document-adapter 0.7.2__tar.gz → 0.8.0__tar.gz

{document_adapter-0.7.2/document_adapter.egg-info → document_adapter-0.8.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: document-adapter
-Version: 0.7.2
+Version: 0.8.0
 Summary: LLM-friendly document template editing (DOCX/PPTX/HWPX) with MCP server and Claude API tool-use support
 Author-email: Son Seongjun <sonsj97@plateer.com>
 License: MIT

{document_adapter-0.7.2 → document_adapter-0.8.0}/document_adapter/base.py

@@ -183,6 +183,38 @@ class TableSchema:
         return d
 
 
+@dataclass
+class ShapeInfo:
+    """PPTX 의 표 외 shape (textbox / placeholder / 도형 내 텍스트) 메타.
+
+    ``shape_id`` 는 슬라이드 내 고유 숫자. ``name`` 은 사람 이름 (예: "Title 1").
+    ``text`` 는 full text. ``text_preview`` 는 40자 컷.
+    """
+    slide_index: int
+    shape_id: int
+    name: str
+    kind: str                  # shape 종류 (placeholder / text_box / group / picture 등)
+    has_text: bool
+    text: str = ""
+    text_preview: str = ""
+    placeholder_type: str | None = None
+
+    def to_dict(self) -> dict[str, Any]:
+        d = {
+            "slide_index": self.slide_index,
+            "shape_id": self.shape_id,
+            "name": self.name,
+            "kind": self.kind,
+            "has_text": self.has_text,
+            "text_preview": self.text_preview,
+        }
+        if self.text and self.text != self.text_preview:
+            d["text"] = self.text
+        if self.placeholder_type:
+            d["placeholder_type"] = self.placeholder_type
+        return d
+
+
 @dataclass
 class CellContent:
     """단일 셀의 전체 내용 + 병합/중첩 메타 + 크기 힌트.
@@ -304,6 +336,35 @@ class DocumentAdapter(ABC):
     def append_row(self, table_index: int, values: list[str]) -> None:
         """표 끝에 새 행 추가."""
 
+    # ---- shape text (v0.8+, PPTX 전용) ----
+    def get_shapes(
+        self,
+        slide_index: int | None = None,
+        min_text_len: int = 1,
+        max_preview: int = 40,
+    ) -> list[ShapeInfo]:
+        """표 외 shape (textbox / placeholder / 도형 텍스트) 목록.
+
+        PPTX 에서만 실질적 의미가 있다. DOCX/HWPX 는 기본 빈 리스트 반환
+        (해당 포맷은 표와 paragraph 위주로 편집).
+        """
+        return []
+
+    def set_shape_text(
+        self,
+        slide_index: int,
+        shape_id: int,
+        text: str,
+    ) -> str:
+        """shape 의 텍스트 프레임을 text 로 교체. 기존 run-level 포맷 보존.
+
+        Returns: 기존 텍스트.
+        PPTX 만 지원 — 그 외 포맷은 NotImplementedForFormat.
+        """
+        raise NotImplementedForFormat(
+            f"{self.format} does not support shape-level text editing"
+        )
+
     # ---- label-based form filling ----
     def fill_form(
         self,

{document_adapter-0.7.2 → document_adapter-0.8.0}/document_adapter/hwpx_adapter.py

@@ -118,11 +118,21 @@ class HwpxAdapter(DocumentAdapter):
                     yield from walk(child_tbl, child_parent, section_name)
 
         for section_name, root in self._pkg.iter_section_roots():
-            # 최상위 <hp:tbl> 찾기: root > hp:p > hp:run > hp:tbl
-            for p in root.findall(HP_P):
-                for run in p.findall(HP_RUN):
-                    for tbl in run.findall(HP_TBL):
-                        yield from walk(tbl, "", section_name)
+            # 섹션의 모든 <hp:tbl> descendant 중 "top-level" 만 선별.
+            # top-level = <hp:tc> (cell) 의 descendant 가 아닌 것.
+            # 이러면 <hp:p>/<hp:run>/<hp:ctrl>/<hp:header>/<hp:footer>/도형 등
+            # 어디에 놓여있든 표를 누락 없이 발견 (xgen-doc2chunk 의 _process_ctrl
+            # 가 처리하던 범위 포함). cell 내부 <hp:tbl> 은 walk 재귀에서 처리.
+            for tbl in root.iter(HP_TBL):
+                parent = tbl.getparent()
+                is_in_cell = False
+                while parent is not None:
+                    if parent.tag == HP_TC:
+                        is_in_cell = True
+                        break
+                    parent = parent.getparent()
+                if not is_in_cell:
+                    yield from walk(tbl, "", section_name)
 
     def _get_table(self, table_index: int) -> tuple[etree._Element, str]:
         """flat_index로 (tbl_element, section_part_name) 반환."""

{document_adapter-0.7.2 → document_adapter-0.8.0}/document_adapter/pptx_adapter.py

@@ -23,6 +23,7 @@ from .base import (
     MergeInfo,
     MergedCellWriteError,
     NotImplementedForFormat,
+    ShapeInfo,
     TableIndexError,
     TableSchema,
 )
@@ -270,6 +271,76 @@ class PptxAdapter(DocumentAdapter):
             char_count=len(text),
         )
 
+    # ---- shape text (v0.8+) ----
+    def get_shapes(
+        self,
+        slide_index: int | None = None,
+        min_text_len: int = 1,
+        max_preview: int = 40,
+    ) -> list[ShapeInfo]:
+        """표 외 shape (textbox / placeholder / 도형 텍스트) 수집.
+
+        ``slide_index`` 는 1-based. None 이면 전체.
+        ``min_text_len`` 미만의 텍스트는 제외 (0 을 주면 빈 shape 도 포함).
+        """
+        shapes_out: list[ShapeInfo] = []
+        for s_idx, slide in enumerate(self._prs.slides, 1):
+            if slide_index is not None and s_idx != slide_index:
+                continue
+            for shape in slide.shapes:
+                if shape.has_table:
+                    continue  # 표는 get_tables 로
+                if not shape.has_text_frame:
+                    continue
+                text = (shape.text_frame.text or "").strip()
+                if len(text) < min_text_len:
+                    continue
+                ph_type = None
+                try:
+                    ph = shape.placeholder_format
+                    if ph is not None and ph.type is not None:
+                        ph_type = str(ph.type).rsplit(".", 1)[-1]
+                except ValueError:
+                    pass
+                kind = "placeholder" if ph_type else "text_box"
+                shapes_out.append(
+                    ShapeInfo(
+                        slide_index=s_idx,
+                        shape_id=shape.shape_id,
+                        name=shape.name,
+                        kind=kind,
+                        has_text=bool(text),
+                        text=text,
+                        text_preview=text[:max_preview],
+                        placeholder_type=ph_type,
+                    )
+                )
+        return shapes_out
+
+    def set_shape_text(
+        self,
+        slide_index: int,
+        shape_id: int,
+        text: str,
+    ) -> str:
+        """shape 의 텍스트를 text 로 교체. 기존 run-level 포맷 보존."""
+        for s_idx, slide in enumerate(self._prs.slides, 1):
+            if s_idx != slide_index:
+                continue
+            for shape in slide.shapes:
+                if shape.shape_id != shape_id:
+                    continue
+                if not shape.has_text_frame:
+                    raise ValueError(
+                        f"shape {shape_id} on slide {slide_index} has no text frame"
+                    )
+                old = shape.text_frame.text or ""
+                _set_text_frame_preserving_format(shape.text_frame, text)
+                return old
+        raise ValueError(
+            f"shape not found: slide_index={slide_index}, shape_id={shape_id}"
+        )
+
     # ---- editing ----
     def render_template(self, context: dict[str, Any]) -> None:
         """paragraph 단위로 {{key}}를 치환. run이 쪼개진 경우를 처리하기 위해

{document_adapter-0.7.2 → document_adapter-0.8.0}/document_adapter/tools.py

@@ -172,6 +172,53 @@ TOOL_DEFINITIONS: list[dict[str, Any]] = [
             "required": ["path", "table_index", "values"],
         },
     },
+    {
+        "name": "get_shapes",
+        "description": (
+            "**PPTX 전용**. 표 외 shape (textbox / placeholder / 도형 내 텍스트) 목록을 "
+            "반환한다. PPTX 는 17 슬라이드에 걸쳐 표 4 개 외에 수십 개의 textbox / "
+            "placeholder 로 내용이 채워져 있으므로, 전체 편집에는 get_tables 만으로 "
+            "부족하다. "
+            "반환: shapes[] — 각 항목에 slide_index (1-based), shape_id (슬라이드 내 고유 "
+            "숫자), name, kind (placeholder/text_box), text/text_preview, placeholder_type. "
+            "DOCX/HWPX 는 빈 리스트 반환 (표와 paragraph 중심이라 shape 편집 개념 약함)."
+        ),
+        "input_schema": {
+            "type": "object",
+            "properties": {
+                "path": {"type": "string"},
+                "slide_index": {
+                    "type": "integer",
+                    "description": "1-based. 생략 시 전체 슬라이드.",
+                },
+                "min_text_len": {
+                    "type": "integer",
+                    "default": 1,
+                    "description": "이 값 미만 길이의 텍스트 shape 는 제외.",
+                },
+            },
+            "required": ["path"],
+        },
+    },
+    {
+        "name": "set_shape_text",
+        "description": (
+            "**PPTX 전용**. shape (textbox/placeholder) 의 텍스트를 교체한다. run-level "
+            "포맷 (폰트/크기/색상) 은 보존. get_shapes 로 slide_index + shape_id 파악 후 "
+            "호출. 표 셀은 이 도구 아닌 set_cell 을 쓸 것."
+        ),
+        "input_schema": {
+            "type": "object",
+            "properties": {
+                "path": {"type": "string"},
+                "slide_index": {"type": "integer", "description": "1-based"},
+                "shape_id": {"type": "integer"},
+                "text": {"type": "string"},
+                "output_path": {"type": "string"},
+            },
+            "required": ["path", "slide_index", "shape_id", "text"],
+        },
+    },
     {
         "name": "fill_form",
         "description": (
@@ -364,6 +411,43 @@ def append_row(path: str, table_index: int, values: list[str],
     }
 
 
+def get_shapes(path: str, slide_index: int | None = None,
+               min_text_len: int = 1) -> dict[str, Any]:
+    doc = load(path)
+    try:
+        shapes = doc.get_shapes(slide_index=slide_index, min_text_len=min_text_len)
+    finally:
+        doc.close()
+    return {
+        "format": doc.format,
+        "source": str(path),
+        "shape_count": len(shapes),
+        "shapes": [s.to_dict() for s in shapes],
+    }
+
+
+def set_shape_text(path: str, slide_index: int, shape_id: int, text: str,
+                   output_path: str | None = None) -> dict[str, Any]:
+    target = Path(output_path) if output_path else Path(path)
+    if output_path and Path(path) != target:
+        shutil.copy2(path, target)
+
+    doc = load(target)
+    try:
+        old = doc.set_shape_text(slide_index, shape_id, text)
+        doc.save()
+    finally:
+        doc.close()
+
+    return {
+        "output_path": str(target),
+        "slide_index": slide_index,
+        "shape_id": shape_id,
+        "previous_text": old,
+        "new_text": text,
+    }
+
+
 def fill_form(path: str, data: dict[str, str],
               direction: str = "auto", strict: bool = False,
               output_path: str | None = None) -> dict[str, Any]:
@@ -392,6 +476,8 @@ TOOL_HANDLERS = {
     "append_to_cell": append_to_cell,
     "append_row": append_row,
     "fill_form": fill_form,
+    "get_shapes": get_shapes,
+    "set_shape_text": set_shape_text,
 }
 
 

{document_adapter-0.7.2 → document_adapter-0.8.0/document_adapter.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: document-adapter
-Version: 0.7.2
+Version: 0.8.0
 Summary: LLM-friendly document template editing (DOCX/PPTX/HWPX) with MCP server and Claude API tool-use support
 Author-email: Son Seongjun <sonsj97@plateer.com>
 License: MIT

{document_adapter-0.7.2 → document_adapter-0.8.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "document-adapter"
-version = "0.7.2"
+version = "0.8.0"
 description = "LLM-friendly document template editing (DOCX/PPTX/HWPX) with MCP server and Claude API tool-use support"
 readme = "README.md"
 requires-python = ">=3.10"

{document_adapter-0.7.2 → document_adapter-0.8.0}/tests/test_smoke.py

@@ -1207,3 +1207,75 @@ def test_fill_form_label_normalization(tmp_path: Path) -> None:
 
     assert result["not_found"] == []
     assert len(result["filled"]) == 1
+
+
+# ---- v0.7.3: <hp:ctrl> 내부 테이블 탐지 회귀 방지 ----
+
+def _make_hwpx_with_ctrl_embedded_table(path: Path) -> None:
+    """기본 2x2 표 1 개에 더해 <hp:ctrl> 내부에 동일 구조 표를 1 개 추가.
+
+    HWPX 양식에서 header/footer/footNote 등은 <hp:ctrl> 하위에 들어가므로
+    LLM 이 편집해야 할 표가 ctrl 내부에 숨어있을 수 있다. adapter 가 이를
+    top-level 로 인식하고 편집 가능해야 한다.
+    """
+    import zipfile
+    from copy import deepcopy
+    from lxml import etree
+
+    # 1. 기본 HWPX 생성
+    doc = HwpxDocument.new()
+    doc.add_paragraph("before")
+    doc.add_table(2, 2)
+    doc.add_paragraph("after")
+    doc.save_to_path(path)
+    doc.close()
+
+    HP = "{http://www.hancom.co.kr/hwpml/2011/paragraph}"
+
+    # 2. section0.xml 에 <hp:ctrl> 래핑된 tbl 추가
+    with zipfile.ZipFile(path, "r") as zin:
+        section_xml = zin.read("Contents/section0.xml")
+        other = {n: zin.read(n) for n in zin.namelist() if n != "Contents/section0.xml"}
+        infos = {n: zin.getinfo(n) for n in zin.namelist()}
+
+    root = etree.fromstring(section_xml)
+    existing_tbl = next(root.iter(f"{HP}tbl"))
+    cloned = deepcopy(existing_tbl)
+    new_p = etree.SubElement(root, f"{HP}p")
+    new_run = etree.SubElement(new_p, f"{HP}run")
+    new_ctrl = etree.SubElement(new_run, f"{HP}ctrl")
+    new_ctrl.append(cloned)
+
+    modified_xml = etree.tostring(
+        root, xml_declaration=True, encoding="UTF-8", standalone=True
+    )
+    with zipfile.ZipFile(path, "w") as zout:
+        for name, info in infos.items():
+            data = modified_xml if name == "Contents/section0.xml" else other[name]
+            new_info = zipfile.ZipInfo(filename=name, date_time=info.date_time)
+            new_info.compress_type = info.compress_type
+            zout.writestr(new_info, data)
+
+
+def test_hwpx_ctrl_embedded_table_is_found(tmp_path: Path) -> None:
+    """<hp:ctrl> 내부 tbl 이 top-level 로 발견되고 편집 가능해야 한다 (v0.7.3 회귀 방지)."""
+    src = tmp_path / "ctrl_embedded.hwpx"
+    _make_hwpx_with_ctrl_embedded_table(src)
+
+    adapter = load(src)
+    try:
+        tables = adapter.get_tables()
+        # 2 개 (top-level + ctrl 내부) 전부 발견
+        assert len(tables) == 2, f"expected 2 tables (top-level + ctrl inner), got {len(tables)}"
+        # 두 번째 (ctrl 내부) 표 편집도 가능해야
+        adapter.set_cell(1, 0, 0, "__CTRL_INNER_EDIT__")
+        adapter.save()
+    finally:
+        adapter.close()
+
+    adapter2 = load(src)
+    try:
+        cell = adapter2.get_cell(1, 0, 0)
+        assert cell.text.strip() == "__CTRL_INNER_EDIT__"
+    finally:
+        adapter2.close()