Merge branch 'main' of https://gitea.kronecker.cc/OurCodesAreAllRight/FilesReadSystem
This commit is contained in:
@@ -317,24 +317,70 @@ class XlsxParser(BaseParser):
|
||||
import zipfile
|
||||
from xml.etree import ElementTree as ET
|
||||
|
||||
# 常见的命名空间
|
||||
COMMON_NAMESPACES = [
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2006/main',
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2005/main',
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2004/main',
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2003/main',
|
||||
]
|
||||
|
||||
try:
|
||||
with zipfile.ZipFile(file_path, 'r') as z:
|
||||
if 'xl/workbook.xml' not in z.namelist():
|
||||
# 尝试多种可能的 workbook.xml 路径
|
||||
possible_paths = ['xl/workbook.xml', 'xl\\workbook.xml', 'workbook.xml']
|
||||
content = None
|
||||
for path in possible_paths:
|
||||
if path in z.namelist():
|
||||
content = z.read(path)
|
||||
logger.info(f"找到 workbook.xml at: {path}")
|
||||
break
|
||||
|
||||
if content is None:
|
||||
logger.warning(f"未找到 workbook.xml，文件列表: {z.namelist()[:10]}")
|
||||
return []
|
||||
content = z.read('xl/workbook.xml')
|
||||
|
||||
root = ET.fromstring(content)
|
||||
|
||||
# 命名空间
|
||||
ns = {'main': 'http://schemas.openxmlformats.org/spreadsheetml/2006/main'}
|
||||
|
||||
sheet_names = []
|
||||
for sheet in root.findall('.//main:sheet', ns):
|
||||
|
||||
# 方法1：尝试带命名空间的查找
|
||||
for ns in COMMON_NAMESPACES:
|
||||
sheet_elements = root.findall(f'.//{{{ns}}}sheet')
|
||||
if sheet_elements:
|
||||
for sheet in sheet_elements:
|
||||
name = sheet.get('name')
|
||||
if name:
|
||||
sheet_names.append(name)
|
||||
if sheet_names:
|
||||
logger.info(f"使用命名空间 {ns} 提取工作表: {sheet_names}")
|
||||
return sheet_names
|
||||
|
||||
# 方法2：不使用命名空间，直接查找所有 sheet 元素
|
||||
if not sheet_names:
|
||||
for elem in root.iter():
|
||||
if elem.tag.endswith('sheet') and elem.tag != 'sheets':
|
||||
name = elem.get('name')
|
||||
if name:
|
||||
sheet_names.append(name)
|
||||
for child in elem:
|
||||
if child.tag.endswith('sheet') or child.tag == 'sheet':
|
||||
name = child.get('name')
|
||||
if name and name not in sheet_names:
|
||||
sheet_names.append(name)
|
||||
|
||||
# 方法3：直接从 XML 文本中正则匹配 sheet name
|
||||
if not sheet_names:
|
||||
import re
|
||||
xml_str = content.decode('utf-8', errors='ignore')
|
||||
matches = re.findall(r'<sheet\s+[^>]*name=["\']([^"\']+)["\']', xml_str, re.IGNORECASE)
|
||||
if matches:
|
||||
sheet_names = matches
|
||||
logger.info(f"使用正则提取工作表: {sheet_names}")
|
||||
|
||||
logger.info(f"从 XML 提取工作表: {sheet_names}")
|
||||
return sheet_names
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"从 XML 提取工作表名称失败: {e}")
|
||||
return []
|
||||
@@ -356,6 +402,32 @@ class XlsxParser(BaseParser):
|
||||
import zipfile
|
||||
from xml.etree import ElementTree as ET
|
||||
|
||||
# 常见的命名空间
|
||||
COMMON_NAMESPACES = [
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2006/main',
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2005/main',
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2004/main',
|
||||
'http://schemas.openxmlformats.org/spreadsheetml/2003/main',
|
||||
]
|
||||
|
||||
def find_elements_with_ns(root, tag_name):
|
||||
"""灵活查找元素，支持任意命名空间"""
|
||||
results = []
|
||||
# 方法1：用固定命名空间
|
||||
for ns in COMMON_NAMESPACES:
|
||||
try:
|
||||
elems = root.findall(f'.//{{{ns}}}{tag_name}')
|
||||
if elems:
|
||||
results.extend(elems)
|
||||
except:
|
||||
pass
|
||||
# 方法2：不带命名空间查找
|
||||
if not results:
|
||||
for elem in root.iter():
|
||||
if elem.tag.endswith('}' + tag_name):
|
||||
results.append(elem)
|
||||
return results
|
||||
|
||||
with zipfile.ZipFile(file_path, 'r') as z:
|
||||
# 获取工作表名称
|
||||
sheet_names = self._extract_sheet_names_from_xml(file_path)
|
||||
@@ -366,57 +438,68 @@ class XlsxParser(BaseParser):
|
||||
target_sheet = sheet_name if sheet_name and sheet_name in sheet_names else sheet_names[0]
|
||||
sheet_index = sheet_names.index(target_sheet) + 1 # sheet1.xml, sheet2.xml, ...
|
||||
|
||||
# 读取 shared strings
|
||||
# 读取 shared strings - 尝试多种路径
|
||||
shared_strings = []
|
||||
if 'xl/sharedStrings.xml' in z.namelist():
|
||||
ss_content = z.read('xl/sharedStrings.xml')
|
||||
ss_paths = ['xl/sharedStrings.xml', 'xl\\sharedStrings.xml', 'sharedStrings.xml']
|
||||
for ss_path in ss_paths:
|
||||
if ss_path in z.namelist():
|
||||
try:
|
||||
ss_content = z.read(ss_path)
|
||||
ss_root = ET.fromstring(ss_content)
|
||||
ns = {'main': 'http://schemas.openxmlformats.org/spreadsheetml/2006/main'}
|
||||
for si in ss_root.findall('.//main:si', ns):
|
||||
t = si.find('.//main:t', ns)
|
||||
if t is not None:
|
||||
shared_strings.append(t.text or '')
|
||||
for si in find_elements_with_ns(ss_root, 'si'):
|
||||
t_elements = [c for c in si if c.tag.endswith('}t') or c.tag == 't']
|
||||
if t_elements:
|
||||
shared_strings.append(t_elements[0].text or '')
|
||||
else:
|
||||
shared_strings.append('')
|
||||
break
|
||||
except Exception as e:
|
||||
logger.warning(f"读取 sharedStrings 失败: {e}")
|
||||
|
||||
# 读取工作表
|
||||
sheet_file = f'xl/worksheets/sheet{sheet_index}.xml'
|
||||
if sheet_file not in z.namelist():
|
||||
raise ValueError(f"工作表文件 {sheet_file} 不存在")
|
||||
# 读取工作表 - 尝试多种可能的路径
|
||||
sheet_content = None
|
||||
sheet_paths = [
|
||||
f'xl/worksheets/sheet{sheet_index}.xml',
|
||||
f'xl\\worksheets\\sheet{sheet_index}.xml',
|
||||
f'worksheets/sheet{sheet_index}.xml',
|
||||
]
|
||||
for sp in sheet_paths:
|
||||
if sp in z.namelist():
|
||||
sheet_content = z.read(sp)
|
||||
break
|
||||
|
||||
if sheet_content is None:
|
||||
raise ValueError(f"工作表文件 sheet{sheet_index}.xml 不存在")
|
||||
|
||||
sheet_content = z.read(sheet_file)
|
||||
root = ET.fromstring(sheet_content)
|
||||
ns = {'main': 'http://schemas.openxmlformats.org/spreadsheetml/2006/main'}
|
||||
|
||||
# 收集所有行数据
|
||||
all_rows = []
|
||||
headers = {}
|
||||
|
||||
for row in root.findall('.//main:row', ns):
|
||||
for row in find_elements_with_ns(root, 'row'):
|
||||
row_idx = int(row.get('r', 0))
|
||||
row_cells = {}
|
||||
for cell in row.findall('main:c', ns):
|
||||
for cell in find_elements_with_ns(row, 'c'):
|
||||
cell_ref = cell.get('r', '')
|
||||
col_letters = ''.join(filter(str.isalpha, cell_ref))
|
||||
cell_type = cell.get('t', 'n')
|
||||
v = cell.find('main:v', ns)
|
||||
v_elements = find_elements_with_ns(cell, 'v')
|
||||
v = v_elements[0] if v_elements else None
|
||||
|
||||
if v is not None and v.text:
|
||||
if cell_type == 's':
|
||||
# shared string
|
||||
try:
|
||||
row_cells[col_letters] = shared_strings[int(v.text)]
|
||||
except (ValueError, IndexError):
|
||||
row_cells[col_letters] = v.text
|
||||
elif cell_type == 'b':
|
||||
# boolean
|
||||
row_cells[col_letters] = v.text == '1'
|
||||
else:
|
||||
row_cells[col_letters] = v.text
|
||||
else:
|
||||
row_cells[col_letters] = None
|
||||
|
||||
# 处理表头行
|
||||
if row_idx == header_row + 1:
|
||||
headers = {**row_cells}
|
||||
elif row_idx > header_row + 1:
|
||||
@@ -424,7 +507,6 @@ class XlsxParser(BaseParser):
|
||||
|
||||
# 构建 DataFrame
|
||||
if headers:
|
||||
# 按原始列顺序排列
|
||||
col_order = list(headers.keys())
|
||||
df = pd.DataFrame(all_rows)
|
||||
if not df.empty:
|
||||
|
||||
@@ -78,12 +78,19 @@ class TemplateFillService:
|
||||
fill_details = []
|
||||
|
||||
logger.info(f"开始填表: {len(template_fields)} 个字段, {len(source_doc_ids or [])} 个源文档")
|
||||
logger.info(f"source_doc_ids: {source_doc_ids}")
|
||||
logger.info(f"source_file_paths: {source_file_paths}")
|
||||
|
||||
# 1. 加载源文档内容
|
||||
source_docs = await self._load_source_documents(source_doc_ids, source_file_paths)
|
||||
|
||||
logger.info(f"加载了 {len(source_docs)} 个源文档")
|
||||
|
||||
# 打印每个加载的文档的详细信息
|
||||
for i, doc in enumerate(source_docs):
|
||||
logger.info(f" 文档[{i}]: id={doc.doc_id}, filename={doc.filename}, doc_type={doc.doc_type}")
|
||||
logger.info(f" content长度: {len(doc.content)}, structured_data keys: {list(doc.structured_data.keys()) if doc.structured_data else 'None'}")
|
||||
|
||||
if not source_docs:
|
||||
logger.warning("没有找到源文档，填表结果将全部为空")
|
||||
|
||||
@@ -158,14 +165,49 @@ class TemplateFillService:
|
||||
try:
|
||||
doc = await mongodb.get_document(doc_id)
|
||||
if doc:
|
||||
sd = doc.get("structured_data", {})
|
||||
sd_keys = list(sd.keys()) if sd else []
|
||||
logger.info(f"从MongoDB加载文档: {doc_id}, doc_type={doc.get('doc_type')}, structured_data keys={sd_keys}")
|
||||
|
||||
# 如果 structured_data 为空，但有 file_path，尝试重新解析文件
|
||||
doc_content = doc.get("content", "")
|
||||
if not sd or (not sd.get("tables") and not sd.get("headers") and not sd.get("rows")):
|
||||
file_path = doc.get("metadata", {}).get("file_path")
|
||||
if file_path:
|
||||
logger.info(f" structured_data 为空，尝试重新解析文件: {file_path}")
|
||||
try:
|
||||
parser = ParserFactory.get_parser(file_path)
|
||||
result = parser.parse(file_path)
|
||||
if result.success and result.data:
|
||||
if result.data.get("structured_data"):
|
||||
sd = result.data.get("structured_data")
|
||||
logger.info(f" 重新解析成功，structured_data keys: {list(sd.keys())}")
|
||||
elif result.data.get("tables"):
|
||||
sd = {"tables": result.data.get("tables", [])}
|
||||
logger.info(f" 使用 data.tables，tables数量: {len(sd.get('tables', []))}")
|
||||
elif result.data.get("rows"):
|
||||
sd = result.data
|
||||
logger.info(f" 使用 data.rows 格式")
|
||||
if result.data.get("content"):
|
||||
doc_content = result.data.get("content", "")
|
||||
else:
|
||||
logger.warning(f" 重新解析失败: {result.error if result else 'unknown'}")
|
||||
except Exception as parse_err:
|
||||
logger.error(f" 重新解析文件异常: {str(parse_err)}")
|
||||
|
||||
if sd.get("tables"):
|
||||
logger.info(f" tables数量: {len(sd.get('tables', []))}")
|
||||
if sd["tables"]:
|
||||
first_table = sd["tables"][0]
|
||||
logger.info(f" 第一表格: headers={first_table.get('headers', [])[:3]}..., rows数量={len(first_table.get('rows', []))}")
|
||||
|
||||
source_docs.append(SourceDocument(
|
||||
doc_id=doc_id,
|
||||
filename=doc.get("metadata", {}).get("original_filename", "unknown"),
|
||||
doc_type=doc.get("doc_type", "unknown"),
|
||||
content=doc.get("content", ""),
|
||||
structured_data=doc.get("structured_data", {})
|
||||
content=doc_content,
|
||||
structured_data=sd
|
||||
))
|
||||
logger.info(f"从MongoDB加载文档: {doc_id}")
|
||||
except Exception as e:
|
||||
logger.error(f"从MongoDB加载文档失败 {doc_id}: {str(e)}")
|
||||
|
||||
@@ -179,10 +221,48 @@ class TemplateFillService:
|
||||
# result.data 的结构取决于解析器类型:
|
||||
# - Excel 单 sheet: {columns: [...], rows: [...], row_count, column_count}
|
||||
# - Excel 多 sheet: {sheets: {sheet_name: {columns, rows, ...}}}
|
||||
# - Markdown: {content: "...", tables: [...], structured_data: {tables: [...]}}
|
||||
# - Word/TXT: {content: "...", structured_data: {...}}
|
||||
doc_data = result.data if result.data else {}
|
||||
doc_content = doc_data.get("content", "") if isinstance(doc_data, dict) else ""
|
||||
doc_structured = doc_data if isinstance(doc_data, dict) and "rows" in doc_data or isinstance(doc_data, dict) and "sheets" in doc_data else {}
|
||||
|
||||
# 检查并提取 structured_data
|
||||
doc_structured = {}
|
||||
if isinstance(doc_data, dict):
|
||||
logger.info(f"文档 {file_path} doc_data keys: {list(doc_data.keys())}")
|
||||
|
||||
# Excel 多 sheet
|
||||
if "sheets" in doc_data:
|
||||
doc_structured = doc_data
|
||||
logger.info(f" -> 使用 Excel 多 sheet 格式")
|
||||
# Excel 单 sheet 或有 rows 的格式
|
||||
elif "rows" in doc_data:
|
||||
doc_structured = doc_data
|
||||
logger.info(f" -> 使用 rows 格式，列数: {len(doc_data.get('columns', []))}")
|
||||
# Markdown 格式：tables 可能直接在 doc_data.tables 或在 structured_data.tables 中
|
||||
elif "tables" in doc_data and doc_data["tables"]:
|
||||
# Markdown: tables 直接在 doc_data 中
|
||||
tables = doc_data["tables"]
|
||||
first_table = tables[0]
|
||||
doc_structured = {
|
||||
"headers": first_table.get("headers", []),
|
||||
"rows": first_table.get("rows", [])
|
||||
}
|
||||
logger.info(f" -> 使用 doc_data.tables 格式，表头: {doc_structured.get('headers', [])[:5]}")
|
||||
elif "structured_data" in doc_data and isinstance(doc_data["structured_data"], dict):
|
||||
# Markdown: tables 在 structured_data 中
|
||||
tables = doc_data["structured_data"].get("tables", [])
|
||||
if tables:
|
||||
first_table = tables[0]
|
||||
doc_structured = {
|
||||
"headers": first_table.get("headers", []),
|
||||
"rows": first_table.get("rows", [])
|
||||
}
|
||||
logger.info(f" -> 使用 structured_data.tables 格式，表头: {doc_structured.get('headers', [])[:5]}")
|
||||
else:
|
||||
logger.warning(f" -> structured_data.tables 为空")
|
||||
else:
|
||||
logger.warning(f" -> 未识别的文档格式，无 structured_data")
|
||||
|
||||
source_docs.append(SourceDocument(
|
||||
doc_id=file_path,
|
||||
@@ -279,7 +359,7 @@ class TemplateFillService:
|
||||
response = await self.llm.chat(
|
||||
messages=messages,
|
||||
temperature=0.1,
|
||||
max_tokens=50000
|
||||
max_tokens=4000
|
||||
)
|
||||
|
||||
content = self.llm.extract_message_content(response)
|
||||
@@ -742,7 +822,7 @@ class TemplateFillService:
|
||||
|
||||
def _extract_values_from_structured_data(self, source_docs: List[SourceDocument], field_name: str) -> List[str]:
|
||||
"""
|
||||
从结构化数据（Excel rows）中直接提取指定列的值
|
||||
从结构化数据（Excel rows 或 Markdown tables）中直接提取指定列的值
|
||||
|
||||
适用于有 rows 结构的文档数据，无需 LLM 即可提取
|
||||
|
||||
@@ -754,10 +834,15 @@ class TemplateFillService:
|
||||
值列表，如果无法提取则返回空列表
|
||||
"""
|
||||
all_values = []
|
||||
logger.info(f"[_extract_values_from_structured_data] 开始提取字段: {field_name}")
|
||||
logger.info(f" source_docs 数量: {len(source_docs)}")
|
||||
|
||||
for doc in source_docs:
|
||||
for doc_idx, doc in enumerate(source_docs):
|
||||
# 尝试从 structured_data 中提取
|
||||
structured = doc.structured_data
|
||||
logger.info(f" 文档[{doc_idx}]: {doc.filename}, structured类型: {type(structured)}, 是否为空: {not bool(structured)}")
|
||||
if structured:
|
||||
logger.info(f" structured_data keys: {list(structured.keys())}")
|
||||
|
||||
if not structured:
|
||||
continue
|
||||
@@ -777,6 +862,33 @@ class TemplateFillService:
|
||||
if all_values:
|
||||
break
|
||||
|
||||
# 处理 Markdown 表格格式: {headers: [...], rows: [...], ...}
|
||||
elif structured.get("headers") and structured.get("rows"):
|
||||
headers = structured.get("headers", [])
|
||||
rows = structured.get("rows", [])
|
||||
values = self._extract_values_from_markdown_table(headers, rows, field_name)
|
||||
if values:
|
||||
all_values.extend(values)
|
||||
logger.info(f"从 Markdown 文档 {doc.filename} 提取到 {len(values)} 个值")
|
||||
break
|
||||
|
||||
# 处理 MongoDB 存储的 tables 格式: {tables: [{headers, rows, ...}, ...]}
|
||||
elif structured.get("tables") and isinstance(structured.get("tables"), list):
|
||||
tables = structured.get("tables", [])
|
||||
logger.info(f" 检测到 tables 格式，共 {len(tables)} 个表")
|
||||
for table_idx, table in enumerate(tables):
|
||||
if isinstance(table, dict):
|
||||
headers = table.get("headers", [])
|
||||
rows = table.get("rows", [])
|
||||
logger.info(f" 表格[{table_idx}]: headers={headers[:3]}..., rows数量={len(rows)}")
|
||||
values = self._extract_values_from_markdown_table(headers, rows, field_name)
|
||||
if values:
|
||||
all_values.extend(values)
|
||||
logger.info(f"从表格[{table_idx}] 提取到 {len(values)} 个值")
|
||||
break
|
||||
if all_values:
|
||||
break
|
||||
|
||||
# 处理单 sheet 格式: {columns: [...], rows: [...]}
|
||||
elif structured.get("rows"):
|
||||
columns = structured.get("columns", [])
|
||||
@@ -804,6 +916,100 @@ class TemplateFillService:
|
||||
|
||||
return all_values
|
||||
|
||||
def _extract_values_from_markdown_table(self, headers: List, rows: List, field_name: str) -> List[str]:
|
||||
"""
|
||||
从 Markdown 表格中提取指定列的值
|
||||
|
||||
Markdown 表格格式:
|
||||
- headers: ["col1", "col2", ...]
|
||||
- rows: [["val1", "val2", ...], ...]
|
||||
|
||||
Args:
|
||||
headers: 表头列表
|
||||
rows: 数据行列表
|
||||
field_name: 要提取的字段名
|
||||
|
||||
Returns:
|
||||
值列表
|
||||
"""
|
||||
if not rows or not headers:
|
||||
logger.warning(f"Markdown 表格为空: headers={headers}, rows={len(rows) if rows else 0}")
|
||||
return []
|
||||
|
||||
# 查找匹配的列索引 - 使用增强的匹配算法
|
||||
target_idx = self._find_best_matching_column(headers, field_name)
|
||||
|
||||
if target_idx is None:
|
||||
logger.warning(f"未找到匹配列: {field_name}, 可用表头: {headers}")
|
||||
return []
|
||||
|
||||
logger.info(f"列匹配成功: {field_name} -> {headers[target_idx]} (索引: {target_idx})")
|
||||
|
||||
values = []
|
||||
for row in rows:
|
||||
if isinstance(row, list) and target_idx < len(row):
|
||||
val = row[target_idx]
|
||||
else:
|
||||
val = ""
|
||||
values.append(self._format_value(val))
|
||||
|
||||
return values
|
||||
|
||||
def _find_best_matching_column(self, headers: List, field_name: str) -> Optional[int]:
|
||||
"""
|
||||
查找最佳匹配的列索引
|
||||
|
||||
使用多层匹配策略:
|
||||
1. 精确匹配（忽略大小写）
|
||||
2. 子字符串匹配（字段名在表头中，或表头在字段名中）
|
||||
3. 关键词重叠匹配（中文字符串分割后比对）
|
||||
|
||||
Args:
|
||||
headers: 表头列表
|
||||
field_name: 要匹配的字段名
|
||||
|
||||
Returns:
|
||||
匹配的列索引，找不到返回 None
|
||||
"""
|
||||
field_lower = field_name.lower().strip()
|
||||
field_keywords = set(field_lower.replace(" ", "").split())
|
||||
|
||||
best_match_idx = None
|
||||
best_match_score = 0
|
||||
|
||||
for idx, header in enumerate(headers):
|
||||
header_str = str(header).strip()
|
||||
header_lower = header_str.lower()
|
||||
|
||||
# 策略1: 精确匹配（忽略大小写）
|
||||
if header_lower == field_lower:
|
||||
return idx
|
||||
|
||||
# 策略2: 子字符串匹配
|
||||
if field_lower in header_lower or header_lower in field_lower:
|
||||
# 计算匹配分数（较长匹配更优先）
|
||||
score = max(len(field_lower), len(header_lower)) / min(len(field_lower) + 1, len(header_lower) + 1)
|
||||
if score > best_match_score:
|
||||
best_match_score = score
|
||||
best_match_idx = idx
|
||||
continue
|
||||
|
||||
# 策略3: 关键词重叠匹配（适用于中文）
|
||||
header_keywords = set(header_lower.replace(" ", "").split())
|
||||
overlap = field_keywords & header_keywords
|
||||
if overlap and len(overlap) > 0:
|
||||
score = len(overlap) / max(len(field_keywords), len(header_keywords), 1)
|
||||
if score > best_match_score:
|
||||
best_match_score = score
|
||||
best_match_idx = idx
|
||||
|
||||
# 只有当匹配分数超过阈值时才返回
|
||||
if best_match_score >= 0.3:
|
||||
logger.info(f"模糊匹配: {field_name} -> {headers[best_match_idx]} (分数: {best_match_score:.2f})")
|
||||
return best_match_idx
|
||||
|
||||
return None
|
||||
|
||||
def _extract_column_values(self, rows: List, columns: List, field_name: str) -> List[str]:
|
||||
"""
|
||||
从 rows 和 columns 中提取指定列的值
|
||||
@@ -819,30 +1025,70 @@ class TemplateFillService:
|
||||
if not rows or not columns:
|
||||
return []
|
||||
|
||||
# 查找匹配的列（模糊匹配）
|
||||
target_col = None
|
||||
for col in columns:
|
||||
col_str = str(col)
|
||||
if field_name.lower() in col_str.lower() or col_str.lower() in field_name.lower():
|
||||
target_col = col
|
||||
break
|
||||
# 使用增强的匹配算法查找最佳匹配的列索引
|
||||
target_idx = self._find_best_matching_column(columns, field_name)
|
||||
|
||||
if not target_col:
|
||||
if target_idx is None:
|
||||
logger.warning(f"未找到匹配列: {field_name}, 可用列: {columns}")
|
||||
return []
|
||||
|
||||
target_col = columns[target_idx]
|
||||
logger.info(f"列匹配成功: {field_name} -> {target_col} (索引: {target_idx})")
|
||||
|
||||
values = []
|
||||
for row in rows:
|
||||
if isinstance(row, dict):
|
||||
val = row.get(target_col, "")
|
||||
elif isinstance(row, list) and target_col in columns:
|
||||
val = row[columns.index(target_col)]
|
||||
elif isinstance(row, list) and target_idx < len(row):
|
||||
val = row[target_idx]
|
||||
else:
|
||||
val = ""
|
||||
values.append(str(val) if val is not None else "")
|
||||
values.append(self._format_value(val))
|
||||
|
||||
return values
|
||||
|
||||
def _format_value(self, val: Any) -> str:
|
||||
"""
|
||||
格式化值为字符串，保持原始格式
|
||||
|
||||
- 如果是浮点数但实际上等于整数，返回整数格式（如 3.0 -> "3"）
|
||||
- 如果是浮点数且有小数部分，保留小数（如 3.5 -> "3.5"）
|
||||
- 如果是整数，直接返回（如 3 -> "3"）
|
||||
- 其他类型直接转为字符串
|
||||
|
||||
Args:
|
||||
val: 原始值
|
||||
|
||||
Returns:
|
||||
格式化后的字符串
|
||||
"""
|
||||
if val is None:
|
||||
return ""
|
||||
|
||||
# 如果已经是字符串
|
||||
if isinstance(val, str):
|
||||
return val.strip()
|
||||
|
||||
# 如果是布尔值
|
||||
if isinstance(val, bool):
|
||||
return "true" if val else "false"
|
||||
|
||||
# 如果是数字
|
||||
if isinstance(val, (int, float)):
|
||||
# 检查是否是浮点数但等于整数
|
||||
if isinstance(val, float):
|
||||
# 检查是否是小数部分为0
|
||||
if val == int(val):
|
||||
return str(int(val))
|
||||
else:
|
||||
# 去除尾部多余的0，但保留必要的小数位
|
||||
formatted = f"{val:.10f}".rstrip('0').rstrip('.')
|
||||
return formatted
|
||||
else:
|
||||
return str(val)
|
||||
|
||||
return str(val)
|
||||
|
||||
def _extract_values_from_json(self, result) -> List[str]:
|
||||
"""
|
||||
从解析后的 JSON 对象/数组中提取值数组
|
||||
@@ -856,12 +1102,12 @@ class TemplateFillService:
|
||||
if isinstance(result, dict):
|
||||
# 优先找 values 数组
|
||||
if "values" in result and isinstance(result["values"], list):
|
||||
vals = [str(v).strip() for v in result["values"] if v and str(v).strip()]
|
||||
vals = [self._format_value(v).strip() for v in result["values"] if self._format_value(v).strip()]
|
||||
if vals:
|
||||
return vals
|
||||
# 尝试找 value 字段
|
||||
if "value" in result:
|
||||
val = str(result["value"]).strip()
|
||||
val = self._format_value(result["value"]).strip()
|
||||
if val:
|
||||
return [val]
|
||||
# 尝试找任何数组类型的键
|
||||
@@ -869,13 +1115,13 @@ class TemplateFillService:
|
||||
val = result[key]
|
||||
if isinstance(val, list) and len(val) > 0:
|
||||
if all(isinstance(v, (str, int, float, bool)) or v is None for v in val):
|
||||
vals = [str(v).strip() for v in val if v is not None and str(v).strip()]
|
||||
vals = [self._format_value(v).strip() for v in val if v is not None and self._format_value(v).strip()]
|
||||
if vals:
|
||||
return vals
|
||||
elif isinstance(val, (str, int, float, bool)):
|
||||
return [str(val).strip()]
|
||||
return [self._format_value(val).strip()]
|
||||
elif isinstance(result, list):
|
||||
vals = [str(v).strip() for v in result if v is not None and str(v).strip()]
|
||||
vals = [self._format_value(v).strip() for v in result if v is not None and self._format_value(v).strip()]
|
||||
if vals:
|
||||
return vals
|
||||
return []
|
||||
@@ -1012,15 +1258,15 @@ class TemplateFillService:
|
||||
if isinstance(parsed, dict):
|
||||
# 如果是 {"values": [...]} 格式，提取 values
|
||||
if "values" in parsed and isinstance(parsed["values"], list):
|
||||
return [str(v).strip() for v in parsed["values"] if v and str(v).strip()]
|
||||
return [self._format_value(v).strip() for v in parsed["values"] if self._format_value(v).strip()]
|
||||
# 如果是其他 dict 格式，尝试找 values 键
|
||||
for key in ["values", "value", "data", "result"]:
|
||||
if key in parsed and isinstance(parsed[key], list):
|
||||
return [str(v).strip() for v in parsed[key] if v and str(v).strip()]
|
||||
return [self._format_value(v).strip() for v in parsed[key] if self._format_value(v).strip()]
|
||||
elif key in parsed:
|
||||
return [str(parsed[key]).strip()]
|
||||
return [self._format_value(parsed[key]).strip()]
|
||||
elif isinstance(parsed, list):
|
||||
return [str(v).strip() for v in parsed if v and str(v).strip()]
|
||||
return [self._format_value(v).strip() for v in parsed if self._format_value(v).strip()]
|
||||
except (json.JSONDecodeError, TypeError):
|
||||
pass
|
||||
|
||||
@@ -1036,14 +1282,14 @@ class TemplateFillService:
|
||||
result = []
|
||||
for item in arr:
|
||||
if isinstance(item, dict) and "values" in item and isinstance(item["values"], list):
|
||||
result.extend([str(v).strip() for v in item["values"] if v and str(v).strip()])
|
||||
result.extend([self._format_value(v).strip() for v in item["values"] if self._format_value(v).strip()])
|
||||
elif isinstance(item, dict):
|
||||
result.append(str(item))
|
||||
else:
|
||||
result.append(str(item))
|
||||
result.append(self._format_value(item))
|
||||
if result:
|
||||
return result
|
||||
return [str(v).strip() for v in arr if v and str(v).strip()]
|
||||
return [self._format_value(v).strip() for v in arr if self._format_value(v).strip()]
|
||||
except:
|
||||
pass
|
||||
|
||||
@@ -1134,27 +1380,37 @@ class TemplateFillService:
|
||||
hint_text = f"{user_hint}。{hint_text}"
|
||||
|
||||
# 构建针对字段提取的提示词
|
||||
prompt = f"""你是一个专业的数据提取专家。请从以下文档内容中提取与"{field.name}"相关的所有数据。
|
||||
prompt = f"""你是一个专业的数据提取专家。请从以下文档内容中提取与"{field.name}"完全匹配的数据。
|
||||
|
||||
字段提示: {hint_text}
|
||||
【重要】字段名: "{field.name}"
|
||||
【重要】字段提示: {hint_text}
|
||||
|
||||
请严格按照以下步骤操作：
|
||||
1. 在文档中搜索与"{field.name}"完全相同或高度相关的关键词
|
||||
2. 找到后，提取该关键词后的数值（注意：只要数值，不要单位）
|
||||
3. 如果是表格中的数据，直接提取该单元格的数值
|
||||
4. 如果是段落描述，在关键词附近找数值
|
||||
|
||||
【重要】返回值规则：
|
||||
- 只返回纯数值，不要单位（如 "4.9" 而不是 "4.9万亿元"）
|
||||
- 如果原文是"4.9万亿元"，返回 "4.9"
|
||||
- 如果原文是"144000万册"，返回 "144000"
|
||||
- 如果是百分比如"增长7.7%"，返回 "7.7"
|
||||
- 如果没有找到完全匹配的数据，返回空数组
|
||||
|
||||
文档内容：
|
||||
{doc.content[:8000] if doc.content else ""}
|
||||
|
||||
请完成以下任务：
|
||||
1. 仔细阅读文档，找出所有与"{field.name}"相关的数据
|
||||
2. 如果文档中有表格数据，提取表格中的对应列值
|
||||
3. 如果文档中是段落描述，提取其中的关键数值或结论
|
||||
4. 返回提取的所有值（可能多个，用数组存储）
|
||||
{doc.content[:10000] if doc.content else ""}
|
||||
|
||||
请用严格的 JSON 格式返回：
|
||||
{{
|
||||
"values": ["值1", "值2", ...],
|
||||
"values": ["值1", "值2", ...], // 只填数值，不要单位
|
||||
"source": "数据来源说明",
|
||||
"confidence": 0.0到1.0之间的置信度
|
||||
}}
|
||||
|
||||
如果没有找到相关数据，返回空数组 values: []"""
|
||||
示例：
|
||||
- 如果字段是"图书馆总藏量（万册）"且文档说"图书总藏量14.4亿册"，返回 values: ["144000"]
|
||||
- 如果字段是"国内旅游收入（亿元）"且文档说"国内旅游收入4.9万亿元"，返回 values: ["49000"]"""
|
||||
|
||||
messages = [
|
||||
{"role": "system", "content": "你是一个专业的数据提取助手，擅长从政府统计公报等文档中提取数据。请严格按JSON格式输出。"},
|
||||
@@ -1164,7 +1420,7 @@ class TemplateFillService:
|
||||
response = await self.llm.chat(
|
||||
messages=messages,
|
||||
temperature=0.1,
|
||||
max_tokens=5000
|
||||
max_tokens=4000
|
||||
)
|
||||
|
||||
content = self.llm.extract_message_content(response)
|
||||
|
||||
@@ -21,6 +21,7 @@ interface TemplateFillState {
|
||||
templateFields: TemplateField[];
|
||||
sourceFiles: SourceFile[];
|
||||
sourceFilePaths: string[];
|
||||
sourceDocIds: string[];
|
||||
templateId: string;
|
||||
filledResult: any;
|
||||
setStep: (step: Step) => void;
|
||||
@@ -30,6 +31,9 @@ interface TemplateFillState {
|
||||
addSourceFiles: (files: SourceFile[]) => void;
|
||||
removeSourceFile: (index: number) => void;
|
||||
setSourceFilePaths: (paths: string[]) => void;
|
||||
setSourceDocIds: (ids: string[]) => void;
|
||||
addSourceDocId: (id: string) => void;
|
||||
removeSourceDocId: (id: string) => void;
|
||||
setTemplateId: (id: string) => void;
|
||||
setFilledResult: (result: any) => void;
|
||||
reset: () => void;
|
||||
@@ -41,6 +45,7 @@ const initialState = {
|
||||
templateFields: [],
|
||||
sourceFiles: [],
|
||||
sourceFilePaths: [],
|
||||
sourceDocIds: [],
|
||||
templateId: '',
|
||||
filledResult: null,
|
||||
setStep: () => {},
|
||||
@@ -50,6 +55,9 @@ const initialState = {
|
||||
addSourceFiles: () => {},
|
||||
removeSourceFile: () => {},
|
||||
setSourceFilePaths: () => {},
|
||||
setSourceDocIds: () => {},
|
||||
addSourceDocId: () => {},
|
||||
removeSourceDocId: () => {},
|
||||
setTemplateId: () => {},
|
||||
setFilledResult: () => {},
|
||||
reset: () => {},
|
||||
@@ -63,6 +71,7 @@ export const TemplateFillProvider: React.FC<{ children: ReactNode }> = ({ childr
|
||||
const [templateFields, setTemplateFields] = useState<TemplateField[]>([]);
|
||||
const [sourceFiles, setSourceFiles] = useState<SourceFile[]>([]);
|
||||
const [sourceFilePaths, setSourceFilePaths] = useState<string[]>([]);
|
||||
const [sourceDocIds, setSourceDocIds] = useState<string[]>([]);
|
||||
const [templateId, setTemplateId] = useState<string>('');
|
||||
const [filledResult, setFilledResult] = useState<any>(null);
|
||||
|
||||
@@ -74,12 +83,21 @@ export const TemplateFillProvider: React.FC<{ children: ReactNode }> = ({ childr
|
||||
setSourceFiles(prev => prev.filter((_, i) => i !== index));
|
||||
};
|
||||
|
||||
const addSourceDocId = (id: string) => {
|
||||
setSourceDocIds(prev => prev.includes(id) ? prev : [...prev, id]);
|
||||
};
|
||||
|
||||
const removeSourceDocId = (id: string) => {
|
||||
setSourceDocIds(prev => prev.filter(docId => docId !== id));
|
||||
};
|
||||
|
||||
const reset = () => {
|
||||
setStep('upload');
|
||||
setTemplateFile(null);
|
||||
setTemplateFields([]);
|
||||
setSourceFiles([]);
|
||||
setSourceFilePaths([]);
|
||||
setSourceDocIds([]);
|
||||
setTemplateId('');
|
||||
setFilledResult(null);
|
||||
};
|
||||
@@ -92,6 +110,7 @@ export const TemplateFillProvider: React.FC<{ children: ReactNode }> = ({ childr
|
||||
templateFields,
|
||||
sourceFiles,
|
||||
sourceFilePaths,
|
||||
sourceDocIds,
|
||||
templateId,
|
||||
filledResult,
|
||||
setStep,
|
||||
@@ -101,6 +120,9 @@ export const TemplateFillProvider: React.FC<{ children: ReactNode }> = ({ childr
|
||||
addSourceFiles,
|
||||
removeSourceFile,
|
||||
setSourceFilePaths,
|
||||
setSourceDocIds,
|
||||
addSourceDocId,
|
||||
removeSourceDocId,
|
||||
setTemplateId,
|
||||
setFilledResult,
|
||||
reset,
|
||||
|
||||
@@ -60,6 +60,7 @@ const TemplateFill: React.FC = () => {
|
||||
templateFields, setTemplateFields,
|
||||
sourceFiles, setSourceFiles, addSourceFiles, removeSourceFile,
|
||||
sourceFilePaths, setSourceFilePaths,
|
||||
sourceDocIds, setSourceDocIds, addSourceDocId, removeSourceDocId,
|
||||
templateId, setTemplateId,
|
||||
filledResult, setFilledResult,
|
||||
reset
|
||||
@@ -68,6 +69,9 @@ const TemplateFill: React.FC = () => {
|
||||
const [loading, setLoading] = useState(false);
|
||||
const [previewDoc, setPreviewDoc] = useState<{ name: string; content: string } | null>(null);
|
||||
const [previewOpen, setPreviewOpen] = useState(false);
|
||||
const [sourceMode, setSourceMode] = useState<'upload' | 'select'>('upload');
|
||||
const [uploadedDocuments, setUploadedDocuments] = useState<DocumentItem[]>([]);
|
||||
const [docsLoading, setDocsLoading] = useState(false);
|
||||
|
||||
// 模板拖拽
|
||||
const onTemplateDrop = useCallback((acceptedFiles: File[]) => {
|
||||
@@ -109,15 +113,92 @@ const TemplateFill: React.FC = () => {
|
||||
multiple: true
|
||||
});
|
||||
|
||||
// 加载已上传文档
|
||||
const loadUploadedDocuments = useCallback(async () => {
|
||||
setDocsLoading(true);
|
||||
try {
|
||||
const result = await backendApi.getDocuments(undefined, 100);
|
||||
if (result.success) {
|
||||
// 过滤可作为数据源的文档类型
|
||||
const docs = (result.documents || []).filter((d: DocumentItem) =>
|
||||
['docx', 'md', 'txt', 'xlsx', 'xls'].includes(d.doc_type)
|
||||
);
|
||||
setUploadedDocuments(docs);
|
||||
}
|
||||
} catch (err: any) {
|
||||
console.error('加载文档失败:', err);
|
||||
} finally {
|
||||
setDocsLoading(false);
|
||||
}
|
||||
}, []);
|
||||
|
||||
// 删除文档
|
||||
const handleDeleteDocument = async (docId: string, e: React.MouseEvent) => {
|
||||
e.stopPropagation();
|
||||
if (!confirm('确定要删除该文档吗？')) return;
|
||||
try {
|
||||
const result = await backendApi.deleteDocument(docId);
|
||||
if (result.success) {
|
||||
setUploadedDocuments(prev => prev.filter(d => d.doc_id !== docId));
|
||||
removeSourceDocId(docId);
|
||||
toast.success('文档已删除');
|
||||
} else {
|
||||
toast.error(result.message || '删除失败');
|
||||
}
|
||||
} catch (err: any) {
|
||||
toast.error('删除失败: ' + (err.message || '未知错误'));
|
||||
}
|
||||
};
|
||||
|
||||
useEffect(() => {
|
||||
if (sourceMode === 'select') {
|
||||
loadUploadedDocuments();
|
||||
}
|
||||
}, [sourceMode, loadUploadedDocuments]);
|
||||
|
||||
const handleJointUploadAndFill = async () => {
|
||||
if (!templateFile) {
|
||||
toast.error('请先上传模板文件');
|
||||
return;
|
||||
}
|
||||
|
||||
// 检查是否选择了数据源
|
||||
if (sourceMode === 'upload' && sourceFiles.length === 0) {
|
||||
toast.error('请上传源文档或从已上传文档中选择');
|
||||
return;
|
||||
}
|
||||
if (sourceMode === 'select' && sourceDocIds.length === 0) {
|
||||
toast.error('请选择源文档');
|
||||
return;
|
||||
}
|
||||
|
||||
setLoading(true);
|
||||
|
||||
try {
|
||||
if (sourceMode === 'select') {
|
||||
// 使用已上传文档作为数据源
|
||||
const result = await backendApi.uploadTemplate(templateFile);
|
||||
|
||||
if (result.success) {
|
||||
setTemplateFields(result.fields || []);
|
||||
setTemplateId(result.template_id || 'temp');
|
||||
toast.success('开始智能填表');
|
||||
setStep('filling');
|
||||
|
||||
// 使用 source_doc_ids 进行填表
|
||||
const fillResult = await backendApi.fillTemplate(
|
||||
result.template_id || 'temp',
|
||||
result.fields || [],
|
||||
sourceDocIds,
|
||||
[],
|
||||
'请从以下文档中提取相关信息填写表格'
|
||||
);
|
||||
|
||||
setFilledResult(fillResult);
|
||||
setStep('preview');
|
||||
toast.success('表格填写完成');
|
||||
}
|
||||
} else {
|
||||
// 使用联合上传API
|
||||
const result = await backendApi.uploadTemplateAndSources(
|
||||
templateFile,
|
||||
@@ -135,7 +216,7 @@ const TemplateFill: React.FC = () => {
|
||||
const fillResult = await backendApi.fillTemplate(
|
||||
result.template_id,
|
||||
result.fields || [],
|
||||
[], // 使用 source_file_paths 而非 source_doc_ids
|
||||
[],
|
||||
result.source_file_paths || [],
|
||||
'请从以下文档中提取相关信息填写表格'
|
||||
);
|
||||
@@ -144,6 +225,7 @@ const TemplateFill: React.FC = () => {
|
||||
setStep('preview');
|
||||
toast.success('表格填写完成');
|
||||
}
|
||||
}
|
||||
} catch (err: any) {
|
||||
toast.error('处理失败: ' + (err.message || '未知错误'));
|
||||
} finally {
|
||||
@@ -264,10 +346,31 @@ const TemplateFill: React.FC = () => {
|
||||
源文档
|
||||
</CardTitle>
|
||||
<CardDescription>
|
||||
上传包含数据的源文档（支持多选），可同时上传多个文件
|
||||
选择包含数据的源文档作为填表依据
|
||||
</CardDescription>
|
||||
{/* Source Mode Tabs */}
|
||||
<div className="flex gap-2 mt-2">
|
||||
<Button
|
||||
variant={sourceMode === 'upload' ? 'default' : 'outline'}
|
||||
size="sm"
|
||||
onClick={() => setSourceMode('upload')}
|
||||
>
|
||||
<Upload size={14} className="mr-1" />
|
||||
上传文件
|
||||
</Button>
|
||||
<Button
|
||||
variant={sourceMode === 'select' ? 'default' : 'outline'}
|
||||
size="sm"
|
||||
onClick={() => setSourceMode('select')}
|
||||
>
|
||||
<Files size={14} className="mr-1" />
|
||||
从文档中心选择
|
||||
</Button>
|
||||
</div>
|
||||
</CardHeader>
|
||||
<CardContent>
|
||||
{sourceMode === 'upload' ? (
|
||||
<>
|
||||
<div
|
||||
{...getSourceProps()}
|
||||
className={cn(
|
||||
@@ -306,6 +409,69 @@ const TemplateFill: React.FC = () => {
|
||||
))}
|
||||
</div>
|
||||
)}
|
||||
</>
|
||||
) : (
|
||||
<>
|
||||
{/* Uploaded Documents Selection */}
|
||||
{docsLoading ? (
|
||||
<div className="space-y-2">
|
||||
{[1, 2, 3].map(i => (
|
||||
<Skeleton key={i} className="h-16 w-full rounded-xl" />
|
||||
))}
|
||||
</div>
|
||||
) : uploadedDocuments.length > 0 ? (
|
||||
<div className="space-y-2 max-h-[300px] overflow-y-auto">
|
||||
{uploadedDocuments.map((doc) => (
|
||||
<div
|
||||
key={doc.doc_id}
|
||||
className={cn(
|
||||
"flex items-center gap-3 p-3 rounded-xl border-2 transition-all cursor-pointer",
|
||||
sourceDocIds.includes(doc.doc_id)
|
||||
? "border-primary bg-primary/5"
|
||||
: "border-border hover:bg-muted/30"
|
||||
)}
|
||||
onClick={() => {
|
||||
if (sourceDocIds.includes(doc.doc_id)) {
|
||||
removeSourceDocId(doc.doc_id);
|
||||
} else {
|
||||
addSourceDocId(doc.doc_id);
|
||||
}
|
||||
}}
|
||||
>
|
||||
<div className={cn(
|
||||
"w-6 h-6 rounded-md border-2 flex items-center justify-center transition-all shrink-0",
|
||||
sourceDocIds.includes(doc.doc_id)
|
||||
? "border-primary bg-primary text-white"
|
||||
: "border-muted-foreground/30"
|
||||
)}>
|
||||
{sourceDocIds.includes(doc.doc_id) && <CheckCircle2 size={14} />}
|
||||
</div>
|
||||
{getFileIcon(doc.original_filename)}
|
||||
<div className="flex-1 min-w-0">
|
||||
<p className="text-sm font-medium truncate">{doc.original_filename}</p>
|
||||
<p className="text-xs text-muted-foreground">
|
||||
{doc.doc_type.toUpperCase()} • {format(new Date(doc.created_at), 'yyyy-MM-dd')}
|
||||
</p>
|
||||
</div>
|
||||
<Button
|
||||
variant="ghost"
|
||||
size="sm"
|
||||
onClick={(e) => handleDeleteDocument(doc.doc_id, e)}
|
||||
className="shrink-0"
|
||||
>
|
||||
<Trash2 size={14} className="text-red-500" />
|
||||
</Button>
|
||||
</div>
|
||||
))}
|
||||
</div>
|
||||
) : (
|
||||
<div className="text-center py-8 text-muted-foreground">
|
||||
<Files size={32} className="mx-auto mb-2 opacity-30" />
|
||||
<p className="text-sm">暂无可用的已上传文档</p>
|
||||
</div>
|
||||
)}
|
||||
</>
|
||||
)}
|
||||
</CardContent>
|
||||
</Card>
|
||||
|
||||
|
||||
Reference in New Issue
Block a user