BabelDOC 0.4.2__tar.gz → 0.4.4__tar.gz

{babeldoc-0.4.2 → babeldoc-0.4.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: BabelDOC
-Version: 0.4.2
+Version: 0.4.4
 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-0.4.4/babeldoc/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.4.4"

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/const.py

@@ -3,7 +3,7 @@ import shutil
 import subprocess
 from pathlib import Path
 
-__version__ = "0.4.2"
+__version__ = "0.4.4"
 
 CACHE_FOLDER = Path.home() / ".cache" / "babeldoc"
 

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/docvision/rpc_doclayout2.py

@@ -286,7 +286,7 @@ class RpcDocLayoutModel(DocLayoutModel):
         translate_config,
         save_debug_image,
     ):
-        with ThreadPoolExecutor(max_workers=64) as executor:
+        with ThreadPoolExecutor(max_workers=128) as executor:
             yield from executor.map(
                 self.predict_page,
                 pages,

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/format/pdf/document_il/backend/pdf_creater.py

@@ -331,7 +331,7 @@ class PDFCreater:
         self,
         draw_op: BitStream,
         rectangle: il_version_1.PdfRectangle,
-        line_width: float = 0.5,
+        line_width: float = 0.4,
     ):
         """Draw a rectangle in PDF for visualization purposes.
 

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/format/pdf/document_il/frontend/il_creater.py

@@ -656,7 +656,7 @@ class ILCreater:
                 _, encoding = self.mupdf.xref_get_key(xref_id, "Encoding")
                 if encoding == "/Identity-H" or encoding == "/Identity-V":
                     encoding_length = 2
-                if encoding == "/WinAnsiEncoding":
+                elif encoding == "/WinAnsiEncoding":
                     encoding_length = 1
                 else:
                     _, to_unicode_id = self.mupdf.xref_get_key(xref_id, "ToUnicode")

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/format/pdf/document_il/midend/add_debug_information.py

@@ -3,7 +3,6 @@ import logging
 import babeldoc.format.pdf.document_il.il_version_1 as il_version_1
 from babeldoc.format.pdf.document_il import GraphicState
 from babeldoc.format.pdf.document_il.utils.style_helper import BLUE
-from babeldoc.format.pdf.document_il.utils.style_helper import CYAN
 from babeldoc.format.pdf.document_il.utils.style_helper import ORANGE
 from babeldoc.format.pdf.document_il.utils.style_helper import TEAL
 from babeldoc.format.pdf.document_il.utils.style_helper import YELLOW
@@ -133,9 +132,9 @@ class AddDebugInformation:
                                 char.visual_bbox.box, TEAL, line_width=0.2
                             ),
                         )
-                        page.pdf_rectangle.append(
-                            self._create_rectangle(char.box, CYAN, line_width=0.2),
-                        )
+                        # page.pdf_rectangle.append(
+                        #     self._create_rectangle(char.box, CYAN, line_width=0.2),
+                        # )
 
             for xobj in page.pdf_xobject:
                 new_paragraphs.append(

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/format/pdf/document_il/midend/styles_and_formulas.py

@@ -23,6 +23,7 @@ from babeldoc.format.pdf.document_il.utils.formular_helper import update_formula
 from babeldoc.format.pdf.document_il.utils.layout_helper import LEFT_BRACKET
 from babeldoc.format.pdf.document_il.utils.layout_helper import RIGHT_BRACKET
 from babeldoc.format.pdf.document_il.utils.layout_helper import build_layout_index
+from babeldoc.format.pdf.document_il.utils.layout_helper import calculate_iou_for_boxes
 from babeldoc.format.pdf.document_il.utils.layout_helper import (
     calculate_y_iou_for_boxes,
 )
@@ -66,6 +67,7 @@ class StylesAndFormulas:
         self.merge_overlapping_formulas(page)
         # self.process_page_offsets(page)
         self.process_translatable_formulas(page)
+        self.update_all_formula_data(page)
         self.process_page_offsets(page)
         self.update_all_formula_data(page)
         self.process_page_styles(page)
@@ -476,6 +478,8 @@ class StylesAndFormulas:
             return
 
         for paragraph in page.pdf_paragraph:
+            if paragraph.debug_id is None:
+                continue
             if not paragraph.pdf_paragraph_composition:
                 continue
 
@@ -499,22 +503,28 @@ class StylesAndFormulas:
                     comp = paragraph.pdf_paragraph_composition[j]
                     if comp.pdf_line:
                         for char in reversed(comp.pdf_line.pdf_character):
+                            if not char.pdf_character_id:
+                                continue
                             # 检查 y 坐标是否接近，判断是否在同一行
                             left_iou = calculate_y_iou_for_boxes(char.box, formula.box)
                             if left_iou > 0.6:
                                 left_char = char
                                 break
+                    break
 
                 # 查找右边最近的同一行的文本
                 for j in range(i + 1, len(paragraph.pdf_paragraph_composition)):
                     comp = paragraph.pdf_paragraph_composition[j]
                     if comp.pdf_line:
                         for char in comp.pdf_line.pdf_character:
+                            if not char.pdf_character_id:
+                                continue
                             # 检查 y 坐标是否接近，判断是否在同一行
                             right_iou = calculate_y_iou_for_boxes(char.box, formula.box)
                             if right_iou > 0.6:
                                 right_char = char
                                 break
+                    break
 
                 # If both text segments exist, keep the one with higher IOU
                 if left_char and right_char:
@@ -600,6 +610,9 @@ class StylesAndFormulas:
 
     def is_translatable_formula(self, formula: PdfFormula) -> bool:
         """判断公式是否只包含需要正常翻译的字符（数字、空格和英文逗号）"""
+        if all(char.formula_layout_id for char in formula.pdf_character):
+            return False
+
         text = "".join(char.char_unicode for char in formula.pdf_character)
         if formula.y_offset > 0.1:
             return False
@@ -710,8 +723,11 @@ class StylesAndFormulas:
 
     def merge_overlapping_formulas(self, page: Page):
         """
-        合并 x 轴重叠且 y 轴有交集的相邻公式，或者 x 轴相邻且 y 轴 IOU > 0.5 的公式，
-        或者所有字符的 layout id 都相同的公式
+        合并符合以下条件的公式：
+        1. x 轴重叠且 y 轴有交集的相邻公式，或者
+        2. x 轴相邻且 y 轴 IOU > 0.5 的相邻公式，或者
+        3. 所有字符的 layout id 都相同的相邻公式，或者
+        4. 任意两个公式的 IOU > 0.8
         角标可能会被识别成单独的公式，需要合并
         """
         if not page.pdf_paragraph:
@@ -721,46 +737,76 @@ class StylesAndFormulas:
             if not paragraph.pdf_paragraph_composition:
                 continue
 
-            i = 0
-            while i < len(paragraph.pdf_paragraph_composition) - 1:
-                comp1 = paragraph.pdf_paragraph_composition[i]
-                comp2 = paragraph.pdf_paragraph_composition[i + 1]
-
-                # 检查是否都是公式
-                if comp1.pdf_formula is None or comp2.pdf_formula is None:
-                    i += 1
-                    continue
-
-                formula1 = comp1.pdf_formula
-                formula2 = comp2.pdf_formula
-
-                # 检查合并条件：
-                # 1. x 轴重叠且 y 轴有交集，或者
-                # 2. x 轴相邻且 y 轴 IOU > 0.5，或者
-                # 3. 所有字符的 layout id 都相同
-                should_merge = (
-                    (
-                        self.is_x_axis_contained(formula1.box, formula2.box)
-                        and self.has_y_intersection(formula1.box, formula2.box)
-                    )
-                    or (
-                        self.is_x_axis_adjacent(formula1.box, formula2.box)
-                        and self.calculate_y_iou(formula1.box, formula2.box) > 0.5
-                    )
-                    or (self._have_same_layout_ids(formula1, formula2, page))
-                )
+            # 重复执行合并过程，直到没有更多可以合并的公式
+            merged = True
+            while merged:
+                merged = False
+                for i in range(len(paragraph.pdf_paragraph_composition)):
+                    if merged:
+                        break
+                    comp1 = paragraph.pdf_paragraph_composition[i]
+                    if comp1.pdf_formula is None:
+                        continue
+
+                    for j in range(i + 1, len(paragraph.pdf_paragraph_composition)):
+                        comp2 = paragraph.pdf_paragraph_composition[j]
+                        if comp2.pdf_formula is None:
+                            continue
+
+                        formula1 = comp1.pdf_formula
+                        formula2 = comp2.pdf_formula
+
+                        # 检查合并条件：
+                        # 1. x 轴重叠且 y 轴有交集，或者
+                        # 2. x 轴相邻且 y 轴 IOU > 0.5，或者
+                        # 3. 所有字符的 layout id 都相同，或者
+                        # 4. 任意两个公式的 IOU > 0.8
+                        should_merge = (
+                            (
+                                j == i + 1
+                                and (
+                                    (
+                                        self.is_x_axis_contained(
+                                            formula1.box, formula2.box
+                                        )
+                                        and self.has_y_intersection(
+                                            formula1.box, formula2.box
+                                        )
+                                    )
+                                    or (
+                                        self.is_x_axis_adjacent(
+                                            formula1.box, formula2.box
+                                        )
+                                        and self.calculate_y_iou(
+                                            formula1.box, formula2.box
+                                        )
+                                        > 0.5
+                                    )
+                                )
+                            )
+                            or (self._have_same_layout_ids(formula1, formula2, page))
+                            or (
+                                calculate_iou_for_boxes(formula1.box, formula2.box)
+                                > 0.8
+                            )
+                            or (
+                                calculate_iou_for_boxes(formula2.box, formula1.box)
+                                > 0.8
+                            )
+                        )
 
-                if should_merge:
-                    # 合并公式
-                    merged_formula = self.merge_formulas(formula1, formula2)
-                    paragraph.pdf_paragraph_composition[i] = PdfParagraphComposition(
-                        pdf_formula=merged_formula,
-                    )
-                    # 删除第二个公式
-                    del paragraph.pdf_paragraph_composition[i + 1]
-                    # 不增加 i，因为合并后的公式可能还需要和下一个公式合并
-                else:
-                    i += 1
+                        if should_merge:
+                            # 合并公式
+                            merged_formula = self.merge_formulas(formula1, formula2)
+                            paragraph.pdf_paragraph_composition[i] = (
+                                PdfParagraphComposition(
+                                    pdf_formula=merged_formula,
+                                )
+                            )
+                            # 删除第二个公式
+                            del paragraph.pdf_paragraph_composition[j]
+                            merged = True
+                            break
 
     def _have_same_layout_ids(
         self, formula1: PdfFormula, formula2: PdfFormula, page: Page

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/format/pdf/document_il/midend/typesetting.py

@@ -350,11 +350,12 @@ class TypesettingUnit:
 
             return box
         elif self.formular:
-            if self.formular.x_offset <= 0.5:
-                return self.formular.box
-            formular_box = copy.copy(self.formular.box)
-            formular_box.x2 += self.formular.x_advance
-            return formular_box
+            return self.formular.box
+            # if self.formular.x_offset <= 0.5:
+            #     return self.formular.box
+            # formular_box = copy.copy(self.formular.box)
+            # formular_box.x2 += self.formular.x_advance
+            # return formular_box
         elif self.unicode:
             char_width = self.font.char_lengths(self.unicode, self.font_size)[0]
             if self.x is None or self.y is None or self.scale is None:
@@ -407,8 +408,8 @@ class TypesettingUnit:
         elif self.formular:
             # 创建新的公式对象，保持内部字符的相对位置
             new_chars = []
-            min_x = min(char.visual_bbox.box.x for char in self.formular.pdf_character)
-            min_y = min(char.visual_bbox.box.y for char in self.formular.pdf_character)
+            min_x = self.formular.box.x
+            min_y = self.formular.box.y
 
             for char in self.formular.pdf_character:
                 # 计算相对位置

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/format/pdf/document_il/utils/layout_helper.py

@@ -634,6 +634,7 @@ def get_character_layout(
             "abstract",
             "content",
             "figure_title",
+            "chart_title",
             "table_title",
             "doc_title",
             "footnote",
@@ -722,6 +723,7 @@ def is_text_layout(layout: Layout):
         "footer",
         "seal",
         "text",
+        "chart_title",
     ]
 
 

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/main.py

@@ -21,7 +21,7 @@ from babeldoc.translator.translator import OpenAITranslator
 from babeldoc.translator.translator import set_translate_rate_limiter
 
 logger = logging.getLogger(__name__)
-__version__ = "0.4.2"
+__version__ = "0.4.4"
 
 
 def create_parser():

{babeldoc-0.4.2 → babeldoc-0.4.4}/babeldoc/pdfminer/pdffont.py

@@ -8,6 +8,7 @@ from typing import TYPE_CHECKING
 from typing import Any
 from typing import BinaryIO
 from typing import cast
+import freetype
 
 from babeldoc.pdfminer.casting import safe_float
 from babeldoc.pdfminer.casting import safe_rect_list
@@ -749,99 +750,16 @@ class TrueTypeFont:
     def create_unicode_map(self) -> FileUnicodeMap:
         if b"cmap" not in self.tables:
             raise TrueTypeFont.CMapNotFound
-        (base_offset, length) = self.tables[b"cmap"]
         fp = self.fp
-        fp.seek(base_offset)
-        (version, nsubtables) = cast(tuple[int, int], struct.unpack(">HH", fp.read(4)))
-        subtables: list[tuple[int, int, int]] = []
-        for i in range(nsubtables):
-            subtables.append(
-                cast(tuple[int, int, int], struct.unpack(">HHL", fp.read(8))),
-            )
-        char2gid: dict[int, int] = {}
-        # Only supports subtable type 0, 2 and 4.
-        for platform_id, encoding_id, st_offset in subtables:
-            # Skip non-Unicode cmaps.
-            # https://docs.microsoft.com/en-us/typography/opentype/spec/cmap
-            if not (platform_id == 0 or (platform_id == 3 and encoding_id in [1, 10])):
-                continue
-            fp.seek(base_offset + st_offset)
-            (fmttype, fmtlen, fmtlang) = cast(
-                tuple[int, int, int],
-                struct.unpack(">HHH", fp.read(6)),
-            )
-            if fmttype == 0:
-                char2gid.update(
-                    enumerate(
-                        cast(tuple[int, ...], struct.unpack(">256B", fp.read(256))),
-                    ),
-                )
-            elif fmttype == 2:
-                subheaderkeys = cast(
-                    tuple[int, ...],
-                    struct.unpack(">256H", fp.read(512)),
-                )
-                firstbytes = [0] * 8192
-                for i, k in enumerate(subheaderkeys):
-                    firstbytes[k // 8] = i
-                nhdrs = max(subheaderkeys) // 8 + 1
-                hdrs: list[tuple[int, int, int, int, int]] = []
-                for i in range(nhdrs):
-                    (firstcode, entcount, delta, offset) = cast(
-                        tuple[int, int, int, int],
-                        struct.unpack(">HHhH", fp.read(8)),
-                    )
-                    hdrs.append((i, firstcode, entcount, delta, fp.tell() - 2 + offset))
-                for i, firstcode, entcount, delta, pos in hdrs:
-                    if not entcount:
-                        continue
-                    first = firstcode + (firstbytes[i] << 8)
-                    fp.seek(pos)
-                    for c in range(entcount):
-                        gid = cast(tuple[int], struct.unpack(">H", fp.read(2)))[0]
-                        if gid:
-                            gid += delta
-                        char2gid[first + c] = gid
-            elif fmttype == 4:
-                (segcount, _1, _2, _3) = cast(
-                    tuple[int, int, int, int],
-                    struct.unpack(">HHHH", fp.read(8)),
-                )
-                segcount //= 2
-                ecs = cast(
-                    tuple[int, ...],
-                    struct.unpack(">%dH" % segcount, fp.read(2 * segcount)),
-                )
-                fp.read(2)
-                scs = cast(
-                    tuple[int, ...],
-                    struct.unpack(">%dH" % segcount, fp.read(2 * segcount)),
-                )
-                idds = cast(
-                    tuple[int, ...],
-                    struct.unpack(">%dh" % segcount, fp.read(2 * segcount)),
-                )
-                pos = fp.tell()
-                idrs = cast(
-                    tuple[int, ...],
-                    struct.unpack(">%dH" % segcount, fp.read(2 * segcount)),
-                )
-                for ec, sc, idd, idr in zip(ecs, scs, idds, idrs, strict=False):
-                    if idr:
-                        fp.seek(pos + idr)
-                        for c in range(sc, ec + 1):
-                            b = cast(tuple[int], struct.unpack(">H", fp.read(2)))[0]
-                            char2gid[c] = (b + idd) & 0xFFFF
-                    else:
-                        for c in range(sc, ec + 1):
-                            char2gid[c] = (c + idd) & 0xFFFF
-            else:
-                assert False, str(("Unhandled", fmttype))
-        if not char2gid:
+        char2gid = []
+        try:
+            face = freetype.Face(fp)
+            char2gid = list(face.get_chars())
+        except Exception:
             raise TrueTypeFont.CMapNotFound
         # create unicode map
         unicode_map = FileUnicodeMap()
-        for char, gid in char2gid.items():
+        for char, gid in char2gid:
             unicode_map.add_cid2unichr(gid, char)
         return unicode_map
 

{babeldoc-0.4.2 → babeldoc-0.4.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "BabelDOC"
-version = "0.4.2"
+version = "0.4.4"
 description = "Yet Another Document Translator"
 license = "AGPL-3.0"
 readme = "README.md"
@@ -157,7 +157,7 @@ pythonpath = [".", "src"]
 testpaths = ["tests"]
 
 [bumpver]
-current_version = "0.4.2"
+current_version = "0.4.4"
 version_pattern = "MAJOR.MINOR.PATCH[.PYTAGNUM]"
 
 [bumpver.file_patterns]

{babeldoc-0.4.2 → babeldoc-0.4.4}/uv.lock

@@ -48,7 +48,7 @@ wheels = [
 
 [[package]]
 name = "babeldoc"
-version = "0.4.2"
+version = "0.4.4"
 source = { editable = "." }
 dependencies = [
     { name = "bitstring" },
@@ -2333,6 +2333,7 @@ sdist = { url = "https://files.pythonhosted.org/packages/bd/62/d29612ca33b7844e7
 wheels = [
     { url = "https://files.pythonhosted.org/packages/32/5a/3399a2caf51c91db650de57464465b830c2d4ea15b23d24a98182202b704/pymupdf-1.26.1-cp39-abi3-macosx_10_9_x86_64.whl", hash = "sha256:32296f12a7c7f36febd59cee77823a54490313bcaba9879b17def6518186f94e", size = 23054640, upload-time = "2025-06-11T22:14:07.439Z" },
     { url = "https://files.pythonhosted.org/packages/64/e0/cc3ec6a4d5ada8992b8610f134565ceb517243f12736b50d795cb3459315/pymupdf-1.26.1-cp39-abi3-macosx_11_0_arm64.whl", hash = "sha256:aad7949eca62aca40854510cdb125cf873b181726dc9497a90834200f31faa63", size = 22402766, upload-time = "2025-06-11T22:14:25.557Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/cf/d5b1cd775a17a7b83e25cbf4c46f64cf1352c962ca97646e3e01953cf0df/pymupdf-1.26.1-cp39-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:3b62c4d443121ed9a2eb967c3a0e45f8dbabcc838db8604ece02c4e868808edc", size = 23448474, upload-time = "2025-06-20T17:51:56.446Z" },
     { url = "https://files.pythonhosted.org/packages/82/9f/e7101bd24a0f5cbfa0310c8e5c3a8ec0dd9a86986812ff86ac2fbd273c92/pymupdf-1.26.1-cp39-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:a65c411eb1cbb79e40c307e10fbad23658f19e9d7334ac4de21d24b58009a7b9", size = 24056183, upload-time = "2025-06-11T22:14:53.777Z" },
     { url = "https://files.pythonhosted.org/packages/99/39/23ac15cf0edc2877ef366dc7ae041ac199d212433c2c3113661d1a1d5ad0/pymupdf-1.26.1-cp39-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:26cebdcc1b2b7a7445423599ce2e0000f2be0333cce0fa0e6846e5a7da46f965", size = 24258802, upload-time = "2025-06-11T22:15:20.499Z" },
     { url = "https://files.pythonhosted.org/packages/e1/8c/56bd5951128d5c5c0b64d2942090c2cd7bc44302bac991b941ac736e3d63/pymupdf-1.26.1-cp39-abi3-win32.whl", hash = "sha256:82ed9e106cf564fc959c0691c374ba68443086ba1a1c9f26128eebbc3e6df9e5", size = 16927933, upload-time = "2025-06-11T22:15:59.72Z" },

babeldoc-0.4.2/babeldoc/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.4.2"