# 匯入包
from sqlalchemy import create_engine
import pandas as pd
from string import Template
engine = create_engine("oracle://user:[email protected]***:***/racdb", echo=False)
# 初始化引擎
engine = create_engine('postgresql+psycopg2://' + pg_username + ':' + pg_password + '@' + pg_host + ':' + str(
    pg_port) + '/' + pg_database)
query_sql = """
      select * from $arg1
      """
query_sql = Template(query_sql) # template方法
df = pd.read_sql_query(query_sql .substitute(arg1=tablename),engine) # 配合pandas的方法讀取資料庫值
# 配合pandas的to_sql方法使用十分方便（dataframe物件直接入庫）
df.to_sql(table, engine, if_exists='replace', index=False) #覆蓋入庫
df.to_sql(table, engine, if_exists='append', index=False)  #增量入庫
 

注意：上述df.to_sql的方法實在是太慢太慢了，千萬的資料chunksize設定為萬，上傳了5個小時 鬱悶。查資料後得知以下方法：速度極快！！！！！

def write_to_table(df, table_name, if_exists='fail'):
    import io
    import pandas as pd
    from sqlalchemy import create_engine
    db_engine = create_engine('postgresql://***:***@***:***/***')# 初始化引擎
    string_data_io = io.StringIO()
    df.to_csv(string_data_io, sep='|', index=False)
    pd_sql_engine = pd.io.sql.pandasSQL_builder(db_engine)
    table = pd.io.sql.SQLTable(table_name, pd_sql_engine, frame=df,
                               index=False, if_exists=if_exists,schema = 'goods_code')
    table.create()
    string_data_io.seek(0)
    string_data_io.readline()  # remove header
    with db_engine.connect() as connection:
        with connection.connection.cursor() as cursor:
            copy_cmd = "COPY goods_code.%s FROM STDIN HEADER DELIMITER '|' CSV" %table_name
            cursor.copy_expert(copy_cmd, string_data_io)
        connection.connection.commit()