np.select 멀티프로세싱 적용

import numpy as np
import pandas as pd
import multiprocessing as mp
from functools import partial

def process_chunk(df_chunk, conditions, choices, default):
    # 각 청크에 np.select 적용
    result = np.select(conditions, choices, default=default)
    return pd.Series(result, index=df_chunk.index)

def parallel_select(df, conditions, choices, default='default', num_processes=None):
    if num_processes is None:
        num_processes = mp.cpu_count()
    
    # 데이터 분할
    chunks = np.array_split(df, num_processes)
    
    # 부분 함수 생성 (conditions, choices, default 고정)
    worker = partial(process_chunk, conditions=conditions, choices=choices, default=default)
    
    with mp.Pool(num_processes) as pool:
        results = pool.map(worker, chunks)
    
    # 결과 병합
    return pd.concat(results)

# 대용량 데이터 생성 (1000만 행)
df = pd.DataFrame({
    'col1': np.random.randint(0, 100, 10_000_000),
    'col2': np.random.choice(['A','B','C'], 10_000_000),
    'col3': np.random.randn(10_000_000)
})

# 복잡한 조건 정의
conditions = [
    (df['col1'] > 50) & (df['col2'] == 'A'),
    (df['col1'] < 20) | (df['col3'].abs() > 2),
    df['col2'].isin(['B','C'])
]

choices = ['High A', 'Low or Outlier', 'B or C']

# 멀티프로세싱 적용
df['category'] = parallel_select(df, conditions, choices, default='Other')