487 lines
20 KiB
Python
487 lines
20 KiB
Python
import os
|
||
import re
|
||
import unicodedata
|
||
import json
|
||
import threading
|
||
import pandas as pd
|
||
import requests
|
||
import time
|
||
import random
|
||
from io import StringIO
|
||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||
from app.models import db, Job
|
||
|
||
print("🆕 MODERN webcrawler LOADED! – BATCHED + PROXY + RESUME + ETA + 4x SCRAPER CHUNK-PARALLEL")
|
||
|
||
UPLOAD_FOLDER = '/app/uploads'
|
||
RESULT_FOLDER = '/app/results'
|
||
|
||
SCRAPER_URLS = [
|
||
"http://gmaps-scraper-1:8080",
|
||
"http://gmaps-scraper-2:8080",
|
||
"http://gmaps-scraper-3:8080",
|
||
"http://gmaps-scraper-4:8080",
|
||
]
|
||
|
||
OUTPUT_COLS = ['title', 'category', 'address', 'website', 'phone', 'link']
|
||
|
||
PROXY_URL = "http://bitlleuv-rotate:s5hzse6hz74b@p.webshare.io:80"
|
||
API_PROXIES = {"http": PROXY_URL, "https": PROXY_URL}
|
||
|
||
_job_semaphore = threading.Semaphore(1)
|
||
|
||
# ──────────────────────────────────────────────
|
||
# Tuning
|
||
# ──────────────────────────────────────────────
|
||
BATCH_SIZE = 30 # Keywords pro Scraper-Job
|
||
BATCH_DELAY_MIN = 3 # Sekunden Pause zwischen Chunks (min)
|
||
BATCH_DELAY_MAX = 6 # Sekunden Pause zwischen Chunks (max)
|
||
MAX_TIME = 60 # Sekunden die der Scraper pro Batch hat
|
||
POLL_MAX = 90 # Max. Poll-Versuche pro Batch
|
||
POLL_DELAY_MIN = 2 # Sekunden zwischen Polls (min)
|
||
POLL_DELAY_MAX = 5 # Sekunden zwischen Polls (max)
|
||
STUCK_TIMEOUT = 300 # Sekunden bis Scraper-Neustart (5 Min)
|
||
MAX_RETRIES = 2 # Wiederholversuche pro Batch bei Fehler
|
||
PARALLEL_WORKERS = len(SCRAPER_URLS)
|
||
|
||
_partial_lock = threading.Lock()
|
||
|
||
# ──────────────────────────────────────────────
|
||
# Hilfsfunktionen
|
||
# ──────────────────────────────────────────────
|
||
|
||
def is_blocked(data):
|
||
text = str(data).lower()
|
||
blocked = any(kw in text for kw in ['captcha', 'blocked', 'rate limit', 'too many', '429'])
|
||
if blocked:
|
||
print(f"🚫 BLOCKED: {str(data)[:100]}")
|
||
return blocked
|
||
|
||
def fix_encoding(text):
|
||
if not isinstance(text, str):
|
||
return text
|
||
try:
|
||
return text.encode('latin-1').decode('utf-8')
|
||
except (UnicodeEncodeError, UnicodeDecodeError):
|
||
return text
|
||
|
||
def clean_query(q):
|
||
q = ''.join(c for c in q if unicodedata.category(c) != 'Cc')
|
||
q = ' '.join(q.split())
|
||
return q.strip()
|
||
|
||
def build_input_addresses(df):
|
||
addresses = set()
|
||
for _, row in df.iterrows():
|
||
plz = str(row.get('PLZ', '')).strip()
|
||
stadt = str(row.get('Stadt', '')).strip()
|
||
str_ = str(row.get('Straße', '')).strip()
|
||
nr = str(row.get('Hausnummer', '')).strip()
|
||
zusatz = str(row.get('Zusatz', '')).strip()
|
||
full = f"{str_} {nr} {zusatz} {plz} {stadt}".lower().strip()
|
||
full = ' '.join(full.split())
|
||
addresses.add(full)
|
||
return addresses
|
||
|
||
def normalize_address(addr):
|
||
if not isinstance(addr, str):
|
||
return ''
|
||
addr = fix_encoding(addr)
|
||
return ' '.join(addr.lower().strip().split())
|
||
|
||
def address_in_input(result_addr, input_addresses):
|
||
norm = normalize_address(result_addr)
|
||
for inp_addr in input_addresses:
|
||
plz_match = re.search(r'\b\d{5}\b', inp_addr)
|
||
if not plz_match:
|
||
continue
|
||
plz = plz_match.group()
|
||
if plz not in norm:
|
||
continue
|
||
|
||
parts = inp_addr.split()
|
||
|
||
street = parts[0] if parts else ''
|
||
if len(street) < 4 or street[:5].lower() not in norm:
|
||
continue
|
||
|
||
hausnr = parts[1] if len(parts) > 1 else ''
|
||
if hausnr and not re.search(rf'\b{re.escape(hausnr)}\b', norm):
|
||
continue
|
||
|
||
return True
|
||
return False
|
||
|
||
def format_eta(seconds):
|
||
if seconds < 60:
|
||
return f"{int(seconds)}s"
|
||
h, rem = divmod(int(seconds), 3600)
|
||
m = rem // 60
|
||
return f"{h}h {m:02d}min" if h > 0 else f"{m}min"
|
||
|
||
# ──────────────────────────────────────────────
|
||
# Scraper-Job Cleanup
|
||
# ──────────────────────────────────────────────
|
||
|
||
def _cleanup_scraper_job(scraper_url, scraper_id):
|
||
"""Scraper-Job immer aufräumen wenn wir ihn nicht mehr brauchen"""
|
||
try:
|
||
requests.delete(f"{scraper_url}/api/v1/jobs/{scraper_id}", timeout=10)
|
||
print(f"🗑️ Scraper-Job {scraper_id} gelöscht")
|
||
except Exception as e:
|
||
print(f"⚠️ Cleanup fehlgeschlagen: {e}")
|
||
|
||
# ──────────────────────────────────────────────
|
||
# Scraper-Neustart via Docker SDK
|
||
# ──────────────────────────────────────────────
|
||
|
||
def restart_scraper(scraper_url):
|
||
try:
|
||
import docker
|
||
container_name = scraper_url.split("//")[1].split(":")[0]
|
||
print(f"🔄 Starte {container_name} neu...")
|
||
client = docker.from_env()
|
||
container = client.containers.get(container_name)
|
||
container.restart()
|
||
print(f"✅ {container_name} neu gestartet – warte 15s...")
|
||
time.sleep(15)
|
||
return True
|
||
except Exception as e:
|
||
print(f"⚠️ Scraper-Neustart fehlgeschlagen: {e}")
|
||
return False
|
||
|
||
# ──────────────────────────────────────────────
|
||
# Resume: Progress-File Hilfsfunktionen
|
||
# ──────────────────────────────────────────────
|
||
|
||
def get_progress_path(job_id):
|
||
return os.path.join(RESULT_FOLDER, f"progress_{job_id}.json")
|
||
|
||
def get_partial_path(job_id, suffix):
|
||
return os.path.join(RESULT_FOLDER, f"partial_{job_id}_{suffix}.csv")
|
||
|
||
def load_progress(job_id):
|
||
path = get_progress_path(job_id)
|
||
if os.path.exists(path):
|
||
with open(path, 'r') as f:
|
||
data = json.load(f)
|
||
print(f"🔁 RESUME: ab Batch {data['last_completed_batch'] + 1}/{data['total_batches']}")
|
||
return data
|
||
return None
|
||
|
||
def save_progress(job_id, last_completed_batch, total_batches):
|
||
path = get_progress_path(job_id)
|
||
with open(path, 'w') as f:
|
||
json.dump({"last_completed_batch": last_completed_batch, "total_batches": total_batches}, f)
|
||
|
||
def append_partial(job_id, df_filtered, df_raw):
|
||
with _partial_lock:
|
||
for suffix, df in [('filtered', df_filtered), ('raw', df_raw)]:
|
||
if df is None:
|
||
continue
|
||
path = get_partial_path(job_id, suffix)
|
||
header = not os.path.exists(path)
|
||
df.to_csv(path, mode='a', index=False, header=header, encoding='utf-8-sig', sep=';')
|
||
|
||
def load_partial(job_id):
|
||
results_filtered, results_raw = [], []
|
||
for suffix, lst in [('filtered', results_filtered), ('raw', results_raw)]:
|
||
path = get_partial_path(job_id, suffix)
|
||
if os.path.exists(path):
|
||
try:
|
||
df = pd.read_csv(path, sep=';', encoding='utf-8-sig')
|
||
lst.append(df)
|
||
print(f"📂 Partial {suffix}: {len(df)} Zeilen geladen")
|
||
except Exception as e:
|
||
print(f"⚠️ Partial {suffix} Ladefehler: {e}")
|
||
return results_filtered, results_raw
|
||
|
||
def cleanup_progress(job_id):
|
||
for path in [
|
||
get_progress_path(job_id),
|
||
get_partial_path(job_id, 'filtered'),
|
||
get_partial_path(job_id, 'raw'),
|
||
]:
|
||
if os.path.exists(path):
|
||
os.remove(path)
|
||
|
||
# ──────────────────────────────────────────────
|
||
# CSV Nachbearbeitung
|
||
# ──────────────────────────────────────────────
|
||
|
||
def process_result_csv(raw_bytes, input_df, apply_filter=True):
|
||
try:
|
||
content = raw_bytes.decode('utf-8', errors='replace')
|
||
df_out = pd.read_csv(StringIO(content))
|
||
print(f"📄 Raw result: {df_out.shape}")
|
||
|
||
available = [c for c in OUTPUT_COLS if c in df_out.columns]
|
||
df_out = df_out[available]
|
||
|
||
for col in df_out.columns:
|
||
df_out[col] = df_out[col].apply(fix_encoding)
|
||
|
||
if apply_filter:
|
||
input_addresses = build_input_addresses(input_df)
|
||
before = len(df_out)
|
||
df_out = df_out[
|
||
df_out['address'].apply(lambda a: address_in_input(a, input_addresses))
|
||
]
|
||
print(f"📍 Filter: {before} → {len(df_out)}")
|
||
|
||
df_out = df_out.drop_duplicates(subset=['title', 'address'], keep='first')
|
||
df_out = df_out.dropna(subset=['title'], how='all')
|
||
df_out = df_out[df_out['title'].str.strip().astype(bool)]
|
||
|
||
print(f"✅ Final ({'gefiltert' if apply_filter else 'alle'}): {df_out.shape}")
|
||
return df_out
|
||
except Exception as e:
|
||
print(f"💥 process_result_csv: {e}")
|
||
return None
|
||
|
||
# ──────────────────────────────────────────────
|
||
# Parallel: Einzelnen Batch verarbeiten
|
||
# ──────────────────────────────────────────────
|
||
|
||
def process_batch(batch_idx, batch_queries, scraper_url, filename, job_id, df_input):
|
||
payload = {
|
||
"name": f"{filename.replace('.csv','')}-{job_id}-B{batch_idx+1:03d}",
|
||
"keywords": batch_queries,
|
||
"lang": "de",
|
||
"depth": 1,
|
||
"zoom": 17,
|
||
"radius": 100,
|
||
"max_time": MAX_TIME,
|
||
"fast_mode": False,
|
||
"proxies": [PROXY_URL]
|
||
}
|
||
|
||
for attempt in range(1, MAX_RETRIES + 1):
|
||
scraper_id = None
|
||
try:
|
||
resp = requests.post(f"{scraper_url}/api/v1/jobs", json=payload, timeout=45)
|
||
print(f"📤 Batch {batch_idx+1} → {scraper_url} | {resp.status_code} (Versuch {attempt})")
|
||
|
||
if is_blocked(resp.text):
|
||
print(f"🚫 Batch {batch_idx+1} blocked")
|
||
return None, None
|
||
|
||
if resp.status_code != 201:
|
||
print(f"⚠️ Batch {batch_idx+1} fehlgeschlagen: {resp.text[:100]}")
|
||
if attempt < MAX_RETRIES:
|
||
time.sleep(10)
|
||
continue
|
||
|
||
scraper_id = resp.json()['id']
|
||
print(f"✅ Batch {batch_idx+1} Scraper-ID: {scraper_id}")
|
||
|
||
batch_start_time = time.time()
|
||
|
||
for poll_i in range(1, POLL_MAX + 1):
|
||
r = requests.get(f"{scraper_url}/api/v1/jobs/{scraper_id}", timeout=15)
|
||
data = r.json()
|
||
status = data.get('Status', data.get('status', '?'))
|
||
elapsed = time.time() - batch_start_time
|
||
print(f"⏳ Batch {batch_idx+1} Poll {poll_i}: {status} | {int(elapsed)}s")
|
||
|
||
if status == 'pending' and elapsed > STUCK_TIMEOUT:
|
||
print(f"⚠️ Batch {batch_idx+1} hängt seit {int(elapsed)}s – Neustart {scraper_url}")
|
||
_cleanup_scraper_job(scraper_url, scraper_id)
|
||
scraper_id = None
|
||
restart_scraper(scraper_url)
|
||
break
|
||
|
||
if status in ('ok', 'completed', 'scraped'):
|
||
dl = requests.get(f"{scraper_url}/api/v1/jobs/{scraper_id}/download", timeout=90)
|
||
scraper_id = None
|
||
if dl.status_code == 200:
|
||
df_filtered = process_result_csv(dl.content, df_input, True)
|
||
df_raw = process_result_csv(dl.content, df_input, False)
|
||
print(f"📊 Batch {batch_idx+1}: {len(df_filtered) if df_filtered is not None else 0} filtered")
|
||
return df_filtered, df_raw
|
||
return None, None
|
||
|
||
elif status in ('failed', 'error'):
|
||
print(f"💥 Batch {batch_idx+1}: {status} (Versuch {attempt})")
|
||
_cleanup_scraper_job(scraper_url, scraper_id)
|
||
scraper_id = None
|
||
if attempt < MAX_RETRIES:
|
||
time.sleep(10)
|
||
break
|
||
|
||
time.sleep(random.uniform(POLL_DELAY_MIN, POLL_DELAY_MAX))
|
||
|
||
except Exception as e:
|
||
print(f"💥 Batch {batch_idx+1} Versuch {attempt}: {e}")
|
||
if scraper_id:
|
||
_cleanup_scraper_job(scraper_url, scraper_id)
|
||
scraper_id = None
|
||
if attempt < MAX_RETRIES:
|
||
time.sleep(10)
|
||
|
||
return None, None
|
||
|
||
# ──────────────────────────────────────────────
|
||
# HAUPT-WORKER
|
||
# ──────────────────────────────────────────────
|
||
|
||
def process_file(filename, job_id, app):
|
||
with app.app_context():
|
||
job = Job.query.get(job_id)
|
||
if job:
|
||
job.status = "⏳ Wartet auf anderen Job..."
|
||
db.session.commit()
|
||
|
||
with _job_semaphore:
|
||
print(f"🎯 {filename} Job#{job_id} START!")
|
||
|
||
with app.app_context():
|
||
job = Job.query.get(job_id)
|
||
if not job:
|
||
print("❌ Job missing")
|
||
return
|
||
|
||
try:
|
||
job.status = "📊 parsing CSV"
|
||
db.session.commit()
|
||
|
||
filepath = os.path.join(UPLOAD_FOLDER, filename)
|
||
print(f"📁 {filepath} | {os.path.getsize(filepath)}b")
|
||
|
||
df_input = pd.read_csv(filepath, sep=';', encoding='ISO-8859-1')
|
||
print(f"📊 {df_input.shape}")
|
||
|
||
queries = []
|
||
for _, row in df_input.iterrows():
|
||
parts = [
|
||
str(row.get('PLZ', '')).strip(),
|
||
str(row.get('Stadt', '')).strip(),
|
||
str(row.get('Straße', '')).strip(),
|
||
str(row.get('Hausnummer', '')).strip(),
|
||
str(row.get('Zusatz', '')).strip(),
|
||
]
|
||
q = f"Firmen {' '.join(p for p in parts if p and p != 'nan')}".strip()
|
||
q = clean_query(q)
|
||
if len(q) > 10:
|
||
queries.append(q)
|
||
|
||
total_queries = len(queries)
|
||
print(f"🔍 {total_queries} Queries | Samples: {queries[:3]}")
|
||
if total_queries == 0:
|
||
raise ValueError("Keine gültigen Adressen")
|
||
|
||
batches = (total_queries + BATCH_SIZE - 1) // BATCH_SIZE
|
||
|
||
os.makedirs(RESULT_FOLDER, exist_ok=True)
|
||
progress = load_progress(job_id)
|
||
start_batch = progress['last_completed_batch'] + 1 if progress else 0
|
||
all_results_filtered, all_results_raw = load_partial(job_id) if progress else ([], [])
|
||
|
||
eta_initial = format_eta((batches - start_batch) * ((BATCH_DELAY_MAX + MAX_TIME) / 2) / PARALLEL_WORKERS)
|
||
print(f"📦 {batches} Batches à {BATCH_SIZE} | {PARALLEL_WORKERS}x parallel (Chunk) | Start: {start_batch} | ETA: ~{eta_initial}")
|
||
job_start_time = time.time()
|
||
job.status = f"🔄 Batch {start_batch+1}/{batches} | ⏱️ ~{eta_initial}"
|
||
db.session.commit()
|
||
|
||
completed_count = 0
|
||
|
||
batch_indices = list(range(start_batch, batches))
|
||
chunks = [
|
||
batch_indices[i:i + PARALLEL_WORKERS]
|
||
for i in range(0, len(batch_indices), PARALLEL_WORKERS)
|
||
]
|
||
|
||
with ThreadPoolExecutor(max_workers=PARALLEL_WORKERS) as executor:
|
||
for chunk_idx, chunk in enumerate(chunks):
|
||
futures = {}
|
||
|
||
for batch_idx in chunk:
|
||
batch_start_q = batch_idx * BATCH_SIZE
|
||
batch_end_q = min(batch_start_q + BATCH_SIZE, total_queries)
|
||
batch_queries = queries[batch_start_q:batch_end_q]
|
||
scraper_url = SCRAPER_URLS[batch_idx % len(SCRAPER_URLS)]
|
||
|
||
print(f"\n🚀 Chunk {chunk_idx+1} | Batch {batch_idx+1}/{batches} → {scraper_url}")
|
||
time.sleep(random.uniform(1, 2))
|
||
|
||
future = executor.submit(
|
||
process_batch,
|
||
batch_idx, batch_queries, scraper_url,
|
||
filename, job_id, df_input
|
||
)
|
||
futures[future] = batch_idx
|
||
|
||
for future in as_completed(futures):
|
||
batch_idx = futures[future]
|
||
completed_count += 1
|
||
try:
|
||
df_filtered, df_raw = future.result()
|
||
if df_filtered is not None:
|
||
all_results_filtered.append(df_filtered)
|
||
all_results_raw.append(df_raw)
|
||
append_partial(job_id, df_filtered, df_raw)
|
||
except Exception as e:
|
||
print(f"💥 Batch {batch_idx+1} Exception: {e}")
|
||
|
||
save_progress(job_id, batch_idx, batches)
|
||
|
||
elapsed = time.time() - job_start_time
|
||
if completed_count > 0:
|
||
avg_per_batch = elapsed / completed_count
|
||
remaining = (batches - start_batch - completed_count) * avg_per_batch / PARALLEL_WORKERS
|
||
eta_str = format_eta(remaining)
|
||
else:
|
||
eta_str = "?"
|
||
|
||
job.status = f"🔄 {completed_count}/{batches - start_batch} fertig | ⏱️ ~{eta_str}"
|
||
db.session.commit()
|
||
|
||
if chunk_idx < len(chunks) - 1:
|
||
delay = random.uniform(BATCH_DELAY_MIN, BATCH_DELAY_MAX)
|
||
print(f"⏸️ Chunk {chunk_idx+1} fertig – warte {delay:.1f}s...")
|
||
time.sleep(delay)
|
||
|
||
# ── MERGE & SAVE ──
|
||
job.status = "🔧 merging results"
|
||
db.session.commit()
|
||
|
||
base = filename.replace('.csv', '')
|
||
|
||
if all_results_filtered:
|
||
df_final_filtered = pd.concat(all_results_filtered, ignore_index=True)
|
||
df_final_filtered = df_final_filtered.drop_duplicates(subset=['title', 'address'])
|
||
|
||
out_filtered = f"results_{base}_filtered.csv"
|
||
df_final_filtered.to_csv(
|
||
os.path.join(RESULT_FOLDER, out_filtered),
|
||
index=False, encoding='utf-8-sig', sep=';'
|
||
)
|
||
|
||
out_raw = None
|
||
if all_results_raw:
|
||
df_final_raw = pd.concat(all_results_raw, ignore_index=True)
|
||
out_raw = f"results_{base}_all.csv"
|
||
df_final_raw.to_csv(
|
||
os.path.join(RESULT_FOLDER, out_raw),
|
||
index=False, encoding='utf-8-sig', sep=';'
|
||
)
|
||
|
||
job.result_filename = out_filtered
|
||
job.result_filename_raw = out_raw
|
||
job.status = f"✅ Fertig: {len(df_final_filtered)} Firmen"
|
||
|
||
cleanup_progress(job_id)
|
||
else:
|
||
job.status = "❌ Keine Ergebnisse"
|
||
|
||
db.session.commit()
|
||
print(f"🎉 Job {job_id} komplett!")
|
||
|
||
except Exception as e:
|
||
job.status = f"Failed: {str(e)[:50]}"
|
||
print(f"💥 FATAL: {e}")
|
||
import traceback
|
||
traceback.print_exc()
|
||
db.session.commit()
|
||
|
||
print(f"✅ DONE! Status: {job.status}")
|