fix

docutranslate/agents/segments_agent.py

@@ -1,4 +1,5 @@
 import json
+from json_repair import json_repair
 from dataclasses import dataclass
 from json import JSONDecodeError
 
@@ -28,12 +29,18 @@ class SegmentsTranslateAgent(Agent):
 翻译后的片段应该与源格式尽量相同
 如果待翻译片段已经是目标语言，则保持原样
 # 输出
-翻译后的片段序列，以json文本表示（文本而非代码块）。其中键是片段编号，值是翻译后的片段
+翻译后的片段序列，以json文本表示（注意不是代码块）。其中键是片段编号，值是翻译后的片段。
+返回的json文本必须能被json.loads转换为形如{{"片段编号":"译文"}}的字典。
 # 示例
 ## 输入
 {r'{"0":"hello","1":"apple","2":true,"3":"false"}'}
 ## 输出
 {r'{"0":"你好","1":"苹果","2":true,"3":"错误"}'}
+# 错误输出示例
+"{r'"{\"0\":\"你好\",\"1\":\"苹果\"}"'}"
+# 正确输出示例
+{r'{"0":"你好","1":"苹果"}'}
+警告：绝不要将整个JSON对象用引号包裹成一个字符串。
 """
         if config.custom_prompt:
             self.system_prompt += "\n# 重要规则或背景【非常重要】\n" + config.custom_prompt + '\n'
@@ -41,26 +48,40 @@ class SegmentsTranslateAgent(Agent):
     def send_segments(self, segments: list[str], chunk_size: int):
         indexed_originals = {str(i): text for i, text in enumerate(segments)}
         chunks = flat_json_split(indexed_originals, chunk_size)
-        prompts = [json.dumps(chunk) for chunk in chunks]
+        prompts = [json.dumps(chunk,ensure_ascii=False) for chunk in chunks]
         translated_chunks = super().send_prompts(prompts=prompts)
         indexed_translated = indexed_originals.copy()
         for chunk_str in translated_chunks:
-            translated_part = json.loads(chunk_str)
-            indexed_translated.update(translated_part)
+            try:
+                translated_part = json_repair.loads(chunk_str)
+                for key,val in translated_part:
+                    if key in indexed_translated:
+                        indexed_translated[key]=val
+            except JSONDecodeError as e:
+                self.logger.info(f"json解析错误，解析文本:{chunk_str}，错误:{e.__repr__()}")
+            except ValueError as e:
+                self.logger.info(f"value错误，更新对象:{indexed_translated}，错误:{e.__repr__()}")
+
         return list(indexed_translated.values())
 
     #todo:增加协程粒度
     async def send_segments_async(self, segments: list[str], chunk_size: int):
         indexed_originals = {str(i): text for i, text in enumerate(segments)}
         chunks = flat_json_split(indexed_originals, chunk_size)
-        prompts = [json.dumps(chunk) for chunk in chunks]
+        prompts = [json.dumps(chunk,ensure_ascii=False) for chunk in chunks]
         translated_chunks = await super().send_prompts_async(prompts=prompts)
         indexed_translated = indexed_originals.copy()
         for chunk_str in translated_chunks:
             try:
-                translated_part = json.loads(chunk_str)
-                indexed_translated.update(translated_part)
+                translated_part:dict = json_repair.loads(chunk_str)
+                for key,val in translated_part.items():
+                    if key in indexed_translated:
+                        indexed_translated[key]=val
             except JSONDecodeError as e:
-                self.logger.info(f"json解析错误，解析文本:{chunk_str}，错误:{e.__repr__()}")
+                self.logger.error(f"json解析错误，解析文本:{chunk_str}，错误:{e.__repr__()}")
+            except ValueError as e:
+                self.logger.error(f"value错误，更新对象:{indexed_translated}，错误:{e.__repr__()}")
+            except AttributeError as e:
+                self.logger.error(f"属性错误,chunk_str:{chunk_str}，错误:{e.__repr__()}")
 
         return list(indexed_translated.values())

docutranslate/exporter/xlsx/xlsx2html_exporter.py

@@ -3,6 +3,7 @@ from io import BytesIO
 
 import jinja2
 import openpyxl
+from xlsx2html import xlsx2html
 
 from docutranslate.exporter.base import ExporterConfig
 from docutranslate.exporter.xlsx.base import XlsxExporter
@@ -22,40 +23,5 @@ class Xlsx2HTMLExporter(XlsxExporter):
         self.cdn = config.cdn
 
     def export(self, document: Document) -> Document:
-
-        # 1. 加载工作簿和工作表
-        workbook = openpyxl.load_workbook(BytesIO(document.content))
-        sheet = workbook.active
-
-        # 2. 手动构建HTML字符串
-        table = '<table border="1">\n'
-
-        # 处理表头
-        table += '  <thead>\n    <tr>\n'
-        for cell in sheet[1]:  # 假设第一行是表头
-            table += f'      <th>{cell.value}</th>\n'
-        table += '    </tr>\n  </thead>\n'
-
-        # 处理数据行
-        table += '  <tbody>\n'
-        # iter_rows(min_row=2) 从第二行开始遍历
-        for row in sheet.iter_rows(min_row=2):
-            table += '    <tr>\n'
-            for cell in row:
-                # 处理None值，防止在HTML中显示"None"
-                cell_value = cell.value if cell.value is not None else ""
-                table += f'      <td>{cell_value}</td>\n'
-            table += '    </tr>\n'
-        table += '  </tbody>\n'
-
-        table += '</table>'
-
-        html_template = resource_path("template/xlsx.html").read_text(encoding="utf-8")
-
-        pico = f'<style>{resource_path("static/pico.css").read_text(encoding="utf-8")}</style>' if not self.cdn else r'<link rel="stylesheet" href="https://s4.zstatic.net/ajax/libs/picocss/2.1.1/pico.min.css" integrity="sha512-+4kjFgVD0n6H3xt19Ox84B56MoS7srFn60tgdWFuO4hemtjhySKyW4LnftYZn46k3THUEiTTsbVjrHai+0MOFw==" crossorigin="anonymous" referrerpolicy="no-referrer" />'
-        render = jinja2.Template(html_template).render(
-            title=document.stem,
-            pico=pico,
-            body=table,
-        )
-        return Document.from_bytes(content=render.encode("utf-8"), suffix=".html", stem=document.stem)
+        html_content = xlsx2html(BytesIO(document.content), output=None).getvalue()
+        return Document.from_bytes(content=html_content.encode("utf-8"), suffix=".html", stem=document.stem)