Fix TabView collapsing to zero height after model reset

[juliuspfadt]

Problem

TabViews with dynamic values/source intermittently collapse to zero height (tab bar visible, content folded to nothing) and stay collapsed. Most visible in jaspSem's Moderated Nonlinear Factor Analysis, which nests TabViews 3 levels deep with values: recomputed from checkbox states and factor lists.

Root cause

itemStack (the StackLayout holding tab content) got its height via:

onCurrentIndexChanged: height = Qt.binding( function() { return currentIndex >= 0 ? rep.itemAt(currentIndex).height : 0; });

Two defects:

1. The binding is only installed once currentIndex changes. If it never does (single tab, or index staying at 0), no height binding exists at all.
2. Repeater.itemAt() is a plain function — not notifiable. Any values/source change triggers a full model reset (ListModel::_initTerms → beginResetModel()), destroying and recreating all delegates. If currentIndex stays the same, the binding's only tracked dependencies are currentIndex and the destroyed wrapper's height — it never re-evaluates, keeps referencing the dead item, and the TabView collapses to 0. The collapse latches: StackLayout writes height 0 onto the current item, and since row controls are cached C++-side, nothing re-triggers the binding.

With nested TabViews, a collapse at any level zeroes the level above → the whole stack folds.

Fix

Push-based height tracking: each delegate pushes its height into itemStack.currentTabHeight whenever it is the current item (StackLayout.isCurrentItem attached property, plus Component.onCompleted for creation ordering). Freshly created delegates after a model reset always announce their height, so no stale references are possible. height on itemStack is now a plain declarative binding with guards for empty model / index −1.

No public API change (content, currentIndex alias, add/remove tabs untouched). SEM/PLSSEM content:-style usage unaffected — their initial-height case (defect 1) is fixed too.

Test checklist (jaspSem MNLFA, dataset with ≥6 scale variables)

1. Moderation Options: check/uncheck the invariance checkboxes one by one, incl. unchecking a tab other than the selected one → no fold-up.
2. Same-count resets: swap which checkboxes are checked → inner tabs never fold.
3. Factors: add a 3rd factor while the "Factors" tab is selected (adds "Covariances" entry), remove back to 1, rename factors, add/remove indicators → outer TabViews keep height.
4. Switch tabs at all 3 levels, incl. to a tab whose content grew while hidden.
5. Collapse/re-expand both Sections; also trigger a reset while collapsed, then expand.
6. Plots section: repeat 1–4; toggle interaction terms.
7. Resize window, change interface scale → no fold-ups.
8. Console: no TypeError: Cannot read property 'height' of null, no binding-loop warnings.
9. Regression SEM/PLS-SEM: model TextArea tab has correct height at first open; add/remove/rename tabs; multi-line lavaan syntax grows the TabView.
10. Save/reload a .jasp file with populated nested tabs → all levels render at full height.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR fixes an intermittent UI layout failure where TabView content could collapse to zero height after the underlying tab model resets (common with dynamic/nested TabViews whose model is recomputed). The solution replaces a fragile pull-based height binding (dependent on Repeater.itemAt()) with push-based height tracking from the currently active delegate.

Changes:

• Replace onCurrentIndexChanged-installed height binding with a persistent declarative height binding on the StackLayout.
• Introduce currentTabHeight on the StackLayout, updated by delegates when they become current and when their height changes.
• Add inline documentation explaining why Repeater.itemAt() cannot be safely used as a binding dependency across model resets.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.