feat(storage): app-pause flush + vault search index
Some checks failed
CI / Windows build (push) Has been cancelled

Phase 6 (final storage phase).

- SidecarRepositoryRegistry tracks every open repo; SidecarFlushObserver
  (a WidgetsBindingObserver in main) flushes them all on
  inactive/hidden/paused/detached, awaiting each flush — the last
  strokes can't be lost on app close, not just on the 800ms timer.
- VaultSearchIndex rebuilds by scanning vault sidecars (the source of
  truth) — note titles, OCR text and document names — and search_provider
  queries it, so search spans notes + PDFs. Rebuilt on launch / after
  import.

The vault file-based storage migration (Phases 0-6) is complete:
annotations travel with the file, picked vault folder, atomic autosave,
one Import-file entry, SQLite migrated to sidecars. analyze clean,
tests green.
This commit is contained in:
2026-06-24 23:19:21 +08:00
parent 4886f1b2df
commit 24d13642fd
9 changed files with 749 additions and 68 deletions

View File

@@ -0,0 +1,58 @@
// lib/editor/persistence/sidecar_flush_observer.dart
//
// Phase 6 / §F.3 of the file-based storage plan (docs/plans/2026-06-24-file-
// based-storage.md): app-lifecycle flush hardening.
//
// The per-file SidecarRepository debounces writes by 800 ms. That window is the
// data-loss gap on a Windows tablet: if the OS suspends or closes the app
// before the timer fires, the last strokes never reach disk. This observer
// listens for the app leaving the foreground and DRAINS every open repo's
// pending write before the process can be frozen, so "never lose the last
// strokes on app close" holds even when the editor's own dispose() doesn't run.
//
// Registered once in BadNoteApp; it delegates to
// [SidecarRepositoryRegistry.flushAll], which awaits every repo's flush().
import 'package:flutter/widgets.dart';
import 'sidecar_repository.dart';
/// A [WidgetsBindingObserver] that flushes all open sidecar repositories when
/// the app leaves the foreground (`inactive`/`paused`/`detached`/`hidden`).
class SidecarFlushObserver with WidgetsBindingObserver {
/// Whether the observer is currently registered with the binding.
bool get isAttached => _attached;
bool _attached = false;
/// Register with [WidgetsBinding.instance] so lifecycle changes are observed.
void attach() {
if (_attached) return;
WidgetsBinding.instance.addObserver(this);
_attached = true;
}
/// Stop observing lifecycle changes.
void detach() {
if (!_attached) return;
WidgetsBinding.instance.removeObserver(this);
_attached = false;
}
@override
void didChangeAppLifecycleState(AppLifecycleState state) {
switch (state) {
// Any transition out of the foreground is a potential suspend/kill point:
// drain pending sidecar writes now (the editors' own dispose() may never
// run when the OS freezes the process).
case AppLifecycleState.inactive:
case AppLifecycleState.hidden:
case AppLifecycleState.paused:
case AppLifecycleState.detached:
// Fire-and-forget at the framework boundary, but each write is atomic
// and awaited inside flushAll, so a half-written sidecar is impossible.
SidecarRepositoryRegistry.flushAll();
case AppLifecycleState.resumed:
break;
}
}
}

View File

@@ -29,6 +29,53 @@ import '../engine/stroke_model.dart';
/// Suffix appended to a source-file path to form its sidecar path.
const String kSidecarSuffix = '.badnote.json';
/// Process-wide registry of OPEN [SidecarRepository] instances (Phase 6 / §F.3).
///
/// The 800 ms debounce timer only protects against losing work to a crash that
/// happens *between* edits; it does NOT help when the OS suspends or kills the
/// app mid-window (the main data-loss window on a Windows tablet). The app's
/// lifecycle observer ([SidecarFlushObserver]) calls [flushAll] on
/// `paused`/`inactive`/`detached` to drain every open repo's pending write
/// before the process can be frozen.
///
/// A repo registers itself in [open] and removes itself in [dispose], so the
/// set always reflects exactly the editors holding unsaved sidecar state.
class SidecarRepositoryRegistry {
SidecarRepositoryRegistry._();
static final Set<SidecarRepository> _open = <SidecarRepository>{};
/// The currently open repositories (for tests / inspection).
static Set<SidecarRepository> get open => Set.unmodifiable(_open);
/// Flush every open repository's pending debounced write and await them all.
/// Safe to call repeatedly; a repo with nothing pending is a cheap no-op.
static Future<void> flushAll() async {
// Snapshot first: a flush may complete and (in a future) trigger disposal,
// which mutates `_open` — iterating a copy avoids concurrent-modification.
final repos = List<SidecarRepository>.of(_open);
await Future.wait(repos.map((r) => r.flush()));
}
/// The open repository for [sourceFilePath], or null if none is open. Lets a
/// background task (e.g. OCR) write through the SAME in-memory sidecar the
/// editor holds, instead of racing it with a second open handle.
static SidecarRepository? forPath(String sourceFilePath) {
for (final r in _open) {
if (r.sourceFilePath == sourceFilePath) return r;
}
return null;
}
static void _register(SidecarRepository repo) => _open.add(repo);
static void _unregister(SidecarRepository repo) => _open.remove(repo);
/// Test-only: drop all registrations so one test can't leak repos into the
/// next. Does NOT flush or dispose them.
static void resetForTest() => _open.clear();
}
/// Per-file persistence for the pen editor. Loads the sidecar for a source file
/// path, holds it in memory, and debounces atomic writes back to disk.
class SidecarRepository {
@@ -51,6 +98,12 @@ class SidecarRepository {
Timer? _timer;
bool _disposed = false;
/// Tail of the in-flight write chain. Writes are serialized through this so a
/// debounce-timer write and a concurrent lifecycle [flush] can't race on the
/// same `.tmp`/rename (which would throw on the loser). Each write always
/// persists the LATEST snapshot, so collapsing overlapping writes is safe.
Future<void> _writeChain = Future<void>.value();
/// Open (or create) the repository for [sourceFilePath]. Reads the existing
/// sidecar if present (falling back to its `.bak`), else starts empty.
static Future<SidecarRepository> open(
@@ -66,11 +119,13 @@ class SidecarRepository {
docType: docType,
createdAt: DateTime.now().toUtc(),
);
return SidecarRepository._(
final repo = SidecarRepository._(
sourceFilePath: sourceFilePath,
sidecar: sidecar,
debounce: debounce,
);
SidecarRepositoryRegistry._register(repo);
return repo;
}
// ── Loaded snapshot accessors (read at open) ───────────────────────────────
@@ -99,6 +154,17 @@ class SidecarRepository {
_replace(title: title);
}
/// Replace the handwriting-OCR search text and schedule a save (Phase 6
/// search index). No-op if unchanged.
void scheduleOcrTextSave(String? ocrText) {
final next = (ocrText != null && ocrText.isEmpty) ? null : ocrText;
if (_sidecar.ocrText == next) return;
_replace(ocrText: next, clearOcrText: next == null);
}
/// The OCR text loaded from the sidecar, or null.
String? get loadedOcrText => _sidecar.ocrText;
/// Replace the committed strokes for [pageIndex] and schedule a save.
void scheduleStrokeSave(int pageIndex, List<EditorStroke> strokes) {
final next = Map<int, List<EditorStroke>>.from(_sidecar.strokes);
@@ -165,12 +231,19 @@ class SidecarRepository {
// ── Flush / dispose ────────────────────────────────────────────────────────
/// Write any pending change immediately and wait for it to land.
/// Write any pending change immediately and wait for it (and any in-flight
/// write) to land. If the debounce timer is still armed, fire one final write
/// of the latest snapshot; otherwise just drain whatever write is in flight.
Future<void> flush() async {
if (_timer == null) return;
_timer!.cancel();
_timer = null;
await _write();
if (_timer != null) {
_timer!.cancel();
_timer = null;
await _write();
return;
}
// No pending edit, but a fire-and-forget timer write may still be running:
// await the chain so the bytes are on disk before we return.
await _writeChain;
}
/// Cancel pending timers. Call [flush] first to persist pending writes.
@@ -178,6 +251,7 @@ class SidecarRepository {
_disposed = true;
_timer?.cancel();
_timer = null;
SidecarRepositoryRegistry._unregister(this);
}
// ── Internal ───────────────────────────────────────────────────────────────
@@ -190,6 +264,8 @@ class SidecarRepository {
Map<int, List<EditorStroke>>? strokes,
Map<int, List<SidecarHighlight>>? highlights,
List<SidecarScratchLink>? scratchLinks,
String? ocrText,
bool clearOcrText = false,
}) {
if (_disposed) return;
_sidecar = BadnoteSidecar(
@@ -205,6 +281,7 @@ class SidecarRepository {
highlights: highlights ?? _sidecar.highlights,
bookmarks: _sidecar.bookmarks,
scratchLinks: scratchLinks ?? _sidecar.scratchLinks,
ocrText: clearOcrText ? null : (ocrText ?? _sidecar.ocrText),
);
_timer?.cancel();
_timer = Timer(_debounce, () {
@@ -215,9 +292,17 @@ class SidecarRepository {
});
}
Future<void> _write() async {
final snapshot = _sidecar;
await SidecarStore.writeAtomic(sidecarFile, snapshot);
/// Serialize writes through [_writeChain] so overlapping flushes never race
/// on the temp file. Each link writes the latest in-memory snapshot at the
/// moment it runs; an error in one write doesn't break the chain for the next.
Future<void> _write() {
final next = _writeChain.then((_) async {
final snapshot = _sidecar;
await SidecarStore.writeAtomic(sidecarFile, snapshot);
});
// Keep the chain alive past a failed write (e.g. transient FS error).
_writeChain = next.catchError((_) {});
return next;
}
static String _basename(String path) {

View File

@@ -6,6 +6,7 @@ import 'package:google_fonts/google_fonts.dart';
import 'package:pdfrx/pdfrx.dart';
import 'package:shared_preferences/shared_preferences.dart';
import 'editor/persistence/sidecar_flush_observer.dart';
import 'editor/pdf/pen_capture_region.dart';
import 'l10n/app_localizations.dart';
import 'providers/settings_provider.dart';
@@ -33,11 +34,33 @@ Future<void> main() async {
runApp(const ProviderScope(child: BadNoteApp()));
}
class BadNoteApp extends ConsumerWidget {
class BadNoteApp extends ConsumerStatefulWidget {
const BadNoteApp({super.key});
@override
Widget build(BuildContext context, WidgetRef ref) {
ConsumerState<BadNoteApp> createState() => _BadNoteAppState();
}
class _BadNoteAppState extends ConsumerState<BadNoteApp> {
// Phase 6 / §F.3: flush any open sidecar repos when the app is suspended or
// closed so the last strokes are never lost to an OS kill. Lives for the whole
// app lifetime (attached here, detached on app teardown).
final SidecarFlushObserver _flushObserver = SidecarFlushObserver();
@override
void initState() {
super.initState();
_flushObserver.attach();
}
@override
void dispose() {
_flushObserver.detach();
super.dispose();
}
@override
Widget build(BuildContext context) {
final settings = ref.watch(settingsProvider);
// Material You: prefer the OS dynamic color (Windows/Android system accent);

View File

@@ -1,16 +1,24 @@
import 'package:flutter_riverpod/flutter_riverpod.dart';
import '../models/document.dart';
import '../models/note.dart';
import '../services/database_service.dart';
import '../services/vault_search_index.dart';
import 'document_provider.dart' show vaultServiceProvider;
/// SQLite handle for the FTS/OCR search index. SQLite is now demoted to a
/// rebuildable search cache (the vault sidecars are the source of truth); search
/// is its sole remaining read path until the Phase 6 index rebuild lands.
final databaseServiceProvider = FutureProvider<DatabaseService>((ref) async {
return DatabaseService.getInstance();
/// The search index, rebuilt by SCANNING the vault sidecars (the source of
/// truth) — NOT the demoted SQLite cache (Phase 6, §B/§F). Bumping
/// [searchIndexEpochProvider] (e.g. after an import or note edit) invalidates
/// this provider so the next read re-scans the vault from disk.
final vaultSearchIndexProvider = FutureProvider<VaultSearchIndex>((ref) async {
ref.watch(searchIndexEpochProvider);
final vault = await ref.watch(vaultServiceProvider.future);
final index = VaultSearchIndex(vault);
await index.rebuild();
return index;
});
/// Bump to force the search index to rebuild from disk (e.g. after an import).
final searchIndexEpochProvider = StateProvider<int>((ref) => 0);
final searchQueryProvider = StateProvider<String>((ref) => '');
/// A search result that can be either a note hit or a document hit.
@@ -41,59 +49,45 @@ class DocumentSearchHit extends SearchResult {
final searchResultsProvider = FutureProvider<List<SearchResult>>((ref) async {
final query = ref.watch(searchQueryProvider);
if (query.isEmpty) return [];
if (query.trim().isEmpty) return [];
// Obtain the DB through the provider graph so this participates in
// initialization and disposal like every other consumer.
final db = await ref.watch(databaseServiceProvider.future);
final index = await ref.watch(vaultSearchIndexProvider.future);
// Run the note and document searches concurrently.
final searches = await Future.wait([
db.searchNotes(query),
db.searchDocuments(query),
]);
final noteHits = searches[0] as List<Note>;
final docHits = searches[1] as List<Map<String, dynamic>>;
final hits = await index.search(query);
final results = <SearchResult>[];
// Add note results.
for (final note in noteHits) {
results.add(NoteSearchHit(note: note, snippet: note.title));
}
// Resolve document metadata without an N+1 loop: collect the distinct
// document ids referenced by the hits, look each up exactly once, then
// build the result list from the cached lookups.
final docIds = <String>{
for (final hit in docHits)
if (hit['document_id'] is String) hit['document_id'] as String,
};
final docEntries = await Future.wait(
docIds.map((id) async => MapEntry(id, await db.getDocument(id))),
);
final docsById = <String, Document>{
for (final entry in docEntries)
if (entry.value != null) entry.key: entry.value!,
};
for (final hit in docHits) {
final documentId = hit['document_id'];
if (documentId is! String) continue;
final doc = docsById[documentId];
if (doc == null) continue;
final pageNumber = hit['page_number'];
final content = hit['content'];
results.add(
DocumentSearchHit(
documentId: documentId,
filename: doc.filename,
filePath: doc.filePath,
pageNumber: pageNumber is int ? pageNumber : 0,
snippet: content is String ? content : '',
),
);
for (final hit in hits) {
final entry = hit.entry;
final snippet = hit.snippet.text;
if (entry.isNote) {
// Construct a lightweight Note whose id is the synthetic note path so
// PenNoteScreen re-keys the right sidecar on open. Strokes are hydrated
// lazily by the editor; the search list only needs id/title.
final now = DateTime.now();
results.add(
NoteSearchHit(
note: Note(
id: entry.openPath,
title: entry.title,
createdAt: now,
updatedAt: now,
),
snippet: snippet,
),
);
} else {
results.add(
DocumentSearchHit(
documentId: entry.id,
filename: entry.title,
filePath: entry.openPath,
// The scan-based index matches whole-notebook text, not per-page, so
// the document opens at its first page.
pageNumber: 0,
snippet: snippet,
),
);
}
}
return results;

View File

@@ -1,3 +1,6 @@
import 'dart:io';
import '../editor/persistence/sidecar_repository.dart';
import '../models/note.dart';
import '../models/pen_tool.dart';
import 'database_service.dart';
@@ -6,7 +9,12 @@ import 'stroke_rasterizer.dart';
/// Runs OCR locally: typed text from strokes + handwriting via platform OCR.
class OcrService {
/// Extract searchable text from [note] and merge into the local FTS index.
/// Extract searchable text from [note] and persist it for search. The text is
/// written to the note's `*.badnote.json` sidecar `ocrText` field — the
/// vault-scan source of truth the Phase 6 search index reads — and also
/// appended to the (rebuildable) SQLite FTS cache so legacy callers keep
/// working. [note.id] is the synthetic note path, which is exactly the
/// `sourceFilePath` the editor opened its [SidecarRepository] with.
Future<void> processNote(Note note) async {
final parts = <String>[];
@@ -40,6 +48,22 @@ class OcrService {
final combined = parts.join(' ').trim();
if (combined.isEmpty) return;
// Persist into the note's sidecar so the vault-scan search index finds it.
// Prefer the editor's already-open repo (same in-memory sidecar — no race);
// if the note is closed, open/flush/dispose a transient handle.
final open = SidecarRepositoryRegistry.forPath(note.id);
if (open != null) {
open.scheduleOcrTextSave(combined);
await open.flush();
} else if (await File('${note.id}$kSidecarSuffix').exists()) {
final repo = await SidecarRepository.open(note.id, docType: 'notebook');
repo.scheduleOcrTextSave(combined);
await repo.flush();
repo.dispose();
}
// Also keep the legacy SQLite FTS cache warm (rebuildable; not the source
// of truth). Harmless if the row is never read.
final db = await DatabaseService.getInstance();
await db.appendOcrToFts(note.id, combined);
}

View File

@@ -0,0 +1,180 @@
// lib/services/vault_search_index.dart
//
// Phase 6 of the file-based storage plan (docs/plans/2026-06-24-file-based-
// storage.md §B/§F): the search index, rebuilt by SCANNING the vault sidecars
// (the source of truth) rather than the demoted SQLite cache.
//
// What it indexes, per notebook/note folder, from its `*.badnote.json` sidecar:
// * the title (standalone notebooks) / source filename (file-backed docs),
// * every typed text box — EditorStroke(tool: text).textContent across all
// pages,
// * the handwriting OCR text persisted in the sidecar's `ocrText` field.
//
// It does NOT (yet) index a PDF's embedded text layer — that is a known gap (see
// the REPORT in the task / the class doc below). Matching uses the existing pure
// search primitives (normalize / rank / snippet), so CJK substring search works.
import 'dart:io';
import '../editor/search/search_ranking.dart';
import '../editor/search/search_snippet.dart';
import '../storage/badnote_sidecar.dart';
import '../storage/sidecar_store.dart';
import 'vault_service.dart';
/// One indexed notebook: where it lives and the text harvested from its sidecar.
class VaultSearchEntry {
const VaultSearchEntry({
required this.id,
required this.title,
required this.openPath,
required this.docType,
required this.text,
required this.isNote,
});
/// Stable id (the notebook folder path).
final String id;
/// Display title (note title / source filename).
final String title;
/// Path to pass to the editor: the in-vault source file for docs, or the
/// synthetic `<folder>/notebook` note path for standalone notebooks.
final String openPath;
/// `pdf` / `pptx` / `ppt` / `docx` / `notebook`.
final String docType;
/// All searchable text harvested from the sidecar (title + typed text + OCR),
/// joined for substring matching.
final String text;
/// True for standalone (free-ink) notebooks, false for file-backed documents.
final bool isNote;
}
/// A search hit over the vault: the entry plus a display snippet of the match.
class VaultSearchHit {
const VaultSearchHit({required this.entry, required this.snippet});
final VaultSearchEntry entry;
final Snippet snippet;
}
/// Builds and queries a scan-based full-text index over the vault sidecars.
///
/// The index is the list of [VaultSearchEntry]s built by [rebuild]; query is a
/// pure substring/rank over their harvested text (CJK-safe). Cheap enough to
/// rebuild lazily on demand for a single-user vault; there is no background
/// thread and no persisted index file (the SQLite cache is no longer the search
/// source of truth).
///
/// HONEST SCOPE (what search covers / does NOT):
/// * COVERS: note/doc titles, typed text boxes, and handwriting OCR text that
/// has been persisted into a sidecar's `ocrText` field.
/// * DOES NOT cover: a PDF's embedded (printed) text layer — only the user's
/// annotations are indexed, not the underlying document body. Indexing the
/// PDF text layer would require rendering each page through pdfrx at scan
/// time; deferred. OCR is only present where it has already been run and
/// written back to the sidecar.
class VaultSearchIndex {
VaultSearchIndex(this._vault);
final VaultService _vault;
List<VaultSearchEntry> _entries = const [];
bool _built = false;
/// The entries from the most recent [rebuild] (for tests / inspection).
List<VaultSearchEntry> get entries => List.unmodifiable(_entries);
/// Scan the vault and (re)build the in-memory index. Safe on an empty/missing
/// vault (yields an empty index). Never throws on a single unreadable sidecar.
Future<void> rebuild() async {
final entries = <VaultSearchEntry>[];
final notebooks = await _vault.scanNotebooks();
for (final nb in notebooks) {
final sidecar = await SidecarStore.read(
File('${nb.sourceFilePath}$kVaultSidecarSuffix'),
);
entries.add(
VaultSearchEntry(
id: nb.folderPath,
title: nb.filename,
openPath: nb.sourceFilePath,
docType: nb.docType,
text: _harvest(title: nb.filename, sidecar: sidecar),
isNote: false,
),
);
}
final notes = await _vault.scanNotes();
for (final note in notes) {
final sidecar = await SidecarStore.read(
File('${note.notePath}$kVaultSidecarSuffix'),
);
entries.add(
VaultSearchEntry(
id: note.folderPath,
title: note.title,
openPath: note.notePath,
docType: 'notebook',
text: _harvest(title: note.title, sidecar: sidecar),
isNote: true,
),
);
}
_entries = entries;
_built = true;
}
/// Search the index for [query], rebuilding it first if it has never been
/// built. Returns the best-matching notebooks/notes, most-relevant first. An
/// empty/whitespace query yields no hits.
Future<List<VaultSearchHit>> search(String query) async {
if (query.trim().isEmpty) return const [];
if (!_built) await rebuild();
final sources = <String, String>{
for (final e in _entries) e.id: e.text,
};
final ranked = rankHits(sources, query);
final byId = {for (final e in _entries) e.id: e};
final hits = <VaultSearchHit>[];
for (final hit in ranked) {
final entry = byId[hit.ref];
if (entry == null) continue;
hits.add(VaultSearchHit(entry: entry, snippet: hit.snippet));
}
return hits;
}
/// Concatenate every searchable string from one sidecar: the [title], every
/// typed text box across all pages, and the persisted handwriting OCR text.
static String _harvest({
required String title,
required BadnoteSidecar? sidecar,
}) {
final parts = <String>[title];
if (sidecar != null) {
// Typed text boxes: a stroke carries `textContent` regardless of tool
// (EditorTool has only pen/highlighter/eraser; text is a content flag).
for (final pageStrokes in sidecar.strokes.values) {
for (final stroke in pageStrokes) {
final text = stroke.textContent;
if (text != null && text.trim().isNotEmpty) {
parts.add(text.trim());
}
}
}
final ocr = sidecar.ocrText;
if (ocr != null && ocr.trim().isNotEmpty) parts.add(ocr.trim());
}
return parts.join('\n');
}
}

View File

@@ -222,6 +222,7 @@ class BadnoteSidecar {
List<Bookmark>? bookmarks,
List<SidecarScratchLink>? scratchLinks,
Map<int, String>? legacyAnnotations,
this.ocrText,
this.legacyId,
}) : strokes = strokes ?? <int, List<EditorStroke>>{},
highlights = highlights ?? <int, List<SidecarHighlight>>{},
@@ -263,6 +264,14 @@ class BadnoteSidecar {
/// sidecars.
final Map<int, String> legacyAnnotations;
/// Searchable text recovered from this notebook's handwriting via local OCR
/// (Phase 6 search index). Persisted in the sidecar — the source of truth —
/// so the vault-scan search index can find handwritten notes WITHOUT the
/// (rebuildable, per-device) SQLite cache. Typed text already lives in the
/// strokes' `textContent`, so this holds ONLY the OCR'd handwriting. Null when
/// the notebook has no handwriting or OCR hasn't run.
final String? ocrText;
/// The legacy SQLite row id this sidecar was migrated from (a `documents.id`
/// or `notes.id`). Set ONLY by the one-time migration; it makes the migration
/// idempotent (a re-run recognizes an already-migrated item by this id even if
@@ -295,6 +304,7 @@ class BadnoteSidecar {
for (final entry in legacyAnnotations.entries)
entry.key.toString(): entry.value,
},
if (ocrText != null && ocrText!.isNotEmpty) 'ocrText': ocrText,
if (legacyId != null) 'legacyId': legacyId,
};
@@ -349,6 +359,7 @@ class BadnoteSidecar {
}
return out;
}(),
ocrText: json['ocrText'] as String?,
legacyId: json['legacyId'] as String?,
);
}