vavallee/bindery

Pre-flight Searched existing issues. #2102 added the display name sort that is now labelled "First name A to Z"; nothing covers naming the default or the two gaps below. Checked the roadmap. Problem The Authors grid sort menu reads, in order: authors.sortAZ, authors.sortZA, authors.sortFirstNameAZ, authors.sortFirstNameZA, authors.sortRecent. The first two are labelled only with the letter range, so a menu that names one option "First name" and leaves the other unqualified reads as though last name ordering is missing. It is not missing. It is the default. authorSortOrder's default arm (internal/db/authors.go:179-181) orders by sort_key, sort_name, and sort_name is produced by textutil.SortName (internal/textutil/authorquery.go:70), which inverts a display name into "Last, First" at every provider and import path (openlibrary, googlebooks, hardcover, the list syncer, ABS, the renamer, api/helpers.go). So the default sort already is by last name, including the particle handling that

Current behaviour BookRepo.SetFilePath in internal/db/books.go falls back to a legacy single-column write when GetByID fails or returns no row: _, err2 := r.db.ExecContext(ctx, "UPDATE books SET file_path=?, status=? WHERE id=?", filePath, models.BookStatusImported, id) This fallback has two problems: It changes status without bumping updated_at. Anything that detects edits through updated_at misses the change, including the optimistic hydration write guard proposed for #2758. It runs on r.db rather than r.exec, so inside a transaction-bound repo it bypasses the active transaction. Expected behaviour The fallback sets updated_at = timeValueArg(time.Now().UTC()) alongside status, and runs through r.exec like the other BookRepo writers. Steps to reproduce Code inspection. The path is only reached when GetByID errors or the row is missing, so it is rare in practice. A test can reach it by making GetByID fail on a transaction-bound repo, then checking that updated_at is unchanged and

Current behaviour AuthorAliasRepo.Merge in internal/db/author_aliases.go (around line 368) re-parents the source author's books with "UPDATE books SET author_id = ?, updated_at = ? WHERE author_id = ?", targetID, time.Now().UTC(), sourceID) It passes a raw time.Time, not timeValueArg(...). The modernc.org/sqlite driver then stores Go's default time.String() shape, e.g. 2026-09-25 21:43:14.920986 +0000 UTC. That is the #914 corruption shape that timeArg / timeValueArg in internal/db/books.go were introduced to stop writing. Every author merge writes it again for each re-parented book. Reads still work only because parseFlexibleTime tolerates that layout. Anything that compares or orders books.updated_at as text sees mixed shapes on those rows. Expected behaviour Merge writes updated_at in the canonical RFC3339Nano shape, via timeValueArg(time.Now().UTC()), like every other books writer. Steps to reproduce Merge two authors that each have at least one book. SELECT CAST(updated_at AS TEX

Pre-flight I searched existing issues and this isn't a duplicate I'm running the latest version (check Settings → About) Current behaviour internal/bookhydrate/hardcover.go:306: func editionHasAudioMarker(text string) bool { for _, marker := range []string{"audio", "audible", "mp3", "cd", "cassette"} { if strings.Contains(text, marker) { return true } } return false } The haystack is Format + " " + EditionInfo (isLikelyAudioEdition, :298), and EditionInfo is free text from Hardcover with no controlled vocabulary. Four of the five markers are safe as substrings. "cd" is not: it is two common letters and will match inside an ordinary word. An edition wrongly classified as audio is then eligible for two consequential things: ASIN promotion (maybePromoteASIN, :247) and media type widening (deriveAudiobookMetadataFromEdition, :210-220), which also re-opens the book as wanted through ReevaluateStatus(). I have not observed this in the wild, so I do not know the real frequency

Reported by Daize in Discord: there is no way to monitor or unmonitor a single book from the book page. Confirmed, and it is an omission rather than a decision. The book page shows the monitored state and gives you no way to change it. Where you can change it today Books list, tick a book's checkbox and use the bulk Monitor / Unmonitor action (web/src/pages/BooksPage.tsx:416-417). Works for one book, but you have to go back to the list and select it. Wanted page, "stop monitoring" on a row (web/src/pages/WantedPage.tsx:107, api.updateBook(book.id, { monitored: false })). One direction only. So the capability exists and is reachable from two other screens, just not from the page dedicated to the book. Why this is small PUT /api/v1/book/{id} already accepts it (internal/api/books.go:452, :468): Monitored *bool `json:"monitored"` ... if req.Monitored != nil { book.Monitored = *req.Monitored } And BookDetailPage already has the generic save helper wired to that endpoint (web/s