCenteredWithin and ImageScaleTo are each unique to their side.

Two are genuine

Switch vs ToggleSwitch — identical signature. Three differences, all favouring Switch: it calls ImGuiProbes.MarkItem; it animates from ImGui.GetIO().DeltaTime rather than Hexa's clock, so it needs none of the TickAnimationClockIfUnpumped machinery ToggleSwitch requires to avoid rendering inverted after its first click in a pumpless app; and it draws the visible label itself.

Combo vs EnumCombo — only Combo<TEnum> overlaps; the ISemanticString and string collection overloads are unique. The enum overloads differ in naming (raw Enum.GetNames vs Hexa's ComboEnumHelper<T>.GetName, which is where a display-name override would come from), probe marking, and constraint (where TEnum : Enum vs where T : struct, Enum). If nothing depends on Hexa's naming, EnumCombo is the one to retire.

The through-line

Only the ktsu originals are probe-visible. Switch and Combo are the only two widgets in either family that call ImGuiProbes.MarkItem — SwitchTests and ComboTests need zero Mark calls, while every Hexa-backed suite marks the widget itself or reaches it by geometry. That is the most consequential difference between the two families and a reason to prefer a ktsu original in an app you intend to UI-test.

Changes

• CLAUDE.md: replaces the "both survive on purpose" sentence with a link to a new Hexa-backed vs ktsu widgets section carrying the table above.
• examples/ImGuiWidgetsDemo/HexaWidgetsDemo.cs: one line at the top of the comparison tab noting that most rows are not duplicates, pointing at that section.

Verification

dotnet build clean, 0 warnings.

ImGuiWidgetsDemo.UITests reports 12 of 28 failing both with and without this change — identical counts, so nothing here moved them. The root cause is environmental rather than a defect: examples/ImGuiWidgetsDemo/ktsu.png is a Git LFS pointer, git-lfs is not installed in this sandbox, so it was never smudged and ImageSharp fails to decode it (UnknownImageFormatException, with PNG absent from the decoder list). Every failing test is one that reaches a tab loading that texture. CI has git-lfs, so this should not reproduce there.

---

Generated by Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud