From e9de5c9ab99da211598be83460fc847f0a5a068c Mon Sep 17 00:00:00 2001 From: xunbu Date: Sun, 7 Dec 2025 21:12:48 +0800 Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E5=A4=8D=E9=83=A8=E5=88=86=E6=96=87?= =?UTF-8?q?=E6=9C=AC=E6=9C=AA=E7=BF=BB=E8=AF=91=E9=97=AE=E9=A2=98?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../ai_translator/xlsx_translator.py | 356 +++++++++++------- 1 file changed, 220 insertions(+), 136 deletions(-) diff --git a/docutranslate/translator/ai_translator/xlsx_translator.py b/docutranslate/translator/ai_translator/xlsx_translator.py index a9d2705..d1b0e67 100644 --- a/docutranslate/translator/ai_translator/xlsx_translator.py +++ b/docutranslate/translator/ai_translator/xlsx_translator.py @@ -5,10 +5,9 @@ from dataclasses import dataclass from io import BytesIO from typing import Self, Literal, List, Optional, Dict, Tuple, Set import zipfile -import re # 引入正则用于解析简写 +import re import xml.etree.ElementTree as ET -# 仅导入 openpyxl 的工具函数用于坐标计算，不加载 workbook 对象 from openpyxl.utils.cell import coordinate_to_tuple, range_boundaries from docutranslate.agents.segments_agent import SegmentsTranslateAgentConfig, SegmentsTranslateAgent @@ -21,9 +20,6 @@ class XlsxTranslatorConfig(AiTranslatorConfig): insert_mode: Literal["replace", "append", "prepend"] = "replace" separator: str = "\n" # 指定翻译区域列表。 - # 示例: ["Sheet1!A1:B10", "C", "3"] (支持简写: C代表C列, 3代表第3行) - # 如果不指定表名 (如 "C")，则应用于所有表。 - # 如果为 None 或空列表，则翻译整个文件中的所有文本。 translate_regions: Optional[List[str]] = None @@ -54,59 +50,99 @@ class XlsxTranslator(AiTranslator): self.separator = config.separator self.translate_regions = config.translate_regions - # 命名空间定义 - self.ns = { - 'main': 'http://schemas.openxmlformats.org/spreadsheetml/2006/main', - 'r': 'http://schemas.openxmlformats.org/officeDocument/2006/relationships' - } - ET.register_namespace('', self.ns['main']) + # 我们虽然不依赖它查找，但写入新节点时最好还是带上标准命名空间 + self.NS_MAIN = 'http://schemas.openxmlformats.org/spreadsheetml/2006/main' + ET.register_namespace('', self.NS_MAIN) # ========================================================================= - # 辅助方法：无需加载 Workbook 即可解析结构 + # 核心辅助方法：忽略命名空间查找 + # ========================================================================= + + def _tag_is(self, elem: ET.Element, tag_name: str) -> bool: + """判断元素的标签名是否匹配（忽略命名空间）。""" + return elem.tag.endswith(f"}}{tag_name}") or elem.tag == tag_name + + def _find_child(self, parent: ET.Element, tag_name: str) -> Optional[ET.Element]: + """在直接子节点中查找指定标签（忽略命名空间）。""" + for child in parent: + if self._tag_is(child, tag_name): + return child + return None + + def _find_all_children(self, parent: ET.Element, tag_name: str) -> List[ET.Element]: + """查找所有匹配的直接子节点（忽略命名空间）。""" + return [child for child in parent if self._tag_is(child, tag_name)] + + def _get_child_text(self, parent: ET.Element, tag_name: str) -> Optional[str]: + """获取子节点的文本内容（忽略命名空间）。""" + child = self._find_child(parent, tag_name) + return child.text if child is not None else None + + # ========================================================================= + # 辅助逻辑 # ========================================================================= def _get_shared_strings(self, zf: zipfile.ZipFile) -> List[str]: - """解析共享字符串表，返回字符串列表。""" if "xl/sharedStrings.xml" not in zf.namelist(): return [] - shared_strings = [] with zf.open("xl/sharedStrings.xml") as f: context = ET.iterparse(f, events=("end",)) for event, elem in context: - if elem.tag.endswith('}si'): # shared item - texts = [t.text for t in elem.findall('.//main:t', self.ns) if t.text] + # 匹配 + if self._tag_is(elem, "si"): + # 查找所有 + # 注意：xml.etree 的 findall 只能简单的路径查找， + # 既然我们要忽略命名空间，最好手动遍历子树，但 si 结构简单， + # 这里简化处理：直接遍历 iter 出来的 t 元素（如果在 si 内部） + pass + # 由于 iterparse 是扁平流，很难直接关联 si 和 t。 + # 更稳妥的方式是：当 elem 是 si 时，遍历 elem 的 children + texts = [] + for child in elem.iter(): + if self._tag_is(child, "t") and child.text: + texts.append(child.text) shared_strings.append("".join(texts)) elem.clear() return shared_strings def _get_sheet_mapping(self, zf: zipfile.ZipFile) -> Dict[str, str]: - """ - 获取 Sheet 名称到文件路径的映射。 - 例如: {'Sheet1': 'xl/worksheets/sheet1.xml'} - """ + """建立 SheetName -> ZipFilename 的映射（稳健版）""" sheet_name_to_rid = {} try: with zf.open("xl/workbook.xml") as f: root = ET.fromstring(f.read()) - for sheet in root.findall(".//main:sheet", self.ns): - name = sheet.get("name") - rid = sheet.get(f"{{{self.ns['r']}}}id") - if name and rid: - sheet_name_to_rid[name] = rid + # 查找所有 sheet 标签 + # 需递归查找，因为 sheet 通常在 sheets 节点下 + for sheet in root.iter(): + if self._tag_is(sheet, "sheet"): + name = sheet.get("name") + # id 的属性名通常带有 r: 前缀，这很难忽略命名空间， + # 但 openpyxl 规范里 id 属性几乎总是依赖 relationships 命名空间 + # 这里我们尝试遍历属性找到 key 包含 id 的 + rid = None + for k, v in sheet.attrib.items(): + if k.endswith("id"): # 匹配 r:id + rid = v + break + if name and rid: + sheet_name_to_rid[name] = rid except Exception: return {} rid_to_target = {} try: with zf.open("xl/_rels/workbook.xml.rels") as f: - tree = ET.parse(f) - root = tree.getroot() + root = ET.fromstring(f.read()) for child in root: + # Relationships 里的 tag 也是 Relationship rid = child.get("Id") target = child.get("Target") if rid and target: - if not target.startswith("/"): + target = target.replace("\\", "/") + if target.startswith("/"): + target = target.lstrip("/") + else: target = "xl/" + target rid_to_target[rid] = target except Exception: @@ -116,50 +152,32 @@ class XlsxTranslator(AiTranslator): for name, rid in sheet_name_to_rid.items(): if rid in rid_to_target: mapping[name] = rid_to_target[rid] - return mapping def _normalize_range(self, range_str: str) -> str: - """ - 将用户输入的简写转换为标准范围格式。 - "C" -> "C:C" - "3" -> "3:3" - "A1" -> "A1:A1" (openpyxl range_boundaries 实际上支持 A1，但这里统一处理更安全) - """ - # 纯字母 (例如 "C", "AA") -> 整列 if re.match(r"^[A-Za-z]+$", range_str): return f"{range_str}:{range_str}" - # 纯数字 (例如 "3", "10") -> 整行 if re.match(r"^\d+$", range_str): return f"{range_str}:{range_str}" return range_str def _parse_region_boundaries(self, sheet_mapping: Dict[str, str]) -> Dict[str, List[Tuple]]: - """ - 解析配置的 translate_regions。 - 返回: { 'xl/worksheets/sheet1.xml': [(min_col, min_row, max_col, max_row), ...], ... } - """ if not self.translate_regions: return {} - - region_map = {} # filename -> list of boundaries - global_regions = [] # list of boundaries for all sheets + region_map = {} + global_regions = [] for region in self.translate_regions: sheet_name = None raw_range = region.strip() - if "!" in raw_range: parts = raw_range.split("!", 1) sheet_name = parts[0].strip("'") range_part = self._normalize_range(parts[1]) else: range_part = self._normalize_range(raw_range) - try: - # boundaries: (min_col, min_row, max_col, max_row) boundaries = range_boundaries(range_part) - if sheet_name: filename = sheet_mapping.get(sheet_name) if filename: @@ -171,18 +189,15 @@ class XlsxTranslator(AiTranslator): except Exception as e: self.logger.warning(f"无法解析区域 '{region}': {e}") - # 将全局区域添加到所有已知 Sheet if global_regions: - all_files = set(sheet_mapping.values()) - for f in all_files: + target_files = set(sheet_mapping.values()) + for f in target_files: if f not in region_map: region_map[f] = [] region_map[f].extend(global_regions) - return region_map def _is_in_boundaries(self, col: int, row: int, boundaries_list: List[Tuple]) -> bool: - """检查坐标 (col, row) 是否在给定的边界列表中。""" for (min_col, min_row, max_col, max_row) in boundaries_list: if min_col is not None and col < min_col: continue if min_row is not None and row < min_row: continue @@ -191,20 +206,30 @@ class XlsxTranslator(AiTranslator): return True return False + def _apply_insert_mode(self, original: str, translated: str) -> str: + if self.insert_mode == "append": + return f"{original}{self.separator}{translated}" + elif self.insert_mode == "prepend": + return f"{translated}{self.separator}{original}" + else: + return translated + # ========================================================================= - # 高效 XML 区域提取与重构 + # 区域处理 (使用 Helper 方法，完全解耦命名空间) # ========================================================================= def _get_texts_xml_regions(self, document: Document) -> List[str]: - """使用纯 XML 解析（结合 SharedStrings）提取指定区域文本。""" texts_to_translate = set() with zipfile.ZipFile(BytesIO(document.content), 'r') as zf: shared_strings = self._get_shared_strings(zf) - if not shared_strings: - return [] - sheet_mapping = self._get_sheet_mapping(zf) + + if not sheet_mapping: + all_sheets = [n for n in zf.namelist() if n.startswith("xl/worksheets/sheet") and n.endswith(".xml")] + for s in all_sheets: + sheet_mapping[f"Unknown_{s}"] = s + boundaries_map = self._parse_region_boundaries(sheet_mapping) for filename, boundaries in boundaries_map.items(): @@ -214,80 +239,108 @@ class XlsxTranslator(AiTranslator): with zf.open(filename) as f: context = ET.iterparse(f, events=("end",)) for event, elem in context: - if elem.tag.endswith('}c'): # Cell - r_attr = elem.get('r') # e.g. "C5" - t_attr = elem.get('t') # e.g. "s" + # 匹配 + if self._tag_is(elem, "c"): + r_attr = elem.get('r') + t_attr = elem.get('t') - if r_attr and t_attr == 's': + if r_attr: try: - # 【修正】coordinate_to_tuple 返回 (row, col) row, col = coordinate_to_tuple(r_attr) if self._is_in_boundaries(col, row, boundaries): - v_node = elem.find('main:v', self.ns) - if v_node is not None and v_node.text: - idx = int(v_node.text) - if 0 <= idx < len(shared_strings): - texts_to_translate.add(shared_strings[idx]) + text_found = None + + # Shared String + if t_attr == 's': + v_text = self._get_child_text(elem, "v") + if v_text: + idx = int(v_text) + if 0 <= idx < len(shared_strings): + text_found = shared_strings[idx] + + # Inline String + elif t_attr == 'inlineStr': + is_node = self._find_child(elem, "is") + if is_node is not None: + t_text = self._get_child_text(is_node, "t") + if t_text: + text_found = t_text + + if text_found: + texts_to_translate.add(text_found) except Exception: pass elem.clear() + elif self._tag_is(elem, "row"): + elem.clear() return list(texts_to_translate) def _rebuild_xml_regions(self, original_content_bytes: bytes, translation_map: dict) -> bytes: - """使用纯 XML 重构，修正了坐标解包顺序。""" output_zip_io = BytesIO() - with zipfile.ZipFile(BytesIO(original_content_bytes), 'r') as zf_in: with zipfile.ZipFile(output_zip_io, 'w', zipfile.ZIP_DEFLATED) as zf_out: - shared_strings = self._get_shared_strings(zf_in) sheet_mapping = self._get_sheet_mapping(zf_in) + + if not sheet_mapping: + all_sheets = [n for n in zf_in.namelist() if + n.startswith("xl/worksheets/sheet") and n.endswith(".xml")] + for s in all_sheets: + sheet_mapping[f"Unknown_{s}"] = s + boundaries_map = self._parse_region_boundaries(sheet_mapping) for item in zf_in.infolist(): if item.filename in boundaries_map: boundaries = boundaries_map[item.filename] - with zf_in.open(item.filename) as f: tree = ET.parse(f) root = tree.getroot() - cells_modified = False - for cell in root.findall(".//main:c", self.ns): + + # 查找所有 + # 因为 findall 不支持复杂的 tag.endswith，这里我们遍历所有节点 + # 对于 parse 加载的树，iter() 是高效的 + for cell in root.iter(): + if not self._tag_is(cell, "c"): + continue + r_attr = cell.get('r') t_attr = cell.get('t') - - if r_attr and t_attr == 's': + if r_attr: try: - # 【修正】coordinate_to_tuple 返回 (row, col) row, col = coordinate_to_tuple(r_attr) - if self._is_in_boundaries(col, row, boundaries): - v_node = cell.find('main:v', self.ns) - if v_node is not None and v_node.text: - idx = int(v_node.text) - if 0 <= idx < len(shared_strings): - original_text = shared_strings[idx] + original_text = None - if original_text in translation_map: - translated_text = translation_map[original_text] + if t_attr == 's': + v_text = self._get_child_text(cell, "v") + if v_text: + idx = int(v_text) + if 0 <= idx < len(shared_strings): + original_text = shared_strings[idx] + elif t_attr == 'inlineStr': + is_node = self._find_child(cell, "is") + if is_node is not None: + original_text = self._get_child_text(is_node, "t") - final_text = translated_text - if self.insert_mode == "append": - final_text = original_text + self.separator + translated_text - elif self.insert_mode == "prepend": - final_text = translated_text + self.separator + original_text + if original_text and original_text in translation_map: + final_text = self._apply_insert_mode(original_text, + translation_map[original_text]) - # 转换为 inlineStr - cell.set('t', 'inlineStr') - cell.remove(v_node) - is_node = ET.Element(f"{{{self.ns['main']}}}is") - t_node = ET.SubElement(is_node, f"{{{self.ns['main']}}}t") - t_node.text = final_text - cell.append(is_node) + # 清空旧内容 + for child in list(cell): + cell.remove(child) - cells_modified = True + # 写入新内容 (这里必须使用带有命名空间的标签名，否则Excel不认) + cell.set('t', 'inlineStr') + # 注意：写入时使用 self.NS_MAIN 是必须的，因为这是标准 + is_node = ET.Element(f"{{{self.NS_MAIN}}}is") + t_node = ET.SubElement(is_node, f"{{{self.NS_MAIN}}}t") + t_node.text = final_text + cell.append(is_node) + cells_modified = True except Exception: pass @@ -298,11 +351,10 @@ class XlsxTranslator(AiTranslator): zf_out.writestr(item, zf_in.read(item.filename)) else: zf_out.writestr(item, zf_in.read(item.filename)) - return output_zip_io.getvalue() # ========================================================================= - # 原有全文档逻辑 (针对全文档翻译保持极致速度) + # 全文档处理 (同样使用 Helper) # ========================================================================= def _get_texts_xml_all(self, document: Document) -> List[str]: @@ -311,20 +363,39 @@ class XlsxTranslator(AiTranslator): with zipfile.ZipFile(BytesIO(document.content), 'r') as zf: if "xl/sharedStrings.xml" in zf.namelist(): with zf.open("xl/sharedStrings.xml") as f: - root = ET.fromstring(f.read()) - for node in root.findall('.//main:t', self.ns): - if node.text and node.text.strip(): - texts_to_translate.add(node.text) + context = ET.iterparse(f, events=("end",)) + for event, elem in context: + if self._tag_is(elem, "t"): + if elem.text and elem.text.strip(): + texts_to_translate.add(elem.text) + elem.clear() + + sheet_files = [n for n in zf.namelist() if n.startswith("xl/worksheets/sheet") and n.endswith(".xml")] + for sheet_file in sheet_files: + with zf.open(sheet_file) as f: + context = ET.iterparse(f, events=("end",)) + for event, elem in context: + if self._tag_is(elem, "c"): + if elem.get('t') == 'inlineStr': + is_node = self._find_child(elem, "is") + if is_node is not None: + t_text = self._get_child_text(is_node, "t") + if t_text and t_text.strip(): + texts_to_translate.add(t_text) + elem.clear() + elif self._tag_is(elem, "row"): + elem.clear() for item in zf.infolist(): if item.filename.startswith("xl/tables/table"): with zf.open(item.filename) as f: root = ET.fromstring(f.read()) - for col in root.findall('.//main:tableColumn', self.ns): - if col.get('name'): - texts_to_translate.add(col.get('name')) + for col in root.iter(): + if self._tag_is(col, "tableColumn"): + if col.get('name'): + texts_to_translate.add(col.get('name')) except Exception as e: - self.logger.error(f"XML解析失败: {e}") + self.logger.error(f"XML解析失败: {e}", exc_info=True) return list(texts_to_translate) def _rebuild_xml_all(self, original_content_bytes: bytes, translation_map: dict) -> bytes: @@ -337,41 +408,54 @@ class XlsxTranslator(AiTranslator): if item.filename == "xl/sharedStrings.xml": root = ET.fromstring(content) - for node in root.findall('.//main:t', self.ns): - if node.text in translation_map: - trans = translation_map[node.text] - if self.insert_mode == "append": - node.text = node.text + self.separator + trans - elif self.insert_mode == "prepend": - node.text = trans + self.separator + node.text - else: - node.text = trans - content = ET.tostring(root, encoding='utf-8', xml_declaration=True) + modified = False + for node in root.iter(): + if self._tag_is(node, "t"): + if node.text in translation_map: + node.text = self._apply_insert_mode(node.text, translation_map[node.text]) + modified = True + if modified: + zf_out.writestr(item, ET.tostring(root, encoding='utf-8', xml_declaration=True)) + else: + zf_out.writestr(item, content) + + elif item.filename.startswith("xl/worksheets/sheet") and item.filename.endswith(".xml"): + root = ET.fromstring(content) + modified = False + for cell in root.iter(): + if self._tag_is(cell, "c") and cell.get('t') == 'inlineStr': + is_node = self._find_child(cell, "is") + if is_node is not None: + t_node = self._find_child(is_node, "t") + if t_node is not None and t_node.text in translation_map: + t_node.text = self._apply_insert_mode(t_node.text, + translation_map[t_node.text]) + modified = True + if modified: + zf_out.writestr(item, ET.tostring(root, encoding='utf-8', xml_declaration=True)) + else: + zf_out.writestr(item, content) elif item.filename.startswith("xl/tables/table"): root = ET.fromstring(content) - for col in root.findall('.//main:tableColumn', self.ns): - orig = col.get('name') - if orig in translation_map: - trans = translation_map[orig] - if self.insert_mode == "append": - col.set('name', orig + self.separator + trans) - elif self.insert_mode == "prepend": - col.set('name', trans + self.separator + orig) - else: - col.set('name', trans) - content = ET.tostring(root, encoding='utf-8', xml_declaration=True) - - zf_out.writestr(item, content) + modified = False + for col in root.iter(): + if self._tag_is(col, "tableColumn"): + orig = col.get('name') + if orig in translation_map: + col.set('name', self._apply_insert_mode(orig, translation_map[orig])) + modified = True + if modified: + zf_out.writestr(item, ET.tostring(root, encoding='utf-8', xml_declaration=True)) + else: + zf_out.writestr(item, content) + else: + zf_out.writestr(item, content) return output_zip_io.getvalue() except Exception as e: - self.logger.error(f"XML重构失败: {e}") + self.logger.error(f"XML重构失败: {e}", exc_info=True) return original_content_bytes - # ========================================================================= - # 主入口 - # ========================================================================= - def translate(self, document: Document) -> Self: if self.translate_regions: original_texts = self._get_texts_xml_regions(document)