docvortex 0.2.1__tar.gz → 0.2.3__tar.gz

{docvortex-0.2.1/src/docvortex.egg-info → docvortex-0.2.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docvortex
-Version: 0.2.1
+Version: 0.2.3
 Summary: A fast, multi-format document parsing and conversion engine
 License-Expression: MIT
 Project-URL: Repository, https://github.com/myhloli/DocVortex
@@ -52,7 +52,7 @@ Requires-Dist: twine; extra == "dev"
 Dynamic: license-file
 
 
-![DocVortex overview: native document inputs flow through a unified document model to Markdown, HTML, LaTeX, DOCX, EPUB, PDF and structured content.](https://raw.githubusercontent.com/myhloli/DocVortex/main/docs/images/docvortex-overview.jpg)
+![DocVortex overview: native document inputs flow through a unified document model to Markdown, HTML, LaTeX, DOCX, EPUB, PDF and structured content.](https://gcore.jsdelivr.net/gh/myhloli/DocVortex@main/docs/images/docvortex-overview.jpg)
 
 # DocVortex
 
@@ -61,12 +61,12 @@ A fast, multi-format document parsing and conversion engine.
 DocVortex provides a complete, standalone document pipeline:
 
 ```text
-Document -> ModelJson + Assets -> MiddleJson + Assets -> Render / Export
+Document -> Unified Intermediate Representation -> Render / Export
 ```
 
 Native inputs include text PDFs, DOC/DOCX, PPT/PPTX, XLS/XLSX, RTF, ODT/ODS/ODP,
 EPUB, HTML, OFD and CSV. Output formats include Markdown, HTML, LaTeX, DOCX,
-EPUB, PDF and structured content. Content List V1/V2 are provided by MinerU.
+EPUB, PDF and structured content. 
 
 Native parsing runs without OCR or VLM inference services.
 PDF classification is an explicit document operation; native analysis does not
@@ -81,8 +81,7 @@ docvortex classify report.pdf
 ```
 
 Python 3.10–3.14 is supported. Native parsing does not require OCR/VLM inference
-services. PDF access uses `pypdfium2>=5.10.1,<6`; the
-compatibility matrix also exercises 5.13.0.
+services. 
 
 ## Parse once, export many times
 
@@ -166,28 +165,5 @@ uv build
 
 DocVortex project code is licensed under the [MIT License](LICENSE.md).
 
-Dependencies include `pydantic>=2.12.5,<3` and `numpy>=1.21.6`; the
-installer selects versions compatible with the active Python interpreter.
-On Apple Silicon, Python 3.14 installation requires macOS 14 or newer because
-of the ONNX Runtime dependency used by file-type detection.
-
-See [the standalone example](demo/README.md) for native parsing and portable
-result bundles, and [the validation record](docs/validation.md) for test coverage.
-
-PDFium uses a bundled, pinned CJK fallback font for non-embedded CJK fonts;
-no system font installation is required. See [PDF font policy](docs/PDF_FONTS.md)
-for initialization, diagnostics and replacement boundaries.
-
-PDF output normalizes fullwidth Latin letters, digits and selected technical
-symbols in natural-language text and table cells, preserving Chinese punctuation,
-formulas, code and link targets. See
-[PDF text normalization](docs/PDF_TEXT_NORMALIZATION.md) for scope and API usage.
-
-See [the DocVortex upgrade guide](docs/DOCVORTEX_UPGRADE.md) for package, protocol,
-PDF text rules and publishing configuration changes.
-
 See [rendering ownership](docs/RENDER_OWNERSHIP.md) for the seven engine targets,
 MinerU Content List integration and public fragment helpers.
-
-See [refactor validation](docs/REFACTOR_PROGRESS.md) for the staged internal
-refactoring, compatibility checks, corpus comparisons and measured performance.

{docvortex-0.2.1 → docvortex-0.2.3}/README.md

@@ -1,5 +1,5 @@
 
-![DocVortex overview: native document inputs flow through a unified document model to Markdown, HTML, LaTeX, DOCX, EPUB, PDF and structured content.](https://raw.githubusercontent.com/myhloli/DocVortex/main/docs/images/docvortex-overview.jpg)
+![DocVortex overview: native document inputs flow through a unified document model to Markdown, HTML, LaTeX, DOCX, EPUB, PDF and structured content.](https://gcore.jsdelivr.net/gh/myhloli/DocVortex@main/docs/images/docvortex-overview.jpg)
 
 # DocVortex
 
@@ -8,12 +8,12 @@ A fast, multi-format document parsing and conversion engine.
 DocVortex provides a complete, standalone document pipeline:
 
 ```text
-Document -> ModelJson + Assets -> MiddleJson + Assets -> Render / Export
+Document -> Unified Intermediate Representation -> Render / Export
 ```
 
 Native inputs include text PDFs, DOC/DOCX, PPT/PPTX, XLS/XLSX, RTF, ODT/ODS/ODP,
 EPUB, HTML, OFD and CSV. Output formats include Markdown, HTML, LaTeX, DOCX,
-EPUB, PDF and structured content. Content List V1/V2 are provided by MinerU.
+EPUB, PDF and structured content. 
 
 Native parsing runs without OCR or VLM inference services.
 PDF classification is an explicit document operation; native analysis does not
@@ -28,8 +28,7 @@ docvortex classify report.pdf
 ```
 
 Python 3.10–3.14 is supported. Native parsing does not require OCR/VLM inference
-services. PDF access uses `pypdfium2>=5.10.1,<6`; the
-compatibility matrix also exercises 5.13.0.
+services. 
 
 ## Parse once, export many times
 
@@ -113,28 +112,5 @@ uv build
 
 DocVortex project code is licensed under the [MIT License](LICENSE.md).
 
-Dependencies include `pydantic>=2.12.5,<3` and `numpy>=1.21.6`; the
-installer selects versions compatible with the active Python interpreter.
-On Apple Silicon, Python 3.14 installation requires macOS 14 or newer because
-of the ONNX Runtime dependency used by file-type detection.
-
-See [the standalone example](demo/README.md) for native parsing and portable
-result bundles, and [the validation record](docs/validation.md) for test coverage.
-
-PDFium uses a bundled, pinned CJK fallback font for non-embedded CJK fonts;
-no system font installation is required. See [PDF font policy](docs/PDF_FONTS.md)
-for initialization, diagnostics and replacement boundaries.
-
-PDF output normalizes fullwidth Latin letters, digits and selected technical
-symbols in natural-language text and table cells, preserving Chinese punctuation,
-formulas, code and link targets. See
-[PDF text normalization](docs/PDF_TEXT_NORMALIZATION.md) for scope and API usage.
-
-See [the DocVortex upgrade guide](docs/DOCVORTEX_UPGRADE.md) for package, protocol,
-PDF text rules and publishing configuration changes.
-
 See [rendering ownership](docs/RENDER_OWNERSHIP.md) for the seven engine targets,
 MinerU Content List integration and public fragment helpers.
-
-See [refactor validation](docs/REFACTOR_PROGRESS.md) for the staged internal
-refactoring, compatibility checks, corpus comparisons and measured performance.

{docvortex-0.2.1 → docvortex-0.2.3}/src/docvortex/analyzers/native/epub/converter.py

@@ -72,6 +72,7 @@ class EpubConverter:
                     logger.warning("Skipping unusable EPUB spine item index={} path={!r}: {}", index, path, exc)
                     blocks = []
                 pages.append(blocks)
+            anchors.finalize_links(pages)
             self.pages = pages
         finally:
             package.close()

{docvortex-0.2.1 → docvortex-0.2.3}/src/docvortex/analyzers/native/epub/xhtml.py

@@ -5,8 +5,10 @@ from __future__ import annotations
 import base64
 from dataclasses import dataclass
 
+from lxml import html as lxml_html
 from lxml import etree  # type: ignore[reportMissingImports]
 
+from ....content.spans import normalize_span_dicts
 from ....foundation.image_payload import parse_image_data_uri_strict
 from .._shared.hyperlink import sanitize_hyperlink_target
 from docvortex.content.markup import (
@@ -52,13 +54,25 @@ _NOTE_NON_TEXT_SUBTREES = frozenset(
 )
 
 
+def _unwrap_epub_type_quotes(value: str) -> str:
+    """移除单层成对外围引号，避免把多个各自带引号的 token 当作一个整体。"""
+    normalized = value.strip()
+    quote_pairs = {'"': '"', "'": "'", "“": "”", "‘": "’"}
+    if len(normalized) < 2 or quote_pairs.get(normalized[0]) != normalized[-1]:
+        return normalized
+    inner = normalized[1:-1]
+    if normalized[0] in inner or normalized[-1] in inner:
+        return normalized
+    return inner.strip()
+
+
 def _epub_types(element: etree._Element) -> frozenset[str]:
     """读取 EPUB 命名空间或未命名 type 属性中的结构语义 token。"""
     values: list[str] = []
     for name, value in element.attrib.items():
         local_name = etree.QName(name).localname if name.startswith("{") else name.split(":", 1)[-1]
         if local_name == "type":
-            values.extend(value.casefold().split())
+            values.extend(_unwrap_epub_type_quotes(token) for token in _unwrap_epub_type_quotes(value).casefold().split())
     return frozenset(values)
 
 
@@ -149,11 +163,12 @@ class _EpubAnchorPolicy:
 
 
 class EpubAnchorRegistry:
-    """建立章节路径、标题与 note fragment 到实际 canonical anchor 的别名表。"""
+    """建立章节路径、正文、标题与 note fragment 到实际 canonical anchor 的别名表。"""
 
     def __init__(self, chapters: list[tuple[str, etree._Element]], package: EpubPackage) -> None:
         """预扫描全部选中 XHTML 章节，建立标题、note 与章节起点映射。"""
         self._package = package
+        self._pending_links: dict[str, tuple[str, str | None]] = {}
         documents = [
             MarkupAnchorDocument(
                 key=chapter_path,
@@ -166,6 +181,35 @@ class EpubAnchorRegistry:
         ]
         self._registry = MarkupAnchorRegistry(documents, _EpubAnchorPolicy())
 
+    def register_text_targets(
+        self,
+        chapter_path: str,
+        root: etree._Element,
+        sources: list[tuple[etree._Element, dict[str, object]]],
+    ) -> None:
+        """在章节投影完成后登记真实正文目标，不修改已有标题和脚注身份。"""
+        self._registry.register_text_targets(chapter_path, root, sources)
+
+    def defer_link(self, href: str, *, base_part: str) -> str | None:
+        """暂存尚未物化的包内目标，跨章节投影完成后统一兑现或降级。"""
+        normalized = sanitize_hyperlink_target(href, allowed_schemes=(), allow_relative=True, allow_fragment=True)
+        target = self._package.resolve_reference(normalized, base_part=base_part) if normalized is not None else None
+        if target is None:
+            return None
+        placeholder = f"#epub-pending-{len(self._pending_links)}"
+        self._pending_links[placeholder] = (target.path, target.fragment)
+        return placeholder
+
+    def finalize_links(self, pages: list[list[dict[str, object]]]) -> None:
+        """消除所有解析期链接占位，保留无效链接的文字与样式。"""
+        targets = {
+            placeholder: (f"#{anchor}" if (anchor := self._registry.resolve_target(path, fragment)) else None)
+            for placeholder, (path, fragment) in self._pending_links.items()
+        }
+        for page in pages:
+            _finalize_pending_content(page, targets)
+        self._pending_links.clear()
+
     def heading_anchor(self, heading: etree._Element) -> str | None:
         """返回一个已预扫描 EPUB 标题的规范 anchor。"""
         return self._registry.heading_anchor(heading)
@@ -179,7 +223,7 @@ class EpubAnchorRegistry:
         return self._registry.note_anchor(note)
 
     def resolve_anchor(self, href: str, *, base_part: str) -> str | None:
-        """解析指向正文标题或 note 的 EPUB 包内链接，并返回不带井号的 anchor。"""
+        """解析指向已登记正文、标题或 note 的包内链接，返回不带井号的 anchor。"""
         normalized = sanitize_hyperlink_target(
             href,
             allowed_schemes=(),
@@ -194,7 +238,7 @@ class EpubAnchorRegistry:
         return self._registry.resolve_target(target.path, target.fragment)
 
     def resolve_link(self, href: str, *, base_part: str) -> str | None:
-        """解析安全外部链接或指向已输出标题/note 的 EPUB 内部链接。"""
+        """解析安全外部链接或指向已登记输出块的 EPUB 内部链接。"""
         external = sanitize_hyperlink_target(href)
         if external is not None:
             return external
@@ -217,7 +261,9 @@ class _EpubMarkupContext:
 
     def resolve_link(self, href: str) -> str | None:
         """解析安全外部链接或实际存在的 EPUB 包内 anchor。"""
-        return self.anchors.resolve_link(href, base_part=self.chapter_path)
+        return self.anchors.resolve_link(href, base_part=self.chapter_path) or self.anchors.defer_link(
+            href, base_part=self.chapter_path
+        )
 
     def resolve_image(self, source: str, *, alt: str = "") -> ResolvedMarkupImage | None:
         """读取并严格校验一个 EPUB 包内栅格图片引用。"""
@@ -278,12 +324,50 @@ class EpubChapterConverter:
         if body is None:
             return []
         context = _EpubMarkupContext(self.package, self.chapter_path, self.anchors)
-        return MarkupProjector(
+        projector = MarkupProjector(
             body,
             context,
             self.stylesheet,
             single_document_title=False,
-        ).convert()
+            track_text_sources=True,
+        )
+        blocks = projector.convert()
+        self.anchors.register_text_targets(self.chapter_path, self.root, projector.text_sources)
+        return blocks
+
+
+def _finalize_pending_content(items: list[dict[str, object]], targets: dict[str, str | None]) -> None:
+    """递归处理 raw block 和行内 Span，表格 HTML 同步替换或解包失效链接。"""
+    output: list[dict[str, object]] = []
+    flattened = False
+    for item in items:
+        content = item.get("content")
+        if isinstance(content, list):
+            _finalize_pending_content(content, targets)
+        elif item.get("type") == "table" and isinstance(content, str) and "#epub-pending-" in content:
+            root = lxml_html.fragment_fromstring(content, create_parent="div")
+            for link in root.iter("a"):
+                href = link.get("href")
+                if href in targets:
+                    if target := targets[href]:
+                        link.set("href", target)
+                    else:
+                        link.drop_tag()
+            item["content"] = (root.text or "") + "".join(lxml_html.tostring(child, encoding="unicode") for child in root)
+        url = item.get("url")
+        if item.get("type") == "hyperlink" and isinstance(url, str) and url in targets:
+            if target := targets[url]:
+                item["url"] = target
+            else:
+                if isinstance(content, list):
+                    output.extend(content)
+                flattened = True
+                continue
+        output.append(item)
+    if flattened:
+        # 只有失效链接被解包的 Span 列表需要重新合并，避免重复校验整本正文。
+        output = normalize_span_dicts(output)
+    items[:] = output
 
 
 def convert_svg_spine(

{docvortex-0.2.1 → docvortex-0.2.3}/src/docvortex/content/markup/anchors.py

@@ -154,6 +154,29 @@ class MarkupAnchorRegistry:
             return None
         return self._heading_anchors.get(descendant) or self._note_anchors.get(descendant)
 
+    def register_text_targets(
+        self,
+        document_key: str,
+        root: etree._Element,
+        sources: list[tuple[etree._Element, dict[str, object]]],
+    ) -> None:
+        """按源顺序把实际物化正文的多个 ID 别名绑定到一个块级 anchor。"""
+        blocks_by_source = {source: block for source, block in reversed(sources)}
+        for element in root.iter():
+            if not isinstance(element.tag, str) or not (fragment := element_id(element)):
+                continue
+            block = blocks_by_source.get(element)
+            if block is None:
+                continue
+            target_key = (document_key, fragment)
+            if target_key in self._targets:
+                continue
+            anchor = block.get("anchor")
+            if not isinstance(anchor, str):
+                anchor = canonical_anchor(self._policy.anchor_prefix, document_key, f"text-{fragment}")
+                block["anchor"] = anchor
+            self._targets[target_key] = anchor
+
     def heading_anchor(self, heading: etree._Element) -> str | None:
         """返回一个已登记标题的规范 anchor。"""
         return self._heading_anchors.get(heading)