Add pagination to category search for large categories

Large categories like "Living people" (900k+ articles) were impractical
because the code tried to download all members before displaying results.
Now stops after collecting ~200 articles and provides a "Next page" link.

Also fixes the MediaWiki API continuation protocol: passes the full
continue dict (not just gcmcontinue) so imcontinue responses are handled
properly, and reduces gcmlimit from "max" to 50 so each batch's images
fit in one API response.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

main.py

@@ -348,6 +348,14 @@ class ArticleWithoutImage:
         return f"/?enwp={quote(self.title)}"
 
 
+@dataclasses.dataclass
+class CategoryResult:
+    """Result of a paginated category search."""
+
+    articles: list[ArticleWithoutImage]
+    gcmcontinue: str | None
+
+
 # Common non-content images to ignore when checking if an article has images
 NON_CONTENT_IMAGE_PATTERNS = [
     "OOjs UI icon",
@@ -378,12 +386,16 @@ def has_content_image(images: list[dict]) -> bool:
     return False
 
 
-def get_articles_without_images(category: str) -> list[ArticleWithoutImage]:
+def get_articles_without_images(
+    category: str,
+    limit: int = 200,
+    gcmcontinue: str | None = None,
+) -> CategoryResult:
     """Get articles in a category that don't have images.
 
     Uses generator=categorymembers with prop=images to efficiently check
     multiple articles in a single API request, following continuation until
-    all category members have been processed.
+    the limit is reached or all category members have been processed.
     """
     params = {
         "action": "query",
@@ -391,20 +403,25 @@ def get_articles_without_images(category: str) -> list[ArticleWithoutImage]:
         "gcmtitle": category,
         "gcmtype": "page",  # Only articles, not subcategories or files
         "gcmnamespace": "0",  # Main namespace only
-        "gcmlimit": "max",
+        "gcmlimit": "50",  # Small batches so images fit in one response
         "prop": "images",
-        "imlimit": "max",  # Need enough to check all pages in batch
+        "imlimit": "max",
         "format": "json",
     }
 
     headers = {"User-Agent": WIKIMEDIA_USER_AGENT}
     articles_without_images: list[ArticleWithoutImage] = []
-    continue_token: str | None = None
+    seen_pageids: set[int] = set()
+    next_gcmcontinue: str | None = None
+
+    # Build initial continue params from the external pagination token
+    continue_params: dict[str, str] = {}
+    if gcmcontinue:
+        continue_params = {"gcmcontinue": gcmcontinue, "continue": "gcmcontinue||"}
 
     while True:
         request_params = params.copy()
-        if continue_token:
-            request_params["gcmcontinue"] = continue_token
+        request_params.update(continue_params)
 
         try:
             response = requests.get(
@@ -418,6 +435,11 @@ def get_articles_without_images(category: str) -> list[ArticleWithoutImage]:
 
         pages = data.get("query", {}).get("pages", {})
         for page in pages.values():
+            pageid = page.get("pageid", 0)
+            if not pageid or pageid in seen_pageids:
+                continue
+            seen_pageids.add(pageid)
+
             images = page.get("images", [])
 
             # Skip if page has content images (not just UI icons)
@@ -425,20 +447,29 @@ def get_articles_without_images(category: str) -> list[ArticleWithoutImage]:
                 continue
 
             title = page.get("title", "")
-            pageid = page.get("pageid", 0)
-
-            if title and pageid:
+            if title:
                 articles_without_images.append(
                     ArticleWithoutImage(title=title, pageid=pageid)
                 )
 
-        continue_token = data.get("continue", {}).get("gcmcontinue")
-        if not continue_token:
+        api_continue = data.get("continue")
+        if not api_continue:
             break
 
+        # Only stop at generator boundaries where we have a resumable token
+        gcmc = api_continue.get("gcmcontinue")
+        if gcmc and len(articles_without_images) >= limit:
+            next_gcmcontinue = gcmc
+            break
+
+        continue_params = api_continue
+
     # Sort by title for consistent display
     articles_without_images.sort(key=lambda a: a.title)
-    return articles_without_images
+    return CategoryResult(
+        articles=articles_without_images,
+        gcmcontinue=next_gcmcontinue,
+    )
 
 
 def is_valid_flickr_image_url(url: str) -> bool:
@@ -807,7 +838,8 @@ def category_search() -> str:
         )
 
     log_interaction("search_category", query=category)
-    articles = get_articles_without_images(category)
+    gcmcontinue = flask.request.args.get("gcmcontinue") or None
+    result = get_articles_without_images(category, gcmcontinue=gcmcontinue)
 
     # Get the display name (without Category: prefix)
     category_name = category.replace("Category:", "")
@@ -817,7 +849,8 @@ def category_search() -> str:
         cat=cat,
         category=category,
         category_name=category_name,
-        articles=articles,
+        articles=result.articles,
+        gcmcontinue=result.gcmcontinue,
     )
 
 

templates/category.html

@@ -44,6 +44,12 @@
         {% endfor %}
       </div>
 
+      {% if gcmcontinue %}
+      <div class="mt-3">
+        <a href="{{ url_for('category_search', cat=cat, gcmcontinue=gcmcontinue) }}" class="btn btn-outline-primary">Next page &raquo;</a>
+      </div>
+      {% endif %}
+
       {% else %}
       <div class="alert alert-success mt-3">
         All articles in this category have images!