如何通过Snowflake Snowpark将Pandas DataFrame写入Snowflake数据库？

使用Snowflake Snowpark将Pandas DataFrame（如描述统计、相关矩阵、模型输出等转换结果）写入Snowflake数据库时，报错：

'DataFrame' object has no attribute 'write'

附原代码：

from snowflake.snowpark import Session
from snowflake.snowpark import table
from snowflake.snowpark.functions import udf
from snowflake.snowpark.functions import col
from snowflake.snowpark.types import StringType

sess = None

print('Connecting...')

cnn_params = {
"account": "eu-west-1",
"user": "user",
"password": 'pass',
"warehouse": "xs",
"database": "demodb",
"schema": "demosch",
"role": "ACCOUNTADMIN"
}

try:
    print('session...')
    import pandas as pd
    import numpy as np
    sess = Session.builder.configs(cnn_params).create()
    
    df = pd.DataFrame(np.random.randint(0,100,size=(100, 4)), columns=list('ABCD'))
    #print(df.describe())
    
    df.describe().write.mode("overwrite").save_as_table("describe_output", table_type="temporary")
    
except Exception as e:
    print(e)
finally:
    if sess:
        sess.close()
        print('connection closed...')
print('done.')

df.describe()返回的是Pandas原生DataFrame，而write、save_as_table是Snowpark DataFrame专属的方法，Pandas DataFrame并没有这些属性，因此触发报错。

需要先将Pandas DataFrame转换为Snowpark DataFrame，再调用写入方法。核心是使用Snowpark Session的create_dataframe()方法完成转换。

修改后的代码：

from snowflake.snowpark import Session
import pandas as pd
import numpy as np

sess = None

print('Connecting...')

cnn_params = {
"account": "eu-west-1",
"user": "user",
"password": 'pass',
"warehouse": "xs",
"database": "demodb",
"schema": "demosch",
"role": "ACCOUNTADMIN"
}

try:
    print('session...')
    sess = Session.builder.configs(cnn_params).create()
    
    # 生成Pandas DataFrame并计算描述统计
    pd_df = pd.DataFrame(np.random.randint(0,100,size=(100, 4)), columns=list('ABCD'))
    desc_pd_df = pd_df.describe()
    
    # 将Pandas DataFrame转换为Snowpark DataFrame，reset_index把统计指标转为列
    snowpark_df = sess.create_dataframe(desc_pd_df.reset_index())
    
    # 写入Snowflake临时表
    snowpark_df.write.mode("overwrite").save_as_table("describe_output", table_type="temporary")
    print("数据写入成功")
    
except Exception as e:
    print(e)
finally:
    if sess:
        sess.close()
        print('connection closed...')
print('done.')

• reset_index()：describe()生成的结果中，统计指标（如count、mean）是索引，转换为Snowpark DataFrame时需要将其转为普通列，否则会丢失这部分核心信息。
• sess.create_dataframe()：这是Snowpark将Pandas DataFrame转为Snowpark DataFrame的核心方法，转换后才能使用Snowpark的写入API。

内容的提问来源于stack exchange，提问作者314mip