BabelDOC 0.5.11__tar.gz → 0.5.13__tar.gz

{babeldoc-0.5.11 → babeldoc-0.5.13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: BabelDOC
-Version: 0.5.11
+Version: 0.5.13
 Summary: Yet Another Document Translator
 Project-URL: Homepage, https://github.com/funstory-ai/BabelDOC
 Project-URL: Issues, https://github.com/funstory-ai/BabelDOC/issues

babeldoc-0.5.13/babeldoc/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.5.13"

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/babeldoc_exception/BabelDOCException.py

@@ -11,3 +11,9 @@ class ExtractTextError(Exception):
 class InputFileGeneratedByBabelDOCError(Exception):
     def __init__(self, message):
         super().__init__(message)
+
+
+class ContentFilterError(Exception):
+    def __init__(self, message):
+        super().__init__(message)
+        self.message = message

babeldoc-0.5.13/babeldoc/const.py

@@ -0,0 +1,95 @@
+import itertools
+import multiprocessing as mp
+import os
+import shutil
+import subprocess
+import threading
+from pathlib import Path
+
+__version__ = "0.5.13"
+
+CACHE_FOLDER = Path.home() / ".cache" / "babeldoc"
+
+
+def get_cache_file_path(filename: str, sub_folder: str | None = None) -> Path:
+    if sub_folder is not None:
+        sub_folder = sub_folder.strip("/")
+        sub_folder_path = CACHE_FOLDER / sub_folder
+        sub_folder_path.mkdir(parents=True, exist_ok=True)
+        return sub_folder_path / filename
+    return CACHE_FOLDER / filename
+
+
+try:
+    git_path = shutil.which("git")
+    if git_path is None:
+        raise FileNotFoundError("git executable not found")
+    two_parent = Path(__file__).resolve().parent.parent
+    md_ = two_parent / "docs" / "README.md"
+    if two_parent.name == "site-packages" or not md_.exists():
+        raise FileNotFoundError("not in git repo")
+    WATERMARK_VERSION = (
+        subprocess.check_output(  # noqa: S603
+            [git_path, "describe", "--always"],
+            cwd=Path(__file__).resolve().parent,
+        )
+        .strip()
+        .decode()
+    )
+except (OSError, FileNotFoundError, subprocess.CalledProcessError):
+    WATERMARK_VERSION = f"v{__version__}"
+
+TIKTOKEN_CACHE_FOLDER = CACHE_FOLDER / "tiktoken"
+TIKTOKEN_CACHE_FOLDER.mkdir(parents=True, exist_ok=True)
+os.environ["TIKTOKEN_CACHE_DIR"] = str(TIKTOKEN_CACHE_FOLDER)
+
+
+_process_pool = None
+_process_pool_lock = threading.Lock()
+_ENABLE_PROCESS_POOL = False
+
+
+def enable_process_pool():
+    # Development and Testing ONLY API
+    global _ENABLE_PROCESS_POOL
+    _ENABLE_PROCESS_POOL = True
+
+
+# macos & windows use spawn mode
+# linux use forkserver mode
+
+
+def get_process_pool():
+    if not _ENABLE_PROCESS_POOL:
+        return None
+    global _process_pool
+    with _process_pool_lock:
+        if _process_pool is None:
+            # Create pool only in main process
+            if mp.current_process().name != "MainProcess":
+                return None
+
+            _process_pool = mp.Pool()
+        return _process_pool
+
+
+def close_process_pool():
+    if not _ENABLE_PROCESS_POOL:
+        return None
+    global _process_pool
+    with _process_pool_lock:
+        if _process_pool:
+            _process_pool.close()
+            _process_pool.join()
+            _process_pool = None
+
+
+def batched(iterable, n, *, strict=False):
+    # batched('ABCDEFG', 3) → ABC DEF G
+    if n < 1:
+        raise ValueError("n must be at least one")
+    iterator = iter(iterable)
+    while batch := tuple(itertools.islice(iterator, n)):
+        if strict and len(batch) != n:
+            raise ValueError("batched(): incomplete batch")
+        yield batch

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/docvision/rpc_doclayout6.py

@@ -28,7 +28,9 @@ from babeldoc.format.pdf.document_il.utils.extract_char import (
 )
 from babeldoc.format.pdf.document_il.utils.fontmap import FontMapper
 from babeldoc.format.pdf.document_il.utils.layout_helper import SPACE_REGEX
-from babeldoc.format.pdf.document_il.utils.mupdf_helper import get_no_rotation_img
+from babeldoc.format.pdf.document_il.utils.mupdf_helper import (
+    get_no_rotation_img_multiprocess,
+)
 
 logger = logging.getLogger(__name__)
 DPI = 150
@@ -141,7 +143,7 @@ def predict_layout(
         f"{host}/inference",
         json=request_data,
         headers={"Accept": "application/json", "Content-Type": "application/json"},
-        timeout=1800,
+        timeout=30,
         follow_redirects=True,
     )
 
@@ -233,7 +235,7 @@ def predict_layout2(
             "Content-Type": "application/msgpack",
             "Accept": "application/msgpack",
         },
-        timeout=480,
+        timeout=30,
         follow_redirects=True,
     )
 
@@ -569,18 +571,19 @@ class RpcDocLayoutModel(DocLayoutModel):
 
         return results
 
-    def predict_page(
-        self, page, mupdf_doc: pymupdf.Document, translate_config, save_debug_image
-    ):
+    def predict_page(self, page, pdf_bytes: Path, translate_config, save_debug_image):
         translate_config.raise_if_cancelled()
-        with self.lock:
-            # pix = mupdf_doc[page.page_number].get_pixmap(dpi=72)
-            pix = get_no_rotation_img(mupdf_doc[page.page_number], dpi=DPI)
-        image = np.frombuffer(pix.samples, np.uint8).reshape(
-            pix.height,
-            pix.width,
-            3,
-        )[:, :, ::-1]
+        # doc = pymupdf.open(io.BytesIO(pdf_bytes))
+        # with self.lock:
+        # pix = mupdf_doc[page.page_number].get_pixmap(dpi=72)
+        image = get_no_rotation_img_multiprocess(
+            pdf_bytes.as_posix(), page.page_number, dpi=DPI
+        )
+        # image = np.frombuffer(pix.samples, np.uint8).reshape(
+        #     pix.height,
+        #     pix.width,
+        #     3,
+        # )[:, :, ::-1]
         char_boxes = convert_page_to_char_boxes(page)
         lines = process_page_chars_to_lines(char_boxes)
         predict_result = self.predict_image(image, 800, lines)
@@ -594,11 +597,13 @@ class RpcDocLayoutModel(DocLayoutModel):
         translate_config,
         save_debug_image,
     ):
-        with ThreadPoolExecutor(max_workers=32) as executor:
+        layout_temp_path = translate_config.get_working_file_path("layout.temp.pdf")
+        mupdf_doc.save(layout_temp_path.as_posix())
+        with ThreadPoolExecutor(max_workers=8) as executor:
             yield from executor.map(
                 self.predict_page,
                 pages,
-                (mupdf_doc for _ in range(len(pages))),
+                (layout_temp_path for _ in range(len(pages))),
                 (translate_config for _ in range(len(pages))),
                 (save_debug_image for _ in range(len(pages))),
             )

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/frontend/il_creater.py

@@ -955,7 +955,7 @@ class ILCreater:
             )
             return
 
-        if char_bounding_box:
+        if char_bounding_box and len(char_bounding_box) == 4:
             x_min, y_min, x_max, y_max = char_bounding_box
             factor = 1 / 1000 * pdf_style.font_size
             x_min = x_min * factor

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/midend/detect_scanned_file.py

@@ -9,6 +9,7 @@ from skimage.metrics import structural_similarity
 from babeldoc.babeldoc_exception.BabelDOCException import ScannedPDFError
 from babeldoc.format.pdf.document_il import il_version_1
 from babeldoc.format.pdf.document_il.backend.pdf_creater import PDFCreater
+from babeldoc.format.pdf.document_il.utils.style_helper import BLACK
 from babeldoc.format.pdf.document_il.utils.style_helper import GREEN
 from babeldoc.format.pdf.translation_config import TranslationConfig
 
@@ -143,6 +144,8 @@ class DetectScannedFile:
         for page in docs.page:
             for char in page.pdf_character:
                 char.render_order = None
+                if not char.debug_info:
+                    char.pdf_style.graphic_state = BLACK
 
     def detect_page_is_scanned(
         self, page: il_version_1.Page, pdf: pymupdf.Document, pdf_creater: PDFCreater

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/midend/il_translator.py

@@ -4,12 +4,16 @@ import copy
 import json
 import logging
 import re
+import threading
 from pathlib import Path
 
 import tiktoken
 from tqdm import tqdm
 
+import babeldoc.format.pdf.document_il.il_version_1 as il_version_1
+from babeldoc.babeldoc_exception.BabelDOCException import ContentFilterError
 from babeldoc.format.pdf.document_il import Document
+from babeldoc.format.pdf.document_il import GraphicState
 from babeldoc.format.pdf.document_il import Page
 from babeldoc.format.pdf.document_il import PdfFont
 from babeldoc.format.pdf.document_il import PdfFormula
@@ -34,6 +38,7 @@ from babeldoc.format.pdf.document_il.utils.paragraph_helper import (
 from babeldoc.format.pdf.document_il.utils.paragraph_helper import (
     is_pure_numeric_paragraph,
 )
+from babeldoc.format.pdf.document_il.utils.style_helper import GRAY80
 from babeldoc.format.pdf.translation_config import TranslationConfig
 from babeldoc.translator.translator import BaseTranslator
 from babeldoc.utils.priority_thread_pool_executor import PriorityThreadPoolExecutor
@@ -307,6 +312,8 @@ class ILTranslator:
             self.support_llm_translate = False
 
         self.use_as_fallback = False
+        self.add_content_filter_hint_lock = threading.Lock()
+        self.docs = None
 
     def calc_token_count(self, text: str) -> int:
         try:
@@ -315,6 +322,7 @@ class ILTranslator:
             return 0
 
     def translate(self, docs: Document):
+        self.docs = docs
         tracker = DocumentTranslateTracker()
 
         if not self.translation_config.shared_context_cross_split_part.first_paragraph:
@@ -390,6 +398,7 @@ class ILTranslator:
             executor.submit(
                 self.translate_paragraph,
                 paragraph,
+                page,
                 pbar,
                 tracker.new_paragraph(),
                 page_font_map,
@@ -985,9 +994,58 @@ Now, please carefully read the following text to be translated and directly outp
 
         return final_input
 
+    def add_content_filter_hint(self, page: Page, paragraph: PdfParagraph):
+        with self.add_content_filter_hint_lock:
+            new_box = il_version_1.Box(
+                x=paragraph.box.x,
+                y=paragraph.box.y2,
+                x2=paragraph.box.x2,
+                y2=paragraph.box.y2 + 1.1,
+            )
+            page.pdf_paragraph.append(
+                self._create_text(
+                    "翻译服务检测到内容可能包含不安全或敏感内容，请您避免翻译敏感内容，感谢您的配合。",
+                    GRAY80,
+                    new_box,
+                    1,
+                )
+            )
+            logger.info("success add content filter hint")
+
+    def _create_text(
+        self,
+        text: str,
+        color: GraphicState,
+        box: il_version_1.Box,
+        font_size: float = 4,
+    ):
+        style = il_version_1.PdfStyle(
+            font_id="base",
+            font_size=font_size,
+            graphic_state=color,
+        )
+        return il_version_1.PdfParagraph(
+            first_line_indent=False,
+            box=box,
+            vertical=False,
+            pdf_style=style,
+            unicode=text,
+            pdf_paragraph_composition=[
+                il_version_1.PdfParagraphComposition(
+                    pdf_same_style_unicode_characters=il_version_1.PdfSameStyleUnicodeCharacters(
+                        unicode=text,
+                        pdf_style=style,
+                        debug_info=True,
+                    ),
+                ),
+            ],
+            xobj_id=-1,
+        )
+
     def translate_paragraph(
         self,
         paragraph: PdfParagraph,
+        page: Page,
         pbar: tqdm | None = None,
         tracker: ParagraphTranslateTracker = None,
         page_font_map: dict[str, PdfFont] = None,
@@ -1039,6 +1097,10 @@ Now, please carefully read the following text to be translated and directly outp
                 self.post_translate_paragraph(
                     paragraph, tracker, translate_input, translated_text
                 )
+            except ContentFilterError as e:
+                logger.warning(f"ContentFilterError: {e.message}")
+                self.add_content_filter_hint(page, paragraph)
+                return
             except Exception as e:
                 logger.exception(
                     f"Error translating paragraph. Paragraph: {paragraph.debug_id} ({paragraph.unicode}). Error: {e}. ",

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/midend/il_translator_llm_only.py

@@ -40,9 +40,11 @@ class BatchParagraph:
     def __init__(
         self,
         paragraphs: list[PdfParagraph],
+        pages: list[Page],
         page_tracker: PageTranslateTracker,
     ):
         self.paragraphs = paragraphs
+        self.pages = pages
         self.trackers = [page_tracker.new_paragraph() for _ in paragraphs]
 
 
@@ -112,6 +114,7 @@ class ILTranslatorLLMOnly:
         return None
 
     def translate(self, docs: Document) -> None:
+        self.il_translator.docs = docs
         tracker = DocumentTranslateTracker()
         self.mid = 0
 
@@ -361,8 +364,9 @@ class ILTranslatorLLMOnly:
 
             # Create batch with both paragraphs
             cross_page_paragraphs = [last_curr_paragraph, first_next_paragraph]
+            cross_page_pages = [page_curr, page_next]
             batch_paragraph = BatchParagraph(
-                cross_page_paragraphs, tracker.new_cross_page()
+                cross_page_paragraphs, cross_page_pages, tracker.new_cross_page()
             )
 
             self.mid += 1
@@ -438,7 +442,7 @@ class ILTranslatorLLMOnly:
                 p1.unicode
             ) + self.calc_token_count(p2.unicode)
 
-            batch = BatchParagraph([p1, p2], tracker.new_cross_column())
+            batch = BatchParagraph([p1, p2], [page, page], tracker.new_cross_column())
             self.mid += 1
             executor.submit(
                 self.translate_paragraph,
@@ -523,7 +527,7 @@ class ILTranslatorLLMOnly:
                 self.mid += 1
                 executor.submit(
                     self.translate_paragraph,
-                    BatchParagraph(paragraphs, tracker),
+                    BatchParagraph(paragraphs, [page] * len(paragraphs), tracker),
                     pbar,
                     page_font_map,
                     page_xobj_font_map,
@@ -541,7 +545,7 @@ class ILTranslatorLLMOnly:
             self.mid += 1
             executor.submit(
                 self.translate_paragraph,
-                BatchParagraph(paragraphs, tracker),
+                BatchParagraph(paragraphs, [page] * len(paragraphs), tracker),
                 pbar,
                 page_font_map,
                 page_xobj_font_map,
@@ -630,6 +634,9 @@ class ILTranslatorLLMOnly:
             llm_prompt_parts.append("#role")
             if self.translation_config.custom_system_prompt:
                 llm_prompt_parts.append(self.translation_config.custom_system_prompt)
+                llm_prompt_parts.append(
+                    "When translating, strictly follow the instructions below to ensure translation quality and preserve all formatting, tags, and placeholders:\n"
+                )
             else:
                 llm_prompt_parts.append(
                     f"You are a professional and reliable machine translation engine responsible for translating the input text into {self.translation_config.lang_out}.\n"
@@ -894,6 +901,7 @@ class ILTranslatorLLMOnly:
                         executor.submit(
                             self.il_translator.translate_paragraph,
                             inputs[id_][2],
+                            batch_paragraph.pages[id_],
                             pbar,
                             inputs[id_][3],
                             page_font_map,
@@ -929,6 +937,7 @@ class ILTranslatorLLMOnly:
                 executor.submit(
                     self.il_translator.translate_paragraph,
                     paragraph,
+                    batch_paragraph.pages[i],
                     pbar,
                     tracker,
                     page_font_map,

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/midend/layout_parser.py

@@ -1,4 +1,7 @@
 import logging
+import math
+import os
+from concurrent.futures import ThreadPoolExecutor
 from pathlib import Path
 
 import cv2
@@ -7,7 +10,6 @@ from pymupdf import Document
 
 import babeldoc.format.pdf.document_il.utils.extract_char
 from babeldoc.format.pdf.document_il import il_version_1
-from babeldoc.format.pdf.document_il.utils.mupdf_helper import get_no_rotation_img
 from babeldoc.format.pdf.document_il.utils.style_helper import GREEN
 from babeldoc.format.pdf.translation_config import TranslationConfig
 
@@ -120,7 +122,7 @@ class LayoutParser:
         total = len(docs.page)
         with self.translation_config.progress_monitor.stage_start(
             self.stage_name,
-            total,
+            total * 2,
         ) as progress:
             # Process predictions for each page
             for page, layouts in self.model.handle_document(
@@ -134,9 +136,15 @@ class LayoutParser:
                     # Convert coordinate system from picture to il
                     # system to the il coordinate system
                     x0, y0, x1, y1 = layout.xyxy
-                    pix = get_no_rotation_img(mupdf_doc[page.page_number])
+                    # pix = get_no_rotation_img(mupdf_doc[page.page_number])
                     # pix = mupdf_doc[page.page_number].get_pixmap()
-                    h, w = pix.height, pix.width
+                    # h, w = pix.height, pix.width
+                    box = mupdf_doc[page.page_number].mediabox_size
+                    b_h = math.ceil(box.y)
+                    b_w = math.ceil(box.x)
+                    # if b_h != h or b_w != w:
+                    #     logger.warning(f"page {page.page_number} mediabox is not correct, b_h: {b_h}, h: {h}, b_w: {b_w}, w: {w}")
+                    h, w = b_h, b_w
                     x0, y0, x1, y1 = (
                         np.clip(int(x0 - 1), 0, w - 1),
                         np.clip(int(h - y1 - 1), 0, h - 1),
@@ -157,39 +165,47 @@ class LayoutParser:
                     page_layouts.append(page_layout)
 
                 page.page_layout = page_layouts
-                self.generate_fallback_line_layout_for_page(page)
-                self._save_debug_box_to_page(page)
+                # self.generate_fallback_line_layout_for_page(page)
+                # self._save_debug_box_to_page(page)
                 progress.advance(1)
-
+            with ThreadPoolExecutor(max_workers=os.cpu_count()) as executor:
+                for page in docs.page:
+                    executor.submit(
+                        self.generate_fallback_line_layout_for_page, page, progress
+                    )
         return docs
 
-    def generate_fallback_line_layout_for_page(self, page: il_version_1.Page):
-        exists_page_layouts = page.page_layout
-        char_boxes = babeldoc.format.pdf.document_il.utils.extract_char.convert_page_to_char_boxes(
-            page
-        )
-        if not char_boxes:
-            return
+    def generate_fallback_line_layout_for_page(self, page: il_version_1.Page, progress):
+        try:
+            exists_page_layouts = page.page_layout
+            char_boxes = babeldoc.format.pdf.document_il.utils.extract_char.convert_page_to_char_boxes(
+                page
+            )
+            if not char_boxes:
+                return
 
-        clusters = babeldoc.format.pdf.document_il.utils.extract_char.process_page_chars_to_lines(
-            char_boxes
-        )
-        for cluster in clusters:
-            boxes = [c[0] for c in cluster.chars]
-            min_x = min(b.x for b in boxes)
-            max_x = max(b.x2 for b in boxes)
-            min_y = min(b.y for b in boxes)
-            max_y = max(b.y2 for b in boxes)
-            cluster.chars = il_version_1.Box(min_x, min_y, max_x, max_y)
-            page_layout = il_version_1.PageLayout(
-                id=len(exists_page_layouts) + 1,
-                box=il_version_1.Box(
-                    min_x,
-                    min_y,
-                    max_x,
-                    max_y,
-                ),
-                conf=1,
-                class_name="fallback_line",
+            clusters = babeldoc.format.pdf.document_il.utils.extract_char.process_page_chars_to_lines(
+                char_boxes
             )
-            exists_page_layouts.append(page_layout)
+            for cluster in clusters:
+                boxes = [c[0] for c in cluster.chars]
+                min_x = min(b.x for b in boxes)
+                max_x = max(b.x2 for b in boxes)
+                min_y = min(b.y for b in boxes)
+                max_y = max(b.y2 for b in boxes)
+                cluster.chars = il_version_1.Box(min_x, min_y, max_x, max_y)
+                page_layout = il_version_1.PageLayout(
+                    id=len(exists_page_layouts) + 1,
+                    box=il_version_1.Box(
+                        min_x,
+                        min_y,
+                        max_x,
+                        max_y,
+                    ),
+                    conf=1,
+                    class_name="fallback_line",
+                )
+                exists_page_layouts.append(page_layout)
+            self._save_debug_box_to_page(page)
+        finally:
+            progress.advance(1)

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/midend/styles_and_formulas.py

@@ -462,11 +462,17 @@ class StylesAndFormulas:
                 line.pdf_character[i + 1] if i < len(line.pdf_character) - 1 else None
             )
             isspace = char.char_unicode.isspace() if char.char_unicode else False
+            prev_is_space = (
+                previous_char.char_unicode.isspace()
+                if previous_char and previous_char.char_unicode
+                else False
+            )
 
             is_corner_mark = (
                 (
                     previous_char is not None
                     and not isspace
+                    and not prev_is_space
                     and not first_is_bullet
                     # 角标字体，有 0.76 的角标和 0.799 的大写，这里用 0.79 取中，同时考虑首字母放大的情况
                     and char.pdf_style.font_size
@@ -476,6 +482,7 @@ class StylesAndFormulas:
                 or (
                     previous_char is not None
                     and not isspace
+                    and not prev_is_space
                     and not first_is_bullet
                     # 角标字体，有 0.76 的角标和 0.799 的大写，这里用 0.79 取中，同时考虑首字母放大的情况
                     and char.pdf_style.font_size
@@ -487,6 +494,7 @@ class StylesAndFormulas:
                     previous_char is None
                     and next_char is not None
                     and not isspace
+                    and not prev_is_space
                     and not first_is_bullet
                     # 当前字符字体大小明显小于下一个字符，判定为角标
                     and char.pdf_style.font_size < next_char.pdf_style.font_size * 0.79

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/utils/extract_char.py

@@ -11,6 +11,7 @@ from sklearn.cluster import DBSCAN
 
 import babeldoc.format.pdf.high_level
 import babeldoc.format.pdf.translation_config
+from babeldoc.const import get_process_pool
 from babeldoc.format.pdf.document_il import il_version_1
 
 logger = logging.getLogger(__name__)
@@ -571,6 +572,15 @@ def _merge_lines_on_page(page_lines: list[Line]) -> list[Line]:
 
 def process_page_chars_to_lines(
     chars: list[tuple[il_version_1.Box, str, bool]],
+) -> list[Line]:
+    pool = get_process_pool()
+    if pool is None:
+        return process_page_chars_to_lines_internal(chars)
+    return pool.apply(process_page_chars_to_lines_internal, (chars,))
+
+
+def process_page_chars_to_lines_internal(
+    chars: list[tuple[il_version_1.Box, str, bool]],
 ) -> list[Line]:
     """
     Process characters on a single page to cluster them into lines.

babeldoc-0.5.13/babeldoc/format/pdf/document_il/utils/mupdf_helper.py

@@ -0,0 +1,42 @@
+import numpy as np
+import pymupdf
+
+from babeldoc.const import get_process_pool
+
+
+def get_no_rotation_img(page: pymupdf.Page, dpi: int = 72) -> pymupdf.Pixmap:
+    # return page.get_pixmap(dpi=72)
+    original_rotation = page.rotation
+    page.set_rotation(0)
+    pix = page.get_pixmap(dpi=dpi)
+    page.set_rotation(original_rotation)
+    return pix
+
+
+def get_no_rotation_img_multiprocess_internal(
+    pdf_bytes: str, pagenum: int, dpi: int = 72
+) -> np.ndarray:
+    # return page.get_pixmap(dpi=72)
+    doc = pymupdf.open(pdf_bytes)
+    try:
+        page = doc[pagenum]
+        original_rotation = page.rotation
+        page.set_rotation(0)
+        pix = page.get_pixmap(dpi=dpi)
+        page.set_rotation(original_rotation)
+        return np.frombuffer(pix.samples, np.uint8).reshape(
+            pix.height,
+            pix.width,
+            3,
+        )[:, :, ::-1]
+    finally:
+        doc.close()
+
+
+def get_no_rotation_img_multiprocess(pdf_bytes: str, pagenum: int, dpi: int = 72):
+    pool = get_process_pool()
+    if pool is None:
+        return get_no_rotation_img_multiprocess_internal(pdf_bytes, pagenum, dpi)
+    return pool.apply(
+        get_no_rotation_img_multiprocess_internal, (pdf_bytes, pagenum, dpi)
+    )

{babeldoc-0.5.11 → babeldoc-0.5.13}/babeldoc/format/pdf/document_il/utils/style_helper.py

@@ -29,6 +29,10 @@ BLACK = il_version_1.GraphicState(passthrough_per_char_instruction="0 g 0 G")
 
 WHITE = il_version_1.GraphicState(passthrough_per_char_instruction="1 g 1 G")
 
+GRAY80 = il_version_1.GraphicState(passthrough_per_char_instruction="0.80 g 0.80 G")
+GRAY67 = il_version_1.GraphicState(passthrough_per_char_instruction="0.67 g 0.67 G")
+GRAY33 = il_version_1.GraphicState(passthrough_per_char_instruction="0.33 g 0.33 G")
+
 # Generate all color styles
 RED = il_version_1.GraphicState(
     passthrough_per_char_instruction="1.0000000000 0.2313725490 0.1882352941 rg "