Populate FTS5 index during sync and backfill existing databases#94
Merged
Populate FTS5 index during sync and backfill existing databases#94
Conversation
…29) The messages_fts table was created but never populated, causing CLI search and TUI deep search to always return zero results. This adds FTS population at two points: incrementally during sync (UpsertFTS) and as a one-time backfill for existing databases (BackfillFTS in InitSchema). Fixes #29. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… test - Wrap BackfillFTS delete+insert in a transaction so a mid-operation failure doesn't leave the FTS table empty - Return errors from COUNT(*) queries in InitSchema instead of silently swallowing them - Add TestStore_InitSchema_AutoBackfillFTS to cover the auto-backfill path (empty FTS + existing messages triggers backfill on InitSchema) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
These commands open the database but never called InitSchema, so the auto-backfill for existing databases with empty FTS tables never ran. This was the actual cause of "No messages found" — hasFTSTable() returned true (table exists), but the JOIN against the empty FTS table yielded zero rows. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move the auto-backfill out of InitSchema (which should be pure schema
setup) into a NeedsFTSBackfill() check + explicit BackfillFTS() call
in the CLI commands. This lets search/tui/mcp print a status message
("Building search index... indexed N messages.") so users know what's
happening during the one-time backfill on large archives.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- BackfillFTS now processes in batches of 5000 by ID range instead of one massive query, with an optional progress callback - TUI and MCP launch the backfill in a background goroutine so they start instantly — FTS is only needed for deep search, not aggregates - CLI search blocks with live progress: "Indexed 50000 / 1700000..." - NeedsFTSBackfill detects partial backfills (interrupted) via 90% threshold so they resume on next run - Fix CLAUDE.md stale data reference Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The search command now shows:
- A visual progress bar during one-time FTS index build:
Building search index (one-time)...
[================== ] 60% 1020000 / 1700000
- A "Searching..." indicator while the query executes
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The InitSchema and NeedsFTSBackfill COUNT(*) queries on 1.7M rows take seconds. Move the status message before store.Open so users see feedback instantly. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
COUNT(*) requires a full table scan in SQLite — several seconds on 1.7M rows. MAX(id) and MIN(id) are instant B-tree lookups. NeedsFTSBackfill now compares MAX(rowid) between tables, and BackfillFTS uses the ID range for progress reporting instead of an exact count. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Clamp progress done <= total and guard total > 0 to prevent strings.Repeat panic when cursor overshoots ID range - Use GROUP_CONCAT for from_addr in backfill query to match sync path (was LIMIT 1, inconsistent with persistMessage which joins all froms) - Fix stale comments: "total messages" → "ID range", correct threshold explanation in NeedsFTSBackfill Addresses review findings from jobs #4665, #4669, #4671. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Owner
Author
|
I confirmed that |
wesm
added a commit
to robelkin/msgvault
that referenced
this pull request
Feb 7, 2026
) ## Summary - Populate FTS5 full-text search index during sync (`UpsertFTS` per message) and backfill existing databases on first search/TUI/MCP launch - `BackfillFTS` processes in batches of 5000 by ID range with a live progress bar, avoiding multi-minute blocking on large archives (1.7M+ messages) - TUI and MCP run backfill in background goroutine so they launch instantly - CLI `search` shows `Searching...` immediately and a progress bar during one-time backfill - Use `MAX(id)` B-tree lookups instead of `COUNT(*)` table scans for instant backfill detection ## Details The `messages_fts` FTS5 table was created during `InitSchema` but never populated, causing `search` and TUI deep search to always return zero results. The FTS JOIN against the empty table matched nothing, with no fallback to LIKE. **Store layer** (`internal/store/`): - `UpsertFTS()` — inserts/replaces a single FTS row (called per message during sync) - `BackfillFTS(progress)` — batched bulk population from existing data with progress callback - `NeedsFTSBackfill()` — fast `MAX(id)` check to detect empty/incomplete FTS - `FTS5Available()` — accessor for FTS5 availability **Sync** (`internal/sync/`): - `persistMessage` now calls `UpsertFTS` after storing recipients, keeping FTS current **CLI commands** (`cmd/msgvault/cmd/`): - `search`, `tui`, `mcp` call `InitSchema` + backfill check on startup - `search` blocks with progress bar; `tui`/`mcp` backfill in background Fixes wesm#64. Fixes wesm#29. ## Test plan - [x] `go test -tags fts5 ./internal/store/ -run FTS` — UpsertFTS, BackfillFTS, NeedsFTSBackfill tests - [x] `make test` — full suite passes - [x] `make lint` — clean - [x] Manual: `./msgvault search "thank"` returns results after backfill 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
UpsertFTSper message) and backfill existing databases on first search/TUI/MCP launchBackfillFTSprocesses in batches of 5000 by ID range with a live progress bar, avoiding multi-minute blocking on large archives (1.7M+ messages)searchshowsSearching...immediately and a progress bar during one-time backfillMAX(id)B-tree lookups instead ofCOUNT(*)table scans for instant backfill detectionDetails
The
messages_ftsFTS5 table was created duringInitSchemabut never populated, causingsearchand TUI deep search to always return zero results. The FTS JOIN against the empty table matched nothing, with no fallback to LIKE.Store layer (
internal/store/):UpsertFTS()— inserts/replaces a single FTS row (called per message during sync)BackfillFTS(progress)— batched bulk population from existing data with progress callbackNeedsFTSBackfill()— fastMAX(id)check to detect empty/incomplete FTSFTS5Available()— accessor for FTS5 availabilitySync (
internal/sync/):persistMessagenow callsUpsertFTSafter storing recipients, keeping FTS currentCLI commands (
cmd/msgvault/cmd/):search,tui,mcpcallInitSchema+ backfill check on startupsearchblocks with progress bar;tui/mcpbackfill in backgroundFixes #64. Fixes #29.
Test plan
go test -tags fts5 ./internal/store/ -run FTS— UpsertFTS, BackfillFTS, NeedsFTSBackfill testsmake test— full suite passesmake lint— clean./msgvault search "thank"returns results after backfill🤖 Generated with Claude Code