import json
import re
import os

# Load OCR results
with open(r"D:\workspace\membank\售价\ocr_results.json", "r", encoding="utf-8") as f:
    all_results = json.load(f)

def extract_sku_v4(text_lines):
    """
    Robust extraction:
    Pattern in PDD backend: SKU_name ... [price1] ... [5.00] ... [price2]
    where price2 = price1 - 5
    """
    results = []

    # Build full text with confidences
    items = [(t[0].strip(), float(t[1]) if t[1] else 0.0) for t in text_lines]

    i = 0
    while i < len(items):
        text, conf = items[i]

        # Check for price patterns - find lines with two prices and 5.00 in between
        # Look ahead for price patterns
        prices_in_window = []
        window_items = []

        for j in range(i, min(i + 8, len(items))):
            t = items[j][0]
            nums = re.findall(r'(\d{2,4}\.\d{2})', t)
            if nums:
                prices_in_window.extend([float(n) for n in nums])
                window_items.append((j, t, [float(n) for n in nums]))

        if not prices_in_window:
            i += 1
            continue

        prices_set = set(round(p, 2) for p in prices_in_window)

        # Try to find original price and discounted price (diff = 5)
        found_prices = []
        for p in sorted(prices_set):
            if round(p - 5, 2) in prices_set and p > 15:
                found_prices.append((round(p - 5, 2), p))  # (discounted, original)
            elif round(p + 5, 2) in prices_set and p > 15:
                found_prices.append((p, round(p + 5, 2)))

        if not found_prices:
            i += 1
            continue

        # Now find the SKU name - look backwards from the price
        # The SKU name is typically 1-3 lines before the price
        sku_name = None
        for j in range(i-1, max(0, i-5), -1):
            t = items[j][0]
            # Must contain CPU or platform identifiers
            if re.search(r'(J1900|j1900|J3160|j3160|i3|I3|i5|I5|1037[Uu]|5005[Uu]|N5105|16G[Bb]|预装系统|双盘位|单盘位|三盘位|四盘位|黑裙|黑群晖|飞牛|蜗牛|Nas|NAS|盘位)', t):
                # Not a pure UI element
                if not re.match(r'^(商品|营销|官方|拼多多|商家|查看|修改|活动|价格|关注|折扣|券后|ID)', t):
                    sku_name = t
                    break

        if sku_name:
            for discounted, original in found_prices:
                results.append({
                    'sku': sku_name,
                    'price': discounted,
                    'original_price': original,
                    'confidence': conf
                })

        i += 1

    return results


def normalize_sku_key(name):
    """Create a normalized key for deduplication by extracting core identifiers"""
    name = name.strip()
    # Extract CPU type
    cpu = ''
    cpu_match = re.search(r'(J1900|j1900|J3160|j3160|i3|I3|i5|I5|1037[Uu]|5005[Uu]|N5105|N5095|J3355)', name)
    if cpu_match:
        cpu = cpu_match.group(1).upper().replace('J1900','J1900').replace('J3160','J3160')

    # Extract memory
    mem = ''
    mem_match = re.search(r'(\d+[Gg])[+]?(\d+[Gg])?', name)
    if mem_match:
        mem = mem_match.group(0).upper()

    # Extract form factor
    form = ''
    if '单盘' in name or '单盤' in name:
        form = '单盘'
    elif '双盘' in name or '雙盤' in name:
        form = '双盘'
    elif '三盘' in name:
        form = '三盘'
    elif '四盘' in name:
        form = '四盘'

    # System USB
    if '16G' in name.upper() or '16g' in name:
        if 'windows' in name.lower():
            return 'USB:16GB:windows'
        elif '飞牛' in name or '蜗牛' in name:
            return 'USB:16GB:飞牛'
        return f'USB:16GB:{name[:20]}'

    return f"{cpu}:{mem}:{form}"


def clean_sku_name(name):
    """Clean up OCR artifacts in SKU name"""
    name = name.strip()
    # Remove known noise
    name = re.sub(r'关注券5元优惠券', '', name)
    name = re.sub(r'店铺关注券5元优惠券', '', name)
    name = re.sub(r'[>\s]+$', '', name)
    name = re.sub(r'^[>\s]+', '', name)
    name = re.sub(r'\s+', ' ', name)
    return name


# Collect all unique SKUs
all_entries = []
for img_name, lines in all_results.items():
    entries = extract_sku_v4(lines)
    all_entries.extend(entries)

# Deduplicate - keep highest confidence per normalized key
dedup = {}
for e in all_entries:
    key = normalize_sku_key(e['sku'])
    if key and key not in dedup:
        dedup[key] = e
    elif key and e['confidence'] > dedup[key]['confidence']:
        dedup[key] = e

# Sort by price
sorted_entries = sorted(dedup.values(), key=lambda x: x['price'])

# Print for debugging
print("=" * 60)
print("Extracted unique SKUs:")
for i, e in enumerate(sorted_entries):
    key = normalize_sku_key(e['sku'])
    print(f"  [{key}] {clean_sku_name(e['sku'])} -> ${e['price']} (orig: ${e['original_price']})")

# Build markdown
md = []
md.append("# SKU 售价汇总")
md.append("")
md.append("> 从拼多多商家后台「商品价格管理」截图 OCR 识别提取")
md.append(f"> 提取日期: 2026-06-16")
md.append(f"> 图片数量: {len(all_results)} 张")
md.append(f"> 识别 SKU: {len(sorted_entries)} 个（已去重合并）")
md.append("")
md.append("---")
md.append("")

# Group by platform
groups = {
    '系统U盘 (16GB)': [],
    '单盘位 - 1037U': [],
    '双盘位 - 1037U': [],
    '单盘位 - J1900': [],
    '双盘位 - J1900': [],
    '单盘位 - J3160': [],
    '双盘位 - J3160': [],
    '单盘位 - i3': [],
    '双盘位 - i3': [],
    '双盘位 - i5': [],
    '三盘位/四盘位 - i3/i5': [],
    '5005U 系列': [],
    '其他NAS/黑裙晖': [],
}

for e in sorted_entries:
    name = e['sku']
    key = normalize_sku_key(name)

    if 'USB' in key:
        groups['系统U盘 (16GB)'].append(e)
    elif '1037U' in key:
        if '单盘' in key:
            groups['单盘位 - 1037U'].append(e)
        else:
            groups['双盘位 - 1037U'].append(e)
    elif 'J1900' in key:
        if '单盘' in key:
            groups['单盘位 - J1900'].append(e)
        else:
            groups['双盘位 - J1900'].append(e)
    elif 'J3160' in key:
        if '单盘' in key:
            groups['单盘位 - J3160'].append(e)
        else:
            groups['双盘位 - J3160'].append(e)
    elif 'I3' in key:
        if '三盘' in key or '四盘' in key:
            groups['三盘位/四盘位 - i3/i5'].append(e)
        elif '单盘' in key:
            groups['单盘位 - i3'].append(e)
        else:
            groups['双盘位 - i3'].append(e)
    elif 'I5' in key:
        if '三盘' in key or '四盘' in key:
            groups['三盘位/四盘位 - i3/i5'].append(e)
        else:
            groups['双盘位 - i5'].append(e)
    elif '5005U' in key:
        groups['5005U 系列'].append(e)
    else:
        groups['其他NAS/黑裙晖'].append(e)

for group_name in groups:
    entries = groups[group_name]
    if not entries:
        continue

    md.append(f"## {group_name}")
    md.append("")
    md.append("| SKU 配置 | 券后售价 (¥) | 券前原价 (¥) | 优惠 |")
    md.append("|---------|------------:|------------:|-----|")

    for e in entries:
        name = clean_sku_name(e['sku'])
        md.append(f"| {name} | {e['price']:.2f} | {e['original_price']:.2f} | -5.00 |")

    md.append("")

md.append("---")
md.append("")
md.append("## 说明")
md.append("")
md.append("- **优惠**: 统一使用店铺关注券 -5.00 元")
md.append("- **数据来源**: 拼多多商家后台「商品价格管理」页面截图")
md.append("- **识别工具**: RapidOCR (rapidocr-onnxruntime)")
md.append("- **注意事项**: OCR 识别的 SKU 名称可能有碎片化/错字，建议对比后台实际数据确认")
md.append("- **价格日期**: 2026-06-05 (截图时间)")
md.append("")

# Fix the price summary line
overall_min = min(e['price'] for e in sorted_entries) if sorted_entries else 0
overall_max = max(e['price'] for e in sorted_entries) if sorted_entries else 0
md.append(f"> 价格范围: ¥{overall_min:.2f} ~ ¥{overall_max:.2f}")

output_path = r"D:\workspace\membank\售价\SKU售价汇总.md"
with open(output_path, "w", encoding="utf-8") as f:
    f.write("\n".join(md))

print(f"\n\nSaved to: {output_path}")
print(f"Total unique SKUs: {len(sorted_entries)}")