fix: memory leak in search editor

[Simon Siefke (SimonSiefke)]

Details

Search editor panes registered each input's configuration listener for the pane lifetime, while the model factory kept resolved models after their input was disposed. Closing and reopening search editors therefore retained the input, configuration, and results text model graph.

Change

Configuration listeners are now scoped to the active editor input, and disposing a SearchEditorInput removes its model factory entry.

Before

When opening and closing a search editor 37 times, the search editor input, configuration, and results text model graph grows with each run:

After

No more search editor input or text model leak is detected. The remaining rows belong to the independent search result folder-match leak.

Test Video

test-video.webm

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR improves lifecycle management for Search Editor resources by ensuring per-input listeners and cached models are released when inputs are cleared or disposed.

Changes:

• Add explicit disposal for SearchEditorModel instances so they can be removed from the shared factory cache.
• Dispose the SearchEditorModel from SearchEditorInput.dispose() to avoid retaining cached models after an input is closed.
• Track and clear per-input disposables in SearchEditor to prevent event listener leaks across input switches.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
src/vs/workbench/contrib/searchEditor/browser/searchEditorModel.ts	Adds dispose() to remove a model from the global factory cache.
src/vs/workbench/contrib/searchEditor/browser/searchEditorInput.ts	Ensures the Search Editor model is disposed when the input is disposed.
src/vs/workbench/contrib/searchEditor/browser/searchEditor.ts	Introduces inputDisposables to scope and clear disposables when inputs change/clear.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.