批量提取word文档标题

25-04-08     slbcun     959℃     0
import os
import sys
from collections import defaultdict
from docx import Document
from PyQt5.QtWidgets import (QApplication, QMainWindow, QVBoxLayout, QHBoxLayout, QWidget, 
                            QLabel, QPushButton, QTextEdit, QFileDialog, QSpinBox, 
                            QGroupBox, QProgressBar)
from PyQt5.QtCore import Qt
from PyQt5.QtGui import QFont, QIcon
 
class TitleExtractorApp(QMainWindow):
    def __init__(self):
        super().__init__()
        self.setWindowTitle("Word文档标题提取工具")
        self.setWindowIcon(QIcon('icon.png'))  # 请准备一个图标文件或删除这行
        self.setGeometry(100, 100, 800, 600)
         
        # 设置主窗口样式
        self.setStyleSheet("""
            QMainWindow {
                background-color: #f5f5f5;
            }
            QGroupBox {
                border: 1px solid #ccc;
                border-radius: 5px;
                margin-top: 10px;
                padding-top: 15px;
                font-size: 14px;
            }
            QGroupBox::title {
                subcontrol-origin: margin;
                left: 10px;
                padding: 0 3px;
            }
            QPushButton {
                background-color: #4CAF50;
                border: none;
                color: white;
                padding: 8px 16px;
                text-align: center;
                text-decoration: none;
                font-size: 14px;
                margin: 4px 2px;
                border-radius: 4px;
            }
            QPushButton:hover {
                background-color: #45a049;
            }
            QPushButton:pressed {
                background-color: #3e8e41;
            }
            QTextEdit {
                border: 1px solid #ccc;
                border-radius: 4px;
                padding: 8px;
                font-family: 'Segoe UI', Arial, sans-serif;
            }
            QSpinBox {
                padding: 5px;
                font-size: 14px;
            }
        """)
         
        self.init_ui()
         
    def init_ui(self):
        # 主布局
        main_widget = QWidget()
        main_layout = QVBoxLayout()
         
        # 标题
        title_label = QLabel("Word文档标题提取工具")
        title_label.setFont(QFont('Arial', 16, QFont.Bold))
        title_label.setAlignment(Qt.AlignCenter)
        title_label.setStyleSheet("color: #333; margin-bottom: 20px;")
         
        # 设置组
        settings_group = QGroupBox("提取设置")
        settings_layout = QHBoxLayout()
         
        # 文件夹选择
        folder_layout = QVBoxLayout()
        self.folder_label = QLabel("未选择文件夹")
        self.folder_label.setStyleSheet("color: #666;")
        browse_button = QPushButton("选择文件夹")
        browse_button.clicked.connect(self.select_folder)
        folder_layout.addWidget(QLabel("文档文件夹:"))
        folder_layout.addWidget(self.folder_label)
        folder_layout.addWidget(browse_button)
         
        # 标题级别设置
        level_layout = QVBoxLayout()
        level_label = QLabel("最大标题级别:")
        self.level_spin = QSpinBox()
        self.level_spin.setRange(1, 6)
        self.level_spin.setValue(3)
        level_layout.addWidget(level_label)
        level_layout.addWidget(self.level_spin)
         
        # 进度条
        self.progress_bar = QProgressBar()
        self.progress_bar.setRange(0, 100)
        self.progress_bar.setValue(0)
        self.progress_bar.setTextVisible(False)
         
        # 添加到设置组
        settings_layout.addLayout(folder_layout, 70)
        settings_layout.addLayout(level_layout, 30)
        settings_group.setLayout(settings_layout)
         
        # 操作按钮
        button_layout = QHBoxLayout()
        extract_button = QPushButton("提取标题")
        extract_button.clicked.connect(self.extract_titles)
        extract_button.setStyleSheet("background-color: #2196F3;")
         
        save_button = QPushButton("保存结果")
        save_button.clicked.connect(self.save_results)
        save_button.setStyleSheet("background-color: #FF9800;")
         
        clear_button = QPushButton("清空结果")
        clear_button.clicked.connect(self.clear_results)
        clear_button.setStyleSheet("background-color: #f44336;")
         
        button_layout.addWidget(extract_button)
        button_layout.addWidget(save_button)
        button_layout.addWidget(clear_button)
         
        # 结果显示
        result_group = QGroupBox("提取结果")
        result_layout = QVBoxLayout()
        self.result_text = QTextEdit()
        self.result_text.setReadOnly(True)
        result_layout.addWidget(self.result_text)
        result_group.setLayout(result_layout)
         
        # 组装主布局
        main_layout.addWidget(title_label)
        main_layout.addWidget(settings_group)
        main_layout.addWidget(self.progress_bar)
        main_layout.addLayout(button_layout)
        main_layout.addWidget(result_group)
         
        main_widget.setLayout(main_layout)
        self.setCentralWidget(main_widget)
         
        # 状态栏
        self.statusBar().showMessage("准备就绪")
         
        # 初始化变量
        self.selected_folder = ""
        self.extracted_data = {}
     
    def select_folder(self):
        folder = QFileDialog.getExistingDirectory(self, "选择包含Word文档的文件夹")
        if folder:
            self.selected_folder = folder
            self.folder_label.setText(folder)
            self.statusBar().showMessage(f"已选择文件夹: {folder}")
     
    def extract_titles(self):
        if not self.selected_folder:
            self.statusBar().showMessage("请先选择文件夹!", 3000)
            return
             
        max_level = self.level_spin.value()
        self.result_text.clear()
        self.extracted_data = {}
        file_count = 0
         
        # 统计文件数量用于进度条
        total_files = sum(1 for _, _, files in os.walk(self.selected_folder) 
                         for f in files if f.endswith('.docx'))
        if total_files == 0:
            self.statusBar().showMessage("所选文件夹中没有找到Word文档!", 3000)
            return
             
        processed_files = 0
         
        for root, _, files in os.walk(self.selected_folder):
            for filename in files:
                if filename.endswith('.docx'):
                    file_path = os.path.join(root, filename)
                    try:
                        doc = Document(file_path)
                        file_data = defaultdict(list)
                         
                        for p in doc.paragraphs:
                            if p.style.name.startswith('Heading'):
                                try:
                                    level = int(p.style.name.split()[1])
                                    if level <= max_level:
                                        file_data[level].append(p.text)
                                except (IndexError, ValueError):
                                    continue
                         
                        if file_data:
                            self.extracted_data[filename] = file_data
                             
                        processed_files += 1
                        progress = int((processed_files / total_files) * 100)
                        self.progress_bar.setValue(progress)
                         
                    except Exception as e:
                        self.result_text.append(f"处理文件 {filename} 时出错: {str(e)}\n")
         
        # 显示结果
        self.display_results()
        self.progress_bar.setValue(100)
        self.statusBar().showMessage(f"提取完成!共处理 {processed_files} 个文件", 5000)
     
    def display_results(self):
        self.result_text.clear()
        if not self.extracted_data:
            self.result_text.append("没有提取到任何标题数据")
            return
             
        for filename, levels in self.extracted_data.items():
            self.result_text.append(f"=== {filename} ===")
            for level in sorted(levels.keys()):
                self.result_text.append(f"\n[标题 {level}]")
                for i, title in enumerate(levels[level], 1):
                    self.result_text.append(f"{i}. {title}")
            self.result_text.append("\n")
     
    def save_results(self):
        if not self.extracted_data:
            self.statusBar().showMessage("没有可保存的数据!", 3000)
            return
             
        file_path, _ = QFileDialog.getSaveFileName(self, "保存结果", "", "文本文件 (*.txt)")
        if file_path:
            try:
                with open(file_path, 'w', encoding='utf-8') as f:
                    for filename, levels in self.extracted_data.items():
                        f.write(f"=== {filename} ===\n")
                        for level in sorted(levels.keys()):
                            f.write(f"\n[标题 {level}]\n")
                            for i, title in enumerate(levels[level], 1):
                                f.write(f"{i}. {title}\n")
                        f.write("\n")
                self.statusBar().showMessage(f"结果已保存到: {file_path}", 5000)
            except Exception as e:
                self.statusBar().showMessage(f"保存失败: {str(e)}", 5000)
     
    def clear_results(self):
        self.result_text.clear()
        self.extracted_data = {}
        self.progress_bar.setValue(0)
        self.statusBar().showMessage("已清空结果", 3000)
 
if __name__ == "__main__":
    app = QApplication(sys.argv)
     
    # 设置全局字体
    font = QFont()
    font.setFamily("Segoe UI")
    font.setPointSize(10)
    app.setFont(font)
     
    window = TitleExtractorApp()
    window.show()
    sys.exit(app.exec_())


分享 收藏
发表我的评论

表情

共0条评论
  • 这篇文章还没有收到评论,赶紧来抢沙发吧~