test-mcp/crawler_service_test.py

import os
import re
import asyncio
import json
import traceback
import sys
import threading
import queue
import pymysql
from datetime import datetime, timedelta
from playwright.async_api import async_playwright
from dotenv import load_dotenv
from sql_queries import CrawlerQueries

load_dotenv(override=True)

# 글로벌 중단 제어용 이벤트
crawl_stop_event = threading.Event()

def get_db_connection():
    """MySQL 데이터베이스(TEST) 연결을 반환"""
    return pymysql.connect(
        host=os.getenv('DB_HOST', 'localhost'),
        port=int(os.getenv('DB_PORT', 3306)),
        user=os.getenv('DB_USER', 'root'),
        password=os.getenv('DB_PASSWORD', '45278434'),
        database='PM_proto_test',  # 테스트용 DB 고정
        charset='utf8mb4',
        cursorclass=pymysql.cursors.DictCursor
    )

def clean_date_string(date_str):
    """원본 crawler_service.py와 동일한 날짜 정리 로직"""
    if not date_str: return ""
    match = re.search(r'(\d{2})[./-](\d{2})[./-](\d{2})', date_str)
    if match: return f"20{match.group(1)}.{match.group(2)}.{match.group(3)}"
    return date_str[:10].replace("-", ".")

def parse_log_id(log_id):
    """원본 crawler_service.py와 동일한 로그 ID 파싱 로직"""
    if not log_id or "_" not in log_id: return log_id
    try:
        parts = log_id.split('_')
        if len(parts) >= 4:
            date_part = clean_date_string(parts[1])
            activity = parts[3].strip()
            activity = re.sub(r'\(.*?\)', '', activity).strip()
            return f"{date_part}, {activity}"
    except: pass
    return log_id

def crawler_thread_worker(msg_queue, user_id, password):
    crawl_stop_event.clear()
    if sys.platform == 'win32':
        asyncio.set_event_loop_policy(asyncio.WindowsProactorEventLoopPolicy())
    
    loop = asyncio.new_event_loop()
    asyncio.set_event_loop(loop)
    
    async def run():
        async with async_playwright() as p:
            browser = None
            try:
                msg_queue.put(json.dumps({'type': 'log', 'message': '[TEST] 원본 수집 방식 복구 및 추론 엔진 가동...'}))
                browser = await p.chromium.launch(headless=True, args=[
                    "--no-sandbox", 
                    "--disable-dev-shm-usage",
                    "--disable-blink-features=AutomationControlled"
                ])
                context = await browser.new_context(
                    viewport={'width': 1600, 'height': 900},
                    user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36"
                )
                
                captured_data = {"tree": None, "_is_root_archive": False, "project_list": [], "last_project_data": None}

                async def global_interceptor(response):
                    url = response.url
                    try:
                        if "getAllList" in url:
                            data = await response.json()
                            captured_data["project_list"] = data.get("data", [])
                        elif "getTreeObject" in url:
                            # [복구] 원본과 100% 동일한 루트 판정 로직
                            is_root = False
                            if "params[resourcePath]=" in url:
                                path_val = url.split("params[resourcePath]=")[1].split("&")[0]
                                if path_val in ["%2F", "/"]: is_root = True
                            if is_root:
                                captured_data["tree"] = await response.json()
                                captured_data["_is_root_archive"] = True
                        elif "getData" in url and "overview" in url:
                            captured_data["last_project_data"] = await response.json()
                    except: pass

                context.on("response", global_interceptor)
                page = await context.new_page()
                await page.goto("https://overseas.projectmastercloud.com/dashboard", wait_until="domcontentloaded")
                
                if await page.locator("#login-by-id").is_visible(timeout=10000):
                    await page.click("#login-by-id"); await page.fill("#user_id", user_id); await page.fill("#user_pw", password); await page.click("#login-btn")

                await page.wait_for_selector("h4.list__contents_aria_group_body_list_item_label", timeout=60000)
                await asyncio.sleep(2)

                project_names = list(dict.fromkeys([n.strip() for n in await page.locator("h4.list__contents_aria_group_body_list_item_label").all_inner_texts() if n.strip()]))
                count = len(project_names)

                for i, project_name in enumerate(project_names):
                    if crawl_stop_event.is_set(): break
                    msg_queue.put(json.dumps({'type': 'log', 'message': f'[TEST] [{i+1}/{count}] {project_name} 수집'}))
                    p_match = next((p for p in captured_data["project_list"] if p.get('project_nm') == project_name or p.get('short_nm', '').strip() == project_name), None)
                    current_p_id = p_match.get('project_id') if p_match else None
                    
                    try:
                        # 1. 프로젝트 진입
                        target_el = page.locator(f"h4.list__contents_aria_group_body_list_item_label:has-text('{project_name}')").first
                        await target_el.scroll_into_view_if_needed()
                        box = await target_el.bounding_box()
                        if box: await page.mouse.click(box['x'] + 5, box['y'] + 5)
                        else: await target_el.click(force=True)
                        await page.wait_for_selector("text=활동로그", timeout=30000)
                        
                        # 2. [복구] 최신 파일 수 실측 (원본의 수동 Fetch 방식 그대로)
                        captured_data["tree"] = None; captured_data["_is_root_archive"] = False
                        await page.evaluate("""() => { 
                            const baseUrl = window.location.origin + window.location.pathname.split('/').slice(0, 2).join('/');
                            fetch(`${baseUrl}/archive/getTreeObject?params[storageType]=CLOUD&params[resourcePath]=/`); 
                        }""")
                        for _ in range(30):
                            if captured_data["_is_root_archive"]: break
                            await asyncio.sleep(0.5)

                        actual_count = 0
                        if captured_data["tree"]:
                            tree_data = captured_data["tree"]
                            if isinstance(tree_data, list) and len(tree_data) > 0: tree_data = tree_data[0]
                            if isinstance(tree_data, dict):
                                tree = tree_data.get('currentTreeObject', tree_data)
                                if isinstance(tree, dict):
                                    # 원본 파일 수 합산 로직
                                    total = len(tree.get("file", {}))
                                    folders = tree.get("folder", {})
                                    if isinstance(folders, dict):
                                        for f in folders.values(): total += int(f.get("filesCount", 0))
                                    actual_count = total

                        # 3. 활동로그 전수 수집 (하이브리드 방식: 최상단 우선 확보 + 전수 스크롤)
                        all_logs = []
                        await page.get_by_text("활동로그").first.click()
                        if await page.wait_for_selector("article.archive-modal", timeout=10000):
                            # 날짜 필터 적용 (2020-01-01)
                            inputs = await page.locator("article.archive-modal input").all()
                            for inp in inputs:
                                if (await inp.get_attribute("type")) == "date": await inp.fill("2020-01-01"); break
                            
                            apply_btn = page.locator("article.archive-modal").get_by_text("적용").first
                            if await apply_btn.is_visible():
                                await apply_btn.click()
                                # [핵심] 첫 번째 로그가 나타날 때까지 명시적 대기 (최대 10초)
                                try:
                                    await page.wait_for_selector("article.archive-modal div[id*='_']", timeout=10000)
                                except: pass
                                await asyncio.sleep(2)

                            # (1) 최상단 로그 즉시 확보 (안전장치)
                            first_log_el = await page.locator("article.archive-modal div[id*='_']").first.get_attribute("id")
                            if first_log_el:
                                first_log_text = parse_log_id(first_log_el)
                                if ", " in first_log_text:
                                    d, a = first_log_text.split(", ", 1)
                                    all_logs.append({'date': d, 'activity': a})

                            # (2) 전수 수집을 위한 무한 스크롤 및 지정된 클래스 내 ID 수집
                            last_count = len(all_logs)
                            for _ in range(20):
                                # 스크롤 수행 (사용자가 지정한 log-body 클래스 기준)
                                await page.evaluate("""() => {
                                    const body = document.querySelector('.log-item-wrap.log-body.scrollbar.scroll-container') || 
                                                 document.querySelector('article.archive-modal .modal-body') || 
                                                 document.querySelector('article.archive-modal');
                                    if (body) body.scrollTop = body.scrollHeight;
                                }""")
                                await asyncio.sleep(1.5)
                                
                                # 사용자 지정 클래스 내의 모든 div ID 수집
                                # .log-item-wrap.log-body.scrollbar.scroll-container 내부의 div들을 타겟팅
                                selector = ".log-item-wrap.log-body.scrollbar.scroll-container div"
                                current_elements = await page.locator(selector).all()
                                
                                # 만약 지정된 클래스로 검색되지 않을 경우 기존 div[id*='_']를 백업으로 사용
                                if not current_elements:
                                    current_elements = await page.locator("article.archive-modal div[id*='_']").all()

                                seen_ids = {f"{log['date']}, {log['activity']}" for log in all_logs}
                                for el in current_elements:
                                    log_id = await el.get_attribute("id")
                                    if not log_id: continue
                                    
                                    log_text = parse_log_id(log_id)
                                    if ", " in log_text and log_text not in seen_ids:
                                        d, a = log_text.split(", ", 1)
                                        all_logs.append({'date': d, 'activity': a})
                                        seen_ids.add(log_text)
                                
                                if len(all_logs) == last_count: break
                                last_count = len(all_logs)

                            if not all_logs:
                                msg_queue.put(json.dumps({'type': 'log', 'message': f'   - [주의] {project_name}: 수집된 로그가 없습니다.'}))
                            
                            await page.keyboard.press("Escape")

                        # 4. 파일 수 추론 (전수 보존 모드)
                        history_map = {}
                        curr_calc_count = actual_count
                        
                        if all_logs:
                            # 오늘 날짜 강제 주입 대신 수집된 로그의 실제 날짜 사용
                            for log in all_logs:
                                d_db = log['date'].replace(".", "-")
                                act = log['activity']
                                if d_db not in history_map:
                                    history_map[d_db] = {"log": act, "count": curr_calc_count}
                                
                                if "업로드" in act: curr_calc_count -= 1
                                elif "삭제" in act: curr_calc_count += 1
                                if curr_calc_count < 0: curr_calc_count = 0
                                history_map[d_db]["count"] = curr_calc_count
                        else:
                            # 로그가 전혀 없을 경우에만 기본값 생성
                            today_str = datetime.now().strftime("%Y-%m-%d")
                            history_map[today_str] = {"log": "기존 상태 유지 (활동 없음)", "count": actual_count}

                        # 5. DB 저장
                        if current_p_id:
                            with get_db_connection() as conn:
                                with conn.cursor() as cursor:
                                    for date_key, data in history_map.items():
                                        cursor.execute(CrawlerQueries.UPSERT_HISTORY_WITH_DATE, 
                                                       (current_p_id, date_key, f"{date_key.replace('-', '.')}, {data['log']}", data['count']))
                                conn.commit()
                            msg_queue.put(json.dumps({'type': 'log', 'message': f'   - [성공] 실측 {actual_count}개 기준 시계열 적재 완료'}))

                        await page.goto("https://overseas.projectmastercloud.com/dashboard", wait_until="domcontentloaded")
                        
                    except Exception as e:
                        msg_queue.put(json.dumps({'type': 'log', 'message': f'   - {project_name} 에러: {str(e)}'}))
                        await page.goto("https://overseas.projectmastercloud.com/dashboard")

                msg_queue.put(json.dumps({'type': 'done', 'data': []}))
                
            except Exception as e:
                msg_queue.put(json.dumps({'type': 'log', 'message': f'치명적 오류: {str(e)}'}))
            finally:
                if browser: await browser.close()
                msg_queue.put(None)

    loop.run_until_complete(run())
    loop.close()

async def run_crawler_service():
    msg_queue = queue.Queue()
    thread = threading.Thread(target=crawler_thread_worker, args=(msg_queue, os.getenv("PM_USER_ID"), os.getenv("PM_PASSWORD")))
    thread.start()
    while True:
        try:
            msg = await asyncio.to_thread(msg_queue.get, timeout=1.0)
            if msg is None: break
            yield f"data: {msg}\n\n"
        except queue.Empty:
            if not thread.is_alive(): break
            await asyncio.sleep(0.1)
    thread.join()