Divers, en lien avec DMI04 (gemini_for_enonce.py, smaller images, link to enonce.pdf)

cutleft.py

@@ -26,7 +26,8 @@ if os.path.isfile(path_arg) and path_arg.lower().endswith('.pdf'):
     files = [os.path.basename(path_arg)]
 elif os.path.isdir(path_arg):
     INPUT_DIR = path_arg
-    files = sorted([f for f in os.listdir(INPUT_DIR) if f.lower().endswith('.pdf')])
+    files = sorted([f for f in os.listdir(INPUT_DIR) if f.lower().endswith('.pdf') and
+                    "nonc" not in f.lower()])
 else:
     sys.exit("Error: Input must be a directory or a PDF file.")
 
@@ -83,12 +84,20 @@ def stitch_images(image_list):
 
     return combined
 
+import threading
+pdf_cache_lock = threading.Lock()
+
+
 @lru_cache(maxsize=3)
-def get_pdf_pages(filename):
-    """Caches the heavy PDF rendering step for the current and next files."""
+def _get_pdf_pages_cached(filename):
     pdf_path = os.path.join(INPUT_DIR, filename)
     return convert_from_path(pdf_path)
 
+def get_pdf_pages(filename):
+    """Thread-safe wrapper for the cached PDF conversion."""
+    with pdf_cache_lock:
+        return _get_pdf_pages_cached(filename)
+
 def process_single_pdf(filename, shift_offset=0, max_per_file=5):
     """
     Converts PDF to stitched images.
@@ -137,7 +146,8 @@ def process_single_pdf(filename, shift_offset=0, max_per_file=5):
 
         # 3. Generate Preview (All stitched together, Resized)
         full_stitch = stitch_images(cropped_images)
-        preview_resized = full_stitch.resize(OUTPUT_SIZE, Image.LANCZOS)
+        # preview_resized = full_stitch.resize(OUTPUT_SIZE, Image.LANCZOS)
+        preview_resized = full_stitch.resize(OUTPUT_SIZE, Image.BILINEAR)
 
         schema = {
             "original_filename": filename,
@@ -200,8 +210,6 @@ class ImageReviewer:
         self.current_preview = None # Only stores the resized preview for GUI
         self.is_processing = False
 
-        # Queue for pre-fetched results (index, (preview, splits, schema))
-        self.prefetch_queue = Queue(maxsize=1)
         # Queue for manual re-processing results
         self.manual_queue = Queue()
 
@@ -244,19 +252,15 @@ class ImageReviewer:
         self.trigger_processing(self.files[self.index], self.current_shift)
 
     def prefetch_worker(self):
-        """Background thread to process the NEXT image constantly."""
-        idx_to_process = 0
+        """Background thread to load the NEXT file's PDF pages into RAM."""
+        idx_to_process = -1
         while True:
             target = self.index + 1
-            if target < len(self.files):
-                if idx_to_process != target:
-                    fname = self.files[target]
-                    result = process_single_pdf(fname, shift_offset=0)
-                    if result:
-                        self.prefetch_queue.put((target, result)) # Blocks if full
-                        idx_to_process = target
-
-            time.sleep(0.1)
+            if target < len(self.files) and target != idx_to_process:
+                fname = self.files[target]
+                get_pdf_pages(fname)  # Just calling it warms the lru_cache
+                idx_to_process = target
+            time.sleep(0.05)
 
     def load_current_image(self, use_prefetch=False):
         if self.index >= len(self.files):
@@ -266,21 +270,10 @@ class ImageReviewer:
 
         filename = self.files[self.index]
         self.is_processing = False
+        self.current_shift = 0
 
-        result_found = None
-
-        if use_prefetch and not self.prefetch_queue.empty():
-            q_idx, q_result = self.prefetch_queue.queue[0]
-            if q_idx == self.index:
-                _, result_found = self.prefetch_queue.get()
-                self.current_shift = 0
-                print(f"Loaded {filename} from prefetch.")
-
-        if result_found:
-            self.handle_processing_result(result_found, filename)
-        else:
-            # Not in queue (first load or queue mismatch), process manually
-            self.trigger_processing(filename, self.current_shift)
+        # Always trigger processing. If prefetched, get_pdf_pages returns instantly.
+        self.trigger_processing(filename, self.current_shift)
 
     def trigger_processing(self, filename, shift):
         """Starts a thread to process image so GUI doesn't freeze."""