fix(metadata): pick the right book row per hardcover series position

tests/test_ebooks.py

@@ -140,6 +140,46 @@ def test_hardcover_series_prefers_the_row_with_an_edition_in_the_language(monkey
     assert [b.title for b in books] == ["Super-Papagei", "Der Phantomsee", "Sonderband"]
 
 
+INTRUDER_ROWS = {
+    "name": "Villains",
+    "book_series": [
+        {"position": 1, "book": {
+            "id": 1, "title": "Vicious", "users_count": 1833,
+            "contributions": [{"author": {"name": "V. E. Schwab"}}]}},
+        {"position": 1, "book": {
+            "id": 2, "title": "Die Schoenste im ganzen Land", "users_count": 1,
+            "contributions": [{"author": {"name": "Serena Valentino"}}]}},
+        {"position": 2, "book": {
+            "id": 3, "title": "Coeur de pierre", "users_count": 1,
+            "contributions": [{"author": {"name": "Serena Valentino"}}]}},
+        {"position": 2, "book": {
+            "id": 4, "title": "Cold Hearted", "users_count": 76,
+            "contributions": [{"author": {"name": "Serena Valentino"}}]}},
+    ],
+}
+
+
+def test_hardcover_series_skips_a_foreign_book_filed_under_a_volume(monkeypatch):
+    """Hardcover lets anyone file a book into a series - "Vicious" sits on volume
+    1 of the Disney Villains and has a German edition and 1833 readers, so both
+    tie breakers would hand it the slot."""
+    def handler(request: httpx.Request) -> httpx.Response:
+        import json
+        if "editions" in json.loads(request.content)["query"]:
+            return httpx.Response(200, json={"data": {"editions": [
+                {"book_id": 1, "title": "Vicious - Das Boese in uns"},
+                {"book_id": 2, "title": "Die Schoenste im ganzen Land"},
+            ]}})
+        return httpx.Response(200, json={"data": {"series": [INTRUDER_ROWS]}})
+
+    monkeypatch.setattr(hardcover, "TOKEN", "test-token")
+    mock_client(monkeypatch, hardcover, handler)
+    books = asyncio.run(hardcover.series_books("976", "german"))
+    # volume 2 has no German edition: the most read row beats the French one
+    assert [b.title for b in books] == [
+        "Die Schoenste im ganzen Land", "Cold Hearted"]
+
+
 def test_hardcover_series_survives_a_broken_edition_lookup(monkeypatch):
     def handler(request: httpx.Request) -> httpx.Response:
         import json

wordarr/metadata/hardcover.py

@@ -13,6 +13,7 @@ Series are one query - unlike Audible, Hardcover knows them.
 
 import logging
 import os
+from collections import Counter
 
 import httpx
 
@@ -61,6 +62,7 @@ query Series($id: Int!, $limit: Int!) {
         id
         title
         release_year
+        users_count
         cached_image
         contributions(limit: 3) { author { name } }
       }
@@ -204,10 +206,38 @@ async def search(query: str, language: str = "",
     return list(results.values())
 
 
+def _count(value) -> int:
+    try:
+        return int(str(value))
+    except (TypeError, ValueError):
+        return 0
+
+
+def _author_set(book: dict) -> set[str]:
+    return {name.strip().lower()
+            for name in _authors(book).split(", ") if name.strip()}
+
+
+def _main_author(authors: dict[int, set[str]]) -> str:
+    """The name behind the series, or "" when the rows do not agree on one.
+
+    Translators and publishers ride along in the contributions, so the author of
+    the series is simply the name most rows share - and a row without it is the
+    foreign book that someone filed under this position by mistake.
+    """
+    tally = Counter(name for names in authors.values() for name in names)
+    if not tally:
+        return ""
+    name, shared = tally.most_common(1)[0]
+    return name if shared * 2 > len(authors) else ""
+
+
 async def series_books(series_id: str, language: str = "") -> list[MetadataResult]:
     """Every volume of a series, by position. A position often holds several book
-    rows (the same story as another work); the one with an edition in the wanted
+    rows - other editions of the story, and now and then a foreign book someone
-    language wins the slot, else the first. Unnumbered books go last."""
+    filed here by mistake. The slot goes to the row that is by the author of the
+    series, has an edition in the wanted language and is the most read of those.
+    Unnumbered books go last."""
     if not TOKEN:
         return []
     try:
@@ -223,6 +253,8 @@ async def series_books(series_id: str, language: str = "") -> list[MetadataResul
         slots: dict[int, list[int]] = {}  # volume -> book rows filed under it
         loose: list[int] = []
         books: dict[int, MetadataResult] = {}
+        authors: dict[int, set[str]] = {}
+        readers: dict[int, int] = {}
         for entry in row.get("book_series") or []:
             book = entry.get("book") or {}
             try:
@@ -238,6 +270,8 @@ async def series_books(series_id: str, language: str = "") -> list[MetadataResul
                  "featured_series_position": volume},
                 "",
             )
+            authors[book_id] = _author_set(book)
+            readers[book_id] = _count(book.get("users_count"))
             if volume is None:
                 loose.append(book_id)
             else:
@@ -251,8 +285,14 @@ async def series_books(series_id: str, language: str = "") -> list[MetadataResul
                 log.warning("hardcover series editions failed for %s: %s",
                             series_id, exc)
 
+    author = _main_author(authors)
+
+    def rank(book_id: int) -> tuple:
+        return (author in authors[book_id] if author else True,
+                book_id in localized, readers[book_id])
+
     def pick(candidates: list[int]) -> MetadataResult:
-        best = next((i for i in candidates if i in localized), candidates[0])
+        best = max(candidates, key=rank)
         return localized.get(best) or books[best]
 
     return ([pick(slots[volume]) for volume in sorted(slots)]