Faster pdf export and test support

.gitea/workflows/tests.yml

@@ -41,7 +41,7 @@ jobs:
       id: pytest
       continue-on-error: true
       run: |
-        xvfb-run -a pytest --cov=pyPhotoAlbum --cov-report=xml --cov-report=json --cov-report=html --cov-report=term-missing
+        pytest --cov=pyPhotoAlbum --cov-report=xml --cov-report=json --cov-report=html --cov-report=term-missing
       env:
         QT_QPA_PLATFORM: offscreen
 

pyPhotoAlbum/pdf_exporter.py

@@ -5,9 +5,10 @@ Uses multiprocessing to pre-process images in parallel for faster exports.
 """
 
 import os
+import threading
 from typing import Any, List, Tuple, Optional, Union, Dict
 from dataclasses import dataclass, field
-from concurrent.futures import ProcessPoolExecutor, as_completed
+from concurrent.futures import ProcessPoolExecutor, ThreadPoolExecutor, as_completed
 import multiprocessing
 import io
 from reportlab.lib.pagesizes import A4
@@ -192,15 +193,15 @@ class PDFExporter:
         self.warnings: List[str] = []
         self.current_pdf_page = 1
         self.max_workers = max_workers or multiprocessing.cpu_count()
-        # Cache for pre-processed images: task_id -> PIL Image
         self._processed_images: Dict[str, Image.Image] = {}
 
     def export(self, output_path: str, progress_callback=None) -> Tuple[bool, List[str]]:
         """
         Export the project to PDF.
 
-        Uses multiprocessing to pre-process all images in parallel before
-        assembling the PDF sequentially.
+        Uses multiprocessing to pre-process all images in parallel, then renders
+        each page to its own PDF buffer in parallel (via threads), and finally
+        merges the per-page PDFs into the output file.
 
         Args:
             output_path: Path where PDF should be saved
@@ -210,7 +211,6 @@ class PDFExporter:
             Tuple of (success: bool, warnings: List[str])
         """
         self.warnings = []
-        self.current_pdf_page = 1
         self._processed_images = {}
 
         try:
@@ -221,70 +221,157 @@ class PDFExporter:
 
             # Get page dimensions from project (in mm)
             page_width_mm, page_height_mm = self.project.page_size_mm
-
-            # Bleed expands each page on all sides
             bleed_mm = self.project.page_bleed_mm
             bleed_pt = bleed_mm * self.MM_TO_POINTS
-
-            # Convert to PDF points (base page = cut/trim size)
             page_width_pt = page_width_mm * self.MM_TO_POINTS
             page_height_pt = page_height_mm * self.MM_TO_POINTS
 
-            # Expanded page size includes bleed on all sides
-            expanded_width_pt = page_width_pt + 2 * bleed_pt
-            expanded_height_pt = page_height_pt + 2 * bleed_pt
-
-            # Phase 1: Collect all image tasks and process in parallel
+            # Phase 1: parallel image pre-processing (unchanged)
             if progress_callback:
                 progress_callback(0, total_pages, "Collecting images for processing...")
-
             image_tasks = self._collect_image_tasks(page_width_pt, page_height_pt)
-
             if image_tasks:
                 if progress_callback:
                     progress_callback(0, total_pages, f"Processing {len(image_tasks)} images in parallel...")
                 self._preprocess_images_parallel(image_tasks, progress_callback, total_pages)
 
-            # Phase 2: Build PDF using pre-processed images
-            c = canvas.Canvas(output_path, pagesize=(expanded_width_pt, expanded_height_pt))
-
-            pages_processed = 0
-            for page in self.project.pages:
-                page_name = self.project.get_page_display_name(page)
+            # Phase 2: determine ordered page sequence (inserts blank pages for spread alignment)
+            page_sequence = self._compute_page_sequence()
 
+            # Phase 3: render each page to its own PDF bytes in parallel
+            n = len(page_sequence)
             if progress_callback:
-                    progress_callback(pages_processed, total_pages, f"Assembling {page_name}...")
+                progress_callback(0, total_pages, f"Rendering {n} pages in parallel...")
+            pdf_bytes_list: List[Optional[bytes]] = [None] * n
 
-                if page.is_cover:
-                    self._export_cover(c, page, page_width_pt, page_height_pt)
-                    pages_processed += 1
-                elif page.is_double_spread:
-                    if self.current_pdf_page % 2 == 1:
-                        c.showPage()
-                        self.current_pdf_page += 1
+            with ThreadPoolExecutor(max_workers=self.max_workers) as executor:
+                future_to_idx = {
+                    executor.submit(
+                        self._render_item_to_bytes, item, page_width_pt, page_height_pt, bleed_pt
+                    ): i
+                    for i, item in enumerate(page_sequence)
+                }
+                completed = 0
+                for future in as_completed(future_to_idx):
+                    i = future_to_idx[future]
+                    try:
+                        pdf_bytes_list[i] = future.result()
+                    except Exception as e:
+                        self.warnings.append(f"Error rendering page: {str(e)}")
+                        pdf_bytes_list[i] = self._make_blank_page_bytes(page_width_pt, page_height_pt, bleed_pt)
+                    completed += 1
                     if progress_callback:
-                            progress_callback(pages_processed, total_pages, "Inserting blank page for alignment...")
+                        progress_callback(completed, n, f"Rendering pages: {completed}/{n}...")
 
-                    self._export_spread(c, page, page_width_pt, page_height_pt, bleed_pt)
-                    pages_processed += 2
-                else:
-                    self._export_single_page(c, page, page_width_pt, page_height_pt, bleed_pt)
-                    pages_processed += 1
+            # Phase 4: merge all per-page PDFs into the output file
+            if progress_callback:
+                progress_callback(n, total_pages, "Merging pages...")
+            self._merge_page_pdfs(pdf_bytes_list, output_path)
 
-            c.save()
-
-            # Clean up processed images cache
             self._processed_images = {}
-
             if progress_callback:
                 progress_callback(total_pages, total_pages, "Export complete!")
-
             return True, self.warnings
 
         except Exception as e:
             self.warnings.append(f"Export failed: {str(e)}")
             return False, self.warnings
 
+    def _compute_page_sequence(self) -> List[Tuple[str, Any]]:
+        """
+        Build an ordered list of (page_type, page) items to render.
+
+        Inserts ('blank', None) entries before double-page spreads that would
+        otherwise start on an odd-numbered PDF page (spreads must start on even pages).
+        """
+        sequence: List[Tuple[str, Any]] = []
+        pdf_page_num = 1
+        for page in self.project.pages:
+            if page.is_cover:
+                sequence.append(("cover", page))
+                pdf_page_num += 1
+            elif page.is_double_spread:
+                if pdf_page_num % 2 == 1:
+                    sequence.append(("blank", None))
+                    pdf_page_num += 1
+                sequence.append(("spread", page))
+                pdf_page_num += 2
+            else:
+                sequence.append(("single", page))
+                pdf_page_num += 1
+        return sequence
+
+    def _render_item_to_bytes(
+        self,
+        item: Tuple[str, Any],
+        page_width_pt: float,
+        page_height_pt: float,
+        bleed_pt: float,
+    ) -> bytes:
+        """
+        Render a single page item to a self-contained PDF (as bytes).
+
+        Each call creates its own Canvas / BytesIO so pages can be rendered
+        concurrently without sharing state.
+        """
+        page_type, page = item
+        expanded_width_pt = page_width_pt + 2 * bleed_pt
+        expanded_height_pt = page_height_pt + 2 * bleed_pt
+        buf = io.BytesIO()
+
+        if page_type == "blank":
+            c = canvas.Canvas(buf, pagesize=(expanded_width_pt, expanded_height_pt))
+            c.showPage()
+            c.save()
+
+        elif page_type == "cover":
+            cover_width_mm, cover_height_mm = page.layout.size
+            cover_width_pt = cover_width_mm * self.MM_TO_POINTS
+            cover_height_pt = cover_height_mm * self.MM_TO_POINTS
+            c = canvas.Canvas(buf, pagesize=(cover_width_pt, cover_height_pt))
+            for element in sorted(page.layout.elements, key=lambda x: x.z_index):
+                self._render_element(c, element, 0, cover_width_pt, cover_height_pt, "Cover")
+            c.showPage()
+            c.save()
+
+        elif page_type == "single":
+            c = canvas.Canvas(buf, pagesize=(expanded_width_pt, expanded_height_pt))
+            for element in sorted(page.layout.elements, key=lambda x: x.z_index):
+                self._render_element(c, element, 0, page_width_pt, page_height_pt, page.page_number, bleed_pt)
+            c.showPage()
+            c.save()
+
+        elif page_type == "spread":
+            c = canvas.Canvas(buf, pagesize=(expanded_width_pt, expanded_height_pt))
+            self._export_spread(c, page, page_width_pt, page_height_pt, bleed_pt)
+            c.save()
+
+        return buf.getvalue()
+
+    def _make_blank_page_bytes(
+        self, page_width_pt: float, page_height_pt: float, bleed_pt: float
+    ) -> bytes:
+        """Return a minimal single-blank-page PDF for use as an error placeholder."""
+        buf = io.BytesIO()
+        c = canvas.Canvas(buf, pagesize=(page_width_pt + 2 * bleed_pt, page_height_pt + 2 * bleed_pt))
+        c.showPage()
+        c.save()
+        return buf.getvalue()
+
+    def _merge_page_pdfs(self, pdf_bytes_list: List[Optional[bytes]], output_path: str):
+        """Merge a list of single-page PDF byte strings into one output file."""
+        from pypdf import PdfWriter, PdfReader
+
+        writer = PdfWriter()
+        for pdf_bytes in pdf_bytes_list:
+            if pdf_bytes is None:
+                continue
+            reader = PdfReader(io.BytesIO(pdf_bytes))
+            for page in reader.pages:
+                writer.add_page(page)
+        with open(output_path, "wb") as f:
+            writer.write(f)
+
     def _make_task_id(
         self,
         element: ImageData,
@@ -530,10 +617,6 @@ class PDFExporter:
 
 
         c.showPage()  # Finish cover page
-        self.current_pdf_page += 1
-
-        # Reset page size for content pages
-        c.setPageSize((page_width_pt, page_height_pt))
 
     def _export_single_page(
         self, c: canvas.Canvas, page, page_width_pt: float, page_height_pt: float, bleed_pt: float = 0.0
@@ -548,7 +631,6 @@ class PDFExporter:
             self._render_element(c, element, 0, page_width_pt, page_height_pt, page.page_number, bleed_pt)
 
         c.showPage()  # Finish this page
-        self.current_pdf_page += 1
 
     def _export_spread(
         self, c: canvas.Canvas, page, page_width_pt: float, page_height_pt: float, bleed_pt: float = 0.0
@@ -596,7 +678,6 @@ class PDFExporter:
                 self._render_split_element(params, bleed_pt)
 
         c.showPage()  # Finish left page
-        self.current_pdf_page += 1
 
         # Process elements for right page
         c.setPageSize((expanded_width_pt, expanded_height_pt))
@@ -625,7 +706,6 @@ class PDFExporter:
                 self._render_split_element(params, bleed_pt)
 
         c.showPage()  # Finish right page
-        self.current_pdf_page += 1
 
     def _render_element(
         self,

pyproject.toml

@@ -31,6 +31,7 @@ dependencies = [
     "Pillow>=8.0.0",
     "reportlab>=3.5.0",
     "lxml>=4.6.0",
+    "pypdf>=4.0.0",
 ]
 
 [project.optional-dependencies]