from io import BytesIO
from typing import Optional
from sqlalchemy.orm import Session
from datetime import datetime

from docx import Document
from docx.shared import Inches, Pt, RGBColor
from docx.enum.text import WD_ALIGN_PARAGRAPH

from app.models.project import ClassificationProject, ClassificationResult
from app.models.classification import Category, DataLevel


def generate_classification_report(db: Session, project_id: int) -> bytes:
    """Generate a Word report for a classification project."""
    project = db.query(ClassificationProject).filter(ClassificationProject.id == project_id).first()
    if not project:
        raise ValueError("项目不存在")

    doc = Document()

    # Title
    title = doc.add_heading('数据分类分级项目报告', 0)
    title.alignment = WD_ALIGN_PARAGRAPH.CENTER

    # Basic info
    doc.add_heading('一、项目基本信息', level=1)
    info_table = doc.add_table(rows=4, cols=2)
    info_table.style = 'Light Grid Accent 1'
    info_data = [
        ('项目名称', project.name),
        ('报告生成时间', datetime.now().strftime('%Y-%m-%d %H:%M:%S')),
        ('项目状态', project.status),
        ('模板版本', project.template.version if project.template else 'N/A'),
    ]
    for i, (k, v) in enumerate(info_data):
        info_table.rows[i].cells[0].text = k
        info_table.rows[i].cells[1].text = str(v)

    # Statistics
    doc.add_heading('二、分类分级统计', level=1)
    results = db.query(ClassificationResult).filter(ClassificationResult.project_id == project_id).all()

    total = len(results)
    auto_count = sum(1 for r in results if r.source == 'auto')
    manual_count = sum(1 for r in results if r.source == 'manual')

    level_stats = {}
    for r in results:
        if r.level:
            level_stats[r.level.name] = level_stats.get(r.level.name, 0) + 1

    doc.add_paragraph(f'总字段数: {total}')
    doc.add_paragraph(f'自动识别: {auto_count}')
    doc.add_paragraph(f'人工打标: {manual_count}')

    doc.add_heading('三、分级分布', level=1)
    level_table = doc.add_table(rows=1, cols=3)
    level_table.style = 'Light Grid Accent 1'
    hdr_cells = level_table.rows[0].cells
    hdr_cells[0].text = '分级'
    hdr_cells[1].text = '数量'
    hdr_cells[2].text = '占比'
    for level_name, count in sorted(level_stats.items(), key=lambda x: -x[1]):
        row_cells = level_table.add_row().cells
        row_cells[0].text = level_name
        row_cells[1].text = str(count)
        row_cells[2].text = f'{count / total * 100:.1f}%' if total > 0 else '0%'

    # High risk data
    doc.add_heading('四、高敏感数据清单（L4/L5）', level=1)
    high_risk = [r for r in results if r.level and r.level.code in ('L4', 'L5')]
    if high_risk:
        risk_table = doc.add_table(rows=1, cols=5)
        risk_table.style = 'Light Grid Accent 1'
        hdr = risk_table.rows[0].cells
        hdr[0].text = '字段名'
        hdr[1].text = '所属表'
        hdr[2].text = '分类'
        hdr[3].text = '分级'
        hdr[4].text = '来源'
        for r in high_risk[:100]:  # limit to 100 rows
            row = risk_table.add_row().cells
            row[0].text = r.column.name if r.column else 'N/A'
            row[1].text = r.column.table.name if r.column and r.column.table else 'N/A'
            row[2].text = r.category.name if r.category else 'N/A'
            row[3].text = r.level.name if r.level else 'N/A'
            row[4].text = '自动' if r.source == 'auto' else '人工'
    else:
        doc.add_paragraph('暂无L4/L5级高敏感数据。')

    # Save to bytes
    buffer = BytesIO()
    doc.save(buffer)
    buffer.seek(0)
    return buffer.read()