refactor(editor)!: move the vendored editor into Packages/TableProEditor and rename its modules

.claude/skills/fix-issue/references/verification.md

@@ -129,7 +129,7 @@ The `DEVELOPER_DIR` export above is required; without it SwiftLint aborts with `
 
 Two things to know:
 
-- **SwiftLint's `included:` scope is `TablePro` only.** `Plugins/`, `LocalPackages/`, and the test targets are never linted by a bare `swiftlint lint`. Pass explicit paths to lint a change outside the app target.
+- **SwiftLint's `included:` scope is `TablePro` and `Packages`.** The rest of `Plugins/` and the app's test targets are never linted by a bare `swiftlint lint`. Pass explicit paths to lint a change outside that scope.
 - **Local `swiftformat` is a version behind the repo `.swiftformat`** and rejects `--ifdefindent`, so it cannot run here. Rely on SwiftLint plus reading the diff.
 - **Never remove a `swiftlint:disable force_unwrapping` comment** to satisfy a local run. There are five of them, inline in four files, not in the config: `.swiftlint.yml` enables the rule as an opt-in and sets it to `warning`. The CI toolchain differs from this one, so a disable that looks unnecessary here is load-bearing there.
 

.github/workflows/macos-tests.yml

@@ -12,7 +12,6 @@ on:
       - "TablePro/**"
       - "Plugins/**"
       - "Packages/**"
-      - "LocalPackages/**"
       - "TableProTests/**"
       - "TableProUITests/**"
       - "Native/**"
@@ -95,7 +94,7 @@ jobs:
             --base "$BASE_SHA" \
             --ref "$REF" \
             --skip-release-commit \
-            'TablePro/' 'Plugins/' 'Packages/' 'LocalPackages/' \
+            'TablePro/' 'Plugins/' 'Packages/' \
             'TableProTests/' 'TableProUITests/' 'Native/' 'Configs/' 'Libs/' 'scripts/' \
             '\.github/actions/' '\.github/macos-(ui-)?test-quarantine\.txt$' \
             'TablePro\.xcodeproj/project\.xcworkspace/xcshareddata/swiftpm/Package\.resolved$' \
@@ -145,16 +144,12 @@ jobs:
         if: needs.changes.outputs.oracle == 'true'
         run: swift test --package-path Packages/TableProOracle --force-resolved-versions
 
-      - name: Run CodeEditTextView package tests
-        run: swift test --package-path LocalPackages/CodeEditTextView --force-resolved-versions
 
-      # The editor's own suites, 111 XCTest cases and 102 Swift Testing cases, roughly half of them
-      # TablePro's. Two XCTest cases skip themselves on a headless runner (no screen, no key window),
-      # so the expected result is 111 executed with 2 skipped. Pinned to the tracked Package.resolved
-      # like its siblings; note that this package and CodeEditTextView are path-linked, so a
-      # SwiftTreeSitter bump has to land in both lockfiles or this step fails rather than floating.
-      - name: Run CodeEditSourceEditor package tests
-        run: swift test --package-path LocalPackages/CodeEditSourceEditor --force-resolved-versions
+      # The editor, both targets. Two XCTest cases skip themselves on a headless runner (no screen,
+      # no key window), so the expected result is 112 executed with 2 skipped, plus 102 Swift Testing
+      # cases. Pinned to the tracked Package.resolved like its siblings.
+      - name: Run TableProEditor package tests
+        run: swift test --package-path Packages/TableProEditor --force-resolved-versions
 
   # Compiles the app, both test bundles and all 31 plugins once, then hands the products to the
   # test jobs. That build used to happen inside the same job that ran the tests, which meant the

.gitignore

@@ -63,7 +63,6 @@ playground.xcworkspace
 .swiftpm/
 
 .build/
-LocalPackages/*/Example/**/Package.resolved
 
 # Coverage profile output
 *.profraw

CLAUDE.md

@@ -26,9 +26,9 @@ TablePro is a native macOS database client (SwiftUI + AppKit), a fast, lightweig
     - **Registry-only** (the other 23): MongoDB, Oracle, DuckDB, MSSQL, Cassandra, Etcd, CloudflareD1, CloudflareR2SQL, DynamoDB, BigQuery, Spanner, LibSQL, Snowflake, Elasticsearch, Typesense, Beancount, SurrealDB, Teradata, Trino, Dameng, Kafka, Weaviate, Parquet export. Parquet is registry-only because it links its own copy of DuckDB, which does the encoding, and that is too large to ship in the app for one format. Distributed via [TableProApp/plugins](https://github.com/TableProApp/plugins) `plugins.json`, installed into the user plugins directory.
 - **C bridges**: Each plugin contains its own C bridge module (e.g., `Plugins/MySQLDriverPlugin/CMariaDB/`, `Plugins/PostgreSQLDriverPlugin/CLibPQ/`)
 - **Static libs**: `Libs/` holds pre-built `.a` files and `Libs/ios/` holds the iOS xcframeworks. Both are downloaded by `scripts/download-libs.sh` and are not in git.
-- **SPM deps**: declared in `project.yml`. Vendored forks under `LocalPackages/` (CodeEditSourceEditor, CodeEditTextView) and TablePro's own packages under `Packages/` (TableProCore, TableProOracle, TableProGrammars); remote packages are Sparkle, swift-certificates and Yams. Revisions are pinned by the tracked `Package.resolved` inside each generated `.xcodeproj`.
+- **SPM deps**: declared in `project.yml`. Every local package lives under `Packages/` (TableProCore, TableProOracle, TableProEditor, TableProGrammars); remote packages are Sparkle, swift-certificates and Yams. Revisions are pinned by the tracked `Package.resolved` inside each generated `.xcodeproj`.
     - `Packages/TableProGrammars` is first-party code, not a fork. It holds TablePro's `CodeLanguage`, `GrammarID` and `HighlightQueries` plus the four tree-sitter grammars the app compiles (bash, javascript, json, sql), each with its own upstream MIT licence beside its sources and a record in `licenses.yml`. Every `GrammarID` case but `plainText` resolves to a parser, and a test asserts it: the enum is the list of grammars that ship, not a list of languages tree-sitter supports.
-    - `SWIFT_VERSION` in `Configs/Base.xcconfig` sets the language mode for the Xcode-native targets only. A SwiftPM package takes its mode from its own manifest, so `Packages/TableProCore` and `Packages/TableProOracle` carry `swift-tools-version: 6.0` of their own. The vendored `LocalPackages/` forks stay on 5.9 so they can still take upstream changes, and a remote dependency keeps whatever its own manifest says. Never pass `SWIFT_VERSION=` on an `xcodebuild` command line to test a language-mode change: the override reaches the package targets too and reports their errors as yours.
+    - `SWIFT_VERSION` in `Configs/Base.xcconfig` sets the language mode for the Xcode-native targets only. A SwiftPM package takes its mode from its own manifest, so every package under `Packages/` carries `swift-tools-version: 6.0` of its own, and a remote dependency keeps whatever its own manifest says. `TableProEditor` sets `.swiftLanguageMode(.v5)` on its targets: the manifest is current but the code has not been through the Swift 6 concurrency migration yet, and that is a change with its own tests rather than a side effect of a move. Never pass `SWIFT_VERSION=` on an `xcodebuild` command line to test a language-mode change: the override reaches the package targets too and reports their errors as yours.
 
 ## Build & Development Commands
 
@@ -156,14 +156,14 @@ When adding a new method to the driver protocol: add to `PluginDatabaseDriver` (
 - Unknown types (from future plugins) are valid, they round-trip through Codable
 - Use `DatabaseType.allKnownTypes` (not `allCases`) for the canonical list
 
-### Editor Architecture (CodeEditSourceEditor)
+### Editor Architecture (TableProEditor)
 
 - **`ThemeEngine`**: the `@Observable` singleton that owns the active theme, and the single source of truth for editor colors and fonts
 - **Two font domains, and a control never names a third**: `editorFonts` is the Editor Font setting and belongs to a view that also wears the editor's background, text color and syntax palette (the SQL editor, `JSONCodeEditor`, the fold and import previews). Every control that shows or edits a stored *value* takes `ThemeEngine.valueFont` / `valueFontSwiftUI`, which is the Data Grid Font setting, so one value reads the same in the grid cell, its inline editor, the row inspector, a cell popover and a pop-out window. Naming a system text style instead is the defect: it looks right only while the user leaves both settings equal, which is the default, so it ships invisibly (#2393). The inspector sets it once for the whole editor subtree in `InspectorFieldRow`, past the field label and the value menu, so a new field editor inherits it. The switch that decides it is exhaustive over `FieldEditorKind` on purpose: two `default`-armed switches are what let `.typePicker` escape the domain and render a structure row's Name and Type in two different fonts. AppKit will not honour it on a `Picker(.menu)`'s items, which `NSMenu` draws itself; that is the one place the rule stops.
-- **`TableProEditorTheme`**: adapter to CodeEdit's `EditorTheme` protocol; `ThemeEngine.makeEditorTheme()` builds it
-- **`CompletionEngine`**: framework-agnostic; **`QueryCompletionAdapter`** bridges to CodeEdit's `CodeSuggestionDelegate`
+- **`TableProEditorTheme`**: adapter to `TableProEditorKit`'s `EditorTheme` protocol; `ThemeEngine.makeEditorTheme()` builds it
+- **`CompletionEngine`**: framework-agnostic; **`QueryCompletionAdapter`** bridges to `TableProEditorKit`'s `CodeSuggestionDelegate`
 - Editor tabs are drawn by `EditorTabStrip`, not by native window tabs. A window belongs to exactly one `NSWindow` tab group and that group's bar shows every window in it, so a window hosting several connections could only ever show all of their tabs interleaved. Window tabbing itself stays on AppKit's terms: `TabWindowController` leaves `tabbingMode` at `.automatic`, which is the user's own System Settings preference, and never forces `.preferred`.
-- Cursor model: `cursorPositions: [CursorPosition]` (multi-cursor via CodeEditSourceEditor)
+- Cursor model: `cursorPositions: [CursorPosition]` (multi-cursor via `TableProEditorKit`)
 
 ### Change Tracking Flow
 
@@ -339,7 +339,7 @@ These are **non-negotiable**, never skip them:
 
 4. **Tests**: Every change with testable behavior must include or update unit/function tests. UI and user-flow changes should add or update `TableProUITests` UI automation where the flow runs deterministically; if it can't, note why in the PR description. When tests fail, fix the source code, never adjust tests to match incorrect output. Tests define expected behavior.
 
-5. **Lint after changes**: Run `swiftlint lint --strict` to verify compliance. `.swiftlint.yml` sets `included: [TablePro, Packages]`, so a bare run never sees `Plugins/`, `TableProTests/` or `TableProUITests/`.
+5. **Lint after changes**: Run `swiftlint lint --strict` to verify compliance. `.swiftlint.yml` sets `included: [TablePro, Packages]`, so a bare run covers the app and every local package, including `TableProPluginKit` through the symlink in `TableProCore`. It still does not see the rest of `Plugins/`, `TableProTests/` or `TableProUITests/`.
 
     **Reaching those takes file paths, not directory paths.** SwiftLint applies `included:` to a directory argument and not to a file argument, so `swiftlint lint --strict TableProTests` reports zero violations having linted nothing at all: measured, that command finds 0 and `TableProTests/**/*.swift` finds 1,227. Glob to files for a change outside `included:`, e.g. `swiftlint lint --strict $(git show --name-only --format= HEAD | grep '\.swift$')`. `verify.sh lint` names any directory it was handed and then dropped, so a clean result is not read as coverage it never had.
 

Packages/TableProEditor/ORIGIN.md

@@ -0,0 +1,52 @@
+# TableProEditor
+
+The SQL editor: the text engine, its layout and selection, syntax highlighting, folding, the gutter
+and the completion window. It began as two packages from the
+[CodeEdit](https://github.com/CodeEditApp) project and TablePro has diverged from both far enough
+that they are no longer upstreamable as a whole.
+
+| Target | Vendored from | Licence |
+| --- | --- | --- |
+| `TableProTextEngine`, `TableProTextEngineObjC` | https://github.com/CodeEditApp/CodeEditTextView at `d7ac3f11f` (2025-07-30) | MIT, `LICENSE-CodeEditTextView.md` |
+| `TableProEditorKit` | https://github.com/CodeEditApp/CodeEditSourceEditor at `1fa4d3c3f` (2025-12-31) | MIT, `LICENSE-CodeEditSourceEditor.md` |
+
+Both were vendored in 2026-03 at what was then upstream's tip, and upstream has landed nothing since.
+
+## Why it is one package and not two
+
+Upstream split the engine from the editor because they are two products. Here they are one: 32% of
+the commits that touched either touched both, and the strongest co-change pair in the repository is
+`TableProEditorKit`'s controller against `TableProTextEngine`'s text view, which the old boundary put
+on opposite sides. The targets keep the layering the split was for, and SwiftPM rejects a cycle
+between them just as it rejected one between the packages.
+
+## What TablePro changed
+
+Roughly 7,000 lines of source and 7,400 lines of tests, in three groups.
+
+**Correctness and performance in the engine.** Clip-bounded CoreText drawing, per-line widths in the
+red-black tree so `maxLineWidth` falls when the widest line is deleted, a typesetting fix where a
+line break was used as a length rather than an offset, IME and `NSTextInputClient` range handling,
+pasteboard coercion, invisible-character rendering, and range-safety helpers. Most of this is
+general and upstream would plausibly want it.
+
+**Product features.** Per-statement run controls in the gutter, and the fold placeholder preview.
+These carry no SQL knowledge: `StatementRun` holds an `NSRange` and nothing else, and every SQL
+scanner lives in `TablePro/Core/Utilities/SQL/`.
+
+**Replacements.** The completion window was rewritten from `NSTableView` to SwiftUI, and the fold
+ribbon from upstream's hover-marker strip to macOS disclosure chevrons. Six upstream files were
+deleted outright.
+
+Four subsystems upstream ships were removed because TablePro never used them and two of them were
+running anyway: the minimap, jump to definition, the reformatting guide and HTML tag completion
+(#2877).
+
+## Taking a change from upstream
+
+Do not expect a rebase to work. 42 of 65 files TablePro touched in the engine, and 41 of 70 in the
+editor, collide with what upstream changed in its own last 60 commits. Port a specific fix by reading
+it and writing it here, the way any other bug report is handled.
+
+The reverse direction is open: the correctness and performance work is written against upstream's own
+shapes and roughly 1,060 lines of it lift out cleanly if anyone wants to send it back.