一、引入
上一小节介绍了怎么入库到阿里云的 MaxCompute 数仓，其中涉及到 2 种入库方式，一种是转为阿里云的 DataFrame，然后类似 pandas 的 DataFrame 直接写入 MySQL 的方法，将数据写入表中；另外一种是转为列表，再写入 MaxCompute 表。上一小节主要对后者进行展开描述。
有粉丝私聊我说介绍下第一种，本文就来重点探讨下第一种的处理方式。
前面的数据我们通过 pandas 处理，所以这里涉及到将 pandas 的 DataFrame 转为 PyODPS 的 DataFrame 。
用Pandas DataFrame初始化时，PyODPS DataFrame 会尝试对 NUMPY OBJECT 或 STRING 类型进行推断。如果一整列都为空，则会报错。为避免报错，我们可以设置 unknown_as_string 值为 True，将这些列指定为 STRING 类型。如果 Pandas DataFrame 中包含 LIST 或 DICT 列，系统不会推断该列的类型，必须手动使用 as_type 指定类型。as_type参数类型必须是 DICT。
下面具体来探讨一下。
二、迭代入库逻辑
2.1 新增字段类型映射
阿里云的 DataFrame 拥有自己的类型系统，进行表初始化时，MaxCompute的类型会被转换成对应的DataFrame类型，以便支持更多类型的计算后端。
目前，DataFrame的执行后端支持MaxCompute SQL、Pandas和数据库（MySQL和Postgres）。数据类型的映射关系如下：
| MaxCompute类型 | DataFrame类型 |
|---|---|
| BIGINT | INT64 |
| DOUBLE | FLOAT64 |
| STRING | STRING |
| DATETIME | DATETIME |
| BOOLEAN | BOOLEAN |
| DECIMAL | DECIMAL |
| ARRAY<VALUE_TYPE> | LIST<VALUE_TYPE> |
| MAP<KEY_TYPE, VALUE_TYPE> | DICT<KEY_TYPE, VALUE_TYPE> |
| 当options.sql.use_odps2_extension=True时，还支持以下数据类型。 | |
| TINYINT | INT8 |
| SMALLINT | INT16 |
| INT | INT32 |
| FLOAT | FLOAT32 |
为了保证通用性和准确性，**建议通过 as_type 参数，指定数据类型。**避免由于 PyODPS DataFrame 推断有出入导致发生报错。
所以，需要加一层字段的映射关系。前面已经加了一层字段类型的映射关系，此处直接加上 PyODPS DataFrame 的类型即可：将double映射为float64，将array<string>映射为list<string>。
参考如下：
# 关联入库数据类型
data_type_map = [{"feishu_type": 1 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 2 ,"mc_type": "double" ,"mc_df_type": "float64" }
,{"feishu_type": 3 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 4 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 5 ,"mc_type": "datetime" ,"mc_df_type": "datetime" }
,{"feishu_type": 7 ,"mc_type": "boolean" ,"mc_df_type": "boolean" }
,{"feishu_type": 11 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 13 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 15 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 17 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 18 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 19 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 20 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 21 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 22 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 23 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 1001,"mc_type": "datetime" ,"mc_df_type": "datetime" }
,{"feishu_type": 1002,"mc_type": "datetime" ,"mc_df_type": "datetime" }
,{"feishu_type": 1003,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 1004,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 1005,"mc_type": "string" ,"mc_df_type": "string" }]
2.2 构建 PyODPS 的 DataFrame 的 astype
astype参数的值，可以传递一个字典，格式如：{“字段1”:“字段1类型”, “字段2”:“字段2类型”}。
在生成astype之前，通过merge_list()将字段映射fields_map、飞书列信息feishu_fields和数据类型映射data_type_map三者进行合并，返回store_fields_info_df，所以在生成astype时，只需要从store_fields_info_df中提取tb_field_name和mc_df_type，然后再加上record_id和last_modified_time即可。
def generate_astype(store_fields_info_df):
"""
功能：对整合了飞书列入库的列、飞书列类型、和阿里云 DataFrame 数据类型的 pandas DataFrame，提取字段名和阿里云DataFrame 数据类型
"""
#生成阿里云DataFrame的as_type，用于将处理好的数据转为阿里云DataFrame，格式：{'field_no': 'string', 'field_select': 'float64', 'field_text': 'datatime'}
mc_df_astype = store_fields_info_df.set_index('tb_field_name').to_dict()['mc_df_type']
#加上record_id
mc_df_astype['record_id'] = 'string'
#加一个表更新时间
mc_df_astype['last_modified_time'] = 'DATETIME'
# print(mc_df_astype)
print('成功生成阿里云 DataFrame 的 as_type。方法：generate_astype')
return mc_df_astype
2.3 生成 PyODPS DataFrame 并写入数仓表
def insert_mc_table(feishu_df, mc_table_name, mc_df_astype):
"""
生成 PyODPS DataFrame 并将数据插入 Maxcompute 表
"""
aliyun_df = DataFrame(feishu_df,as_type=mc_df_astype)
aliyun_df.persist(mc_table_name)
print(f'成功将飞书数据写入 MaxCompute 数据表：{mc_table_name}。关联方法：insert_mc_table。')
2.4 迭代定制化函数
定制化函数也需要修改，原有的逻辑不用变，新增对astype的改动即可。
PyODPS DataFrame 包括以下类型：int8，int16，int32，int64，float32，float64，boolean，string，decimal，datetime，list，dict，不含 date，所以日期列不需要处理。
def custom_field(df_return, columns, columns_index, mc_df_astype):
# 2.1 场景一：把数字入库为 int 类型
# 修改 SQL 即可
# cre_ddl = cre_ddl.replace('field_number float','field_number int')
column = Column(name='field_number', type='bigint', comment=columns_index['field_number'][1])
columns[columns_index['field_number'][0]] = column
#修改mc_df_astype
mc_df_astype['field_number'] = 'int64'
# 2.2 场景二：把日期入库为 date 类型
# 修改 df，MySQL会自动截断，Maxcompute不行，需要使用 x.date() 处理
df_return['field_createdtime'] = df_return['field_createdtime'].apply(lambda x:x.date())
# 修改 SQL
# cre_ddl = cre_ddl.replace('field_createdtime datetime','field_createdtime date')
column = Column(name='field_createdtime', type='date', comment=columns_index['field_createdtime'][1])
columns[columns_index['field_createdtime'][0]] = column
# #修改mc_df_astype
# mc_df_astype['field_createdtime'] = 'date'
# 2.3 场景三：日期给定默认最大值
# 修改 df 即可
#默认值改为 2222-01-01 00:00:00
mask = df_return['field_date'] == pd.Timestamp('1970-01-01 08:00:01')
df_return.loc[mask, 'field_date'] = pd.Timestamp('2222-01-01 00:00:00')
# 2.4 场景四：公式保留具体值
# 修改 df
# 修改 SQL
df_return['field_numformula'] = df_return['field_numformula'].apply(lambda x:json.loads(x)['value'][0])
# cre_ddl = cre_ddl.replace('field_numformula varchar(256)','field_numformula int')
column = Column(name='field_numformula', type='bigint', comment=columns_index['field_numformula'][1])
columns[columns_index['field_numformula'][0]] = column
#修改mc_df_astype
mc_df_astype['field_numformula'] = 'int64'
# 创建新的 schema
schema = Schema(columns=columns)
print('定制函数打印数据和建表语句')
print('----------------------------------------------\n', df_return[['field_number','field_createdtime','field_date','field_numformula']].head(5))
print('----------------------------------------------\n', schema.columns)
return df_return, schema
三、整合代码
将上一小节最终的整合代码结合上面的三项内容进行修改。
- 修改
data_type_map：加上键值对mc_df_type； - 使用
generate_astype()+insert_mc_table()替换replace_nan_with_none()和insert_mc_table()； - 替换定制化函数的内容
最终参考代码如下：
import requests
import json
import datetime
import pandas as pd
from sqlalchemy import create_engine, text
from urllib.parse import urlparse, parse_qs
from odps.models import Schema, Column
import math
def get_table_params(bitable_url):
# bitable_url = "https://feishu.cn/base/aaaaaaaa?table=tblccc&view=vewddd"
parsed_url = urlparse(bitable_url) #解析url：(ParseResult(scheme='https', netloc='feishu.cn', path='/base/aaaaaaaa', params='', query='table=tblccc&view=vewddd', fragment='')
query_params = parse_qs(parsed_url.query) #解析url参数：{'table': ['tblccc'], 'view': ['vewddd']}
app_token = parsed_url.path.split('/')[-1]
table_id = query_params.get('table', [None])[0]
view_id = query_params.get('view', [None])[0]
print(f'成功解析链接，app_token：{app_token}，table_id：{table_id}，view_id：{view_id}。关联方法：get_table_params。')
return app_token, table_id, view_id
def get_tenant_access_token(app_id, app_secret):
url = "https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal"
payload = json.dumps({
"app_id": app_id,
"app_secret": app_secret
})
headers = {'Content-Type': 'application/json'}
response = requests.request("POST", url, headers=headers, data=payload)
tenant_access_token = response.json()['tenant_access_token']
print(f'成功获取tenant_access_token：{tenant_access_token}。关联函数：get_table_params。')
return tenant_access_token
def get_bitable_datas(tenant_access_token, app_token, table_id, view_id, page_token='', page_size=20):
url = f"https://open.feishu.cn/open-apis/bitable/v1/apps/{app_token}/tables/{table_id}/records/search?page_size={page_size}&page_token={page_token}&user_id_type=user_id"
payload = json.dumps({"view_id": view_id})
headers = {
'Content-Type': 'application/json',
'Authorization': f'Bearer {tenant_access_token}'
}
response = requests.request("POST", url, headers=headers, data=payload)
print(f'成功获取page_token为【{page_token}】的数据。关联函数：get_bitable_datas。')
return response.json()
def get_all_bitable_datas(tenant_access_token, app_token, table_id, view_id, page_token='', page_size=20):
has_more = True
feishu_datas = []
while has_more:
response = get_bitable_datas(tenant_access_token, app_token, table_id, view_id, page_token, page_size)
if response['code'] == 0:
page_token = response['data'].get('page_token')
has_more = response['data'].get('has_more')
# print(response['data'].get('items'))
# print('\n--------------------------------------------------------------------\n')
feishu_datas.extend(response['data'].get('items'))
else:
raise Exception(response['msg'])
print(f'成功获取飞书多维表所有数据，返回 feishu_datas。关联函数：get_all_bitable_datas。')
return feishu_datas
def get_bitable_fields(tenant_access_token, app_token, table_id, page_size=500):
url = f"https://open.feishu.cn/open-apis/bitable/v1/apps/{app_token}/tables/{table_id}/fields?page_size={page_size}"
payload = ''
headers = {'Authorization': f'Bearer {tenant_access_token}'}
response = requests.request("GET", url, headers=headers, data=payload)
field_infos = response.json().get('data').get('items')
print('成功获取飞书字段信息，关联函数：get_bitable_fields。')
return field_infos
def merge_list(ls_from, ls_join, on=None, left_on=None, right_on=None):
"""将两个[{},{}]结构的数据合并"""
df_from = pd.DataFrame(ls_from)
df_join = pd.DataFrame(ls_join)
if on is not None:
df_merge = df_from.merge(df_join, how='left', on=on)
else:
df_merge = df_from.merge(df_join, how='left', left_on=left_on, right_on=right_on) # , suffixes=('', '_y')
print(f'成功合并列表或DataFrame。关联方法：merge_list。')
return df_merge
def extract_key_fields(feishu_datas, store_fields_info_df):
"""处理飞书数据类型编号的数据"""
print('开始处理飞书多维表关键字段数据...')
# 需要record_id 和 订单号，用于和数据库数据匹配
df_feishu = pd.DataFrame(feishu_datas)
df_return = pd.DataFrame()
#根据列的数据类型，分别处理对应的数据。注意：仅返回以下列举的数据类型，如果fields_map的内容包含按钮、流程等数据类型的飞书列，忽略。
for index, row in store_fields_info_df.iterrows():
if row['type'] == 1: #文本
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:x.get(row['field_name'],[{}])[0].get('text'))
elif row['type'] in (2, 3, 4, 7, 13, 1005): #数字、单选、多选、复选框、手机号、自动编号
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:x.get(row['field_name']))
elif row['type'] in (5, 1001, 1002): #日期(包含创建和更新)，需要加 8 小时，即 8*60*60*1000=28800 秒
df_return[row['tb_field_name']] = pd.to_datetime(df_feishu['fields'].apply(lambda x:28800 + int(x.get(row['field_name'],1000)/1000)), unit='s')
elif row['type'] in(11, 23, 1003, 1004): #人员、群组、创建人、修改人，遍历取name
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x: ','.join([i.get('name') for i in x.get(row['field_name'],[{"name":""}])])) # 需要遍历
elif row['type'] == 15: #链接
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:x.get(row['field_name'],{}).get('link'))
elif row['type'] == 17: #附件，遍历取url
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:[i.get('url') for i in x.get(row['field_name'],[{}])]) #需要遍历
elif row['type'] in(18, 21): #单向关联、双向关联，取link_record_ids
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:x.get(row['field_name'],{}).get('link_record_ids'))
elif row['type'] in(19, 20): #查找引用和公式
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:json.dumps(x.get(row['field_name'])))
elif row['type'] == 22: #地理位置
df_return[row['tb_field_name']] = df_feishu['fields'].apply(lambda x:x.get(row['field_name'],{}).get('location'))
#加上record_id
df_return['record_id'] = df_feishu.record_id
#加上表更新字段
df_return['last_modified_time'] = datetime.datetime.now()
print(f'成功提取入库字段的数据。关联方法：extract_key_fields。')
return df_return
def generate_create_schema(store_fields_info_df):
columns = []
columns_index = {}
for index, row in store_fields_info_df.iterrows():
name = row['tb_field_name']; type=row['mc_type']; comment=row['feishu_field_name']
# print(name,type,comment)
columns.append(Column(name=name, type=type, comment=comment))
columns_index[name] = (index, comment)
columns.append(Column(name='record_id', type='string', comment='飞书表行record_id'))
columns.append(Column(name='last_modified_time', type='datetime', comment='数据更新时间'))
schema = Schema(columns=columns)
print(f'成功生成 MaxCompute 建表 schema。关联方法：generate_create_schema。')
return schema, columns, columns_index
def cre_mc_table(db_table_name, schema):
if o.exist_table(db_table_name):
print(f'表单 {db_table_name} 已存在，不需要新建。关联方法：cre_mc_table。')
else:
table = o.create_table(db_table_name, schema, if_not_exists=True)
print(f'成功创建 MaxCompute 表：{db_table_name}。关联方法：cre_mc_table。')
def generate_astype(store_fields_info_df):
"""
功能：对整合了飞书列入库的列、飞书列类型、和阿里云 DataFrame 数据类型的 pandas DataFrame，提取字段名和阿里云DataFrame 数据类型
"""
#生成阿里云DataFrame的as_type，用于将处理好的数据转为阿里云DataFrame，格式：{'field_no': 'string', 'field_select': 'float64', 'field_text': 'datatime'}
mc_df_astype = store_fields_info_df.set_index('tb_field_name').to_dict()['mc_df_type']
#加上record_id
mc_df_astype['record_id'] = 'string'
#加一个表更新时间
mc_df_astype['last_modified_time'] = 'DATETIME'
# print(mc_df_astype)
print('成功生成阿里云 DataFrame 的 as_type。方法：generate_astype')
return mc_df_astype
def insert_mc_table(feishu_df, mc_table_name, mc_df_astype):
"""
生成 PyODPS DataFrame 并将数据插入 Maxcompute 表
"""
aliyun_df = DataFrame(feishu_df,as_type=mc_df_astype)
aliyun_df.persist(mc_table_name)
print(f'成功将飞书数据写入 MaxCompute 数据表：{mc_table_name}。关联方法：insert_mc_table。')
def custom_field(df_return, columns, columns_index, mc_df_astype):
# 2.1 场景一：把数字入库为 int 类型
# 修改 SQL 即可
# cre_ddl = cre_ddl.replace('field_number float','field_number int')
column = Column(name='field_number', type='bigint', comment=columns_index['field_number'][1])
columns[columns_index['field_number'][0]] = column
#修改mc_df_astype
mc_df_astype['field_number'] = 'int64'
# 2.2 场景二：把日期入库为 date 类型
# 修改 df，MySQL会自动截断，Maxcompute不行，需要使用 x.date() 处理
df_return['field_createdtime'] = df_return['field_createdtime'].apply(lambda x:x.date())
# 修改 SQL
# cre_ddl = cre_ddl.replace('field_createdtime datetime','field_createdtime date')
column = Column(name='field_createdtime', type='date', comment=columns_index['field_createdtime'][1])
columns[columns_index['field_createdtime'][0]] = column
# #修改mc_df_astype
# mc_df_astype['field_createdtime'] = 'date'
# 2.3 场景三：日期给定默认最大值
# 修改 df 即可
#默认值改为 2222-01-01 00:00:00
mask = df_return['field_date'] == pd.Timestamp('1970-01-01 08:00:01')
df_return.loc[mask, 'field_date'] = pd.Timestamp('2222-01-01 00:00:00')
# 2.4 场景四：公式保留具体值
# 修改 df
# 修改 SQL
df_return['field_numformula'] = df_return['field_numformula'].apply(lambda x:json.loads(x)['value'][0])
# cre_ddl = cre_ddl.replace('field_numformula varchar(256)','field_numformula int')
column = Column(name='field_numformula', type='bigint', comment=columns_index['field_numformula'][1])
columns[columns_index['field_numformula'][0]] = column
#修改mc_df_astype
mc_df_astype['field_numformula'] = 'int64'
# 创建新的 schema
schema = Schema(columns=columns)
print('定制函数打印数据和建表语句')
print('----------------------------------------------\n', df_return[['field_number','field_createdtime','field_date','field_numformula']].head(5))
print('----------------------------------------------\n', schema.columns)
return df_return, schema, mc_df_astype
def main(mc_table_name, bitable_url, fields_map):
# 基本配置
app_token, table_id, view_id = get_table_params(bitable_url)
app_id = 'your_app_id'
app_secret = 'your_app_secret'
tenant_access_token = get_tenant_access_token(app_id, app_secret)
page_size = 50
# 获取飞书多维表所有数据
feishu_datas = get_all_bitable_datas(tenant_access_token, app_token, table_id, view_id, page_size=page_size)
#获取飞书字段信息
feishu_fields = get_bitable_fields(tenant_access_token, app_token, table_id)
# 以 fields_map 为准关联数据
store_fields_info_df = merge_list(fields_map, feishu_fields, left_on='feishu_field_name', right_on='field_name')
# 处理入库字段数据
feishu_df = extract_key_fields(feishu_datas, store_fields_info_df)
# 关联入库数据类型
data_type_map = [{"feishu_type": 1 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 2 ,"mc_type": "double" ,"mc_df_type": "float64" }
,{"feishu_type": 3 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 4 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 5 ,"mc_type": "datetime" ,"mc_df_type": "datetime" }
,{"feishu_type": 7 ,"mc_type": "boolean" ,"mc_df_type": "boolean" }
,{"feishu_type": 11 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 13 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 15 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 17 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 18 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 19 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 20 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 21 ,"mc_type": "array<string>" ,"mc_df_type": "list<string>" }
,{"feishu_type": 22 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 23 ,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 1001,"mc_type": "datetime" ,"mc_df_type": "datetime" }
,{"feishu_type": 1002,"mc_type": "datetime" ,"mc_df_type": "datetime" }
,{"feishu_type": 1003,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 1004,"mc_type": "string" ,"mc_df_type": "string" }
,{"feishu_type": 1005,"mc_type": "string" ,"mc_df_type": "string" }]
store_fields_info_df = merge_list(store_fields_info_df, data_type_map, left_on='type', right_on='feishu_type')
# 生成 MaxCompute schema
schema, columns, columns_index = generate_create_schema(store_fields_info_df)
# 生成 DataFrame astype
mc_df_astype = generate_astype(store_fields_info_df)
# 定制化
feishu_df, schema, mc_df_astype = custom_field(feishu_df, columns, columns_index, mc_df_astype)
# 建 MaxCompute 数据表
cre_mc_table(mc_table_name, schema)
# MaxCompute 表插入数据
insert_mc_table(feishu_df, mc_table_name, mc_df_astype)
print(f'成功将飞书多维表（{bitable_url}）的数据入库到 mysql 数据表：{mc_table_name}。')
if __name__ == '__main__':
mc_table_name = 'for_ods.feishu_data_type_test'
bitable_url = "https://forchangesz.feishu.cn/base/SpY3b9LMFaodpOsE0kdcGEyonbg?table=tbl5BZE0Aubjz5Yy&view=vewDM4NGlP"
fields_map = [{'tb_field_name': 'field_text','feishu_field_name': '文本'}
,{'tb_field_name': 'field_email','feishu_field_name': 'email'}
,{'tb_field_name': 'field_select','feishu_field_name': '单选'}
,{'tb_field_name': 'field_mobile','feishu_field_name': '电话号码'}
,{'tb_field_name': 'field_no','feishu_field_name': '自动编号'}
,{'tb_field_name': 'field_member1','feishu_field_name': '人员1'}
,{'tb_field_name': 'field_group1','feishu_field_name': '群组1'}
,{'tb_field_name': 'field_creator','feishu_field_name': '创建人'}
,{'tb_field_name': 'field_modifier','feishu_field_name': '修改人'}
,{'tb_field_name': 'field_member2','feishu_field_name': '人员2'}
,{'tb_field_name': 'field_group2','feishu_field_name': '群组2'}
,{'tb_field_name': 'field_url','feishu_field_name': '超链接'}
,{'tb_field_name': 'field_location','feishu_field_name': '地理位置'}
,{'tb_field_name': 'field_findnum','feishu_field_name': '查找引用数值'}
,{'tb_field_name': 'field_numformula','feishu_field_name': '数字公式'}
,{'tb_field_name': 'field_number','feishu_field_name': '数字'}
,{'tb_field_name': 'field_progress','feishu_field_name': '进度'}
,{'tb_field_name': 'field_money','feishu_field_name': '货币'}
,{'tb_field_name': 'field_Rating','feishu_field_name': '评分'}
,{'tb_field_name': 'field_bool','feishu_field_name': '复选框'}
,{'tb_field_name': 'field_date','feishu_field_name': '日期'}
,{'tb_field_name': 'field_createdtime','feishu_field_name': '创建时间'}
,{'tb_field_name': 'field_updatedtime','feishu_field_name': '更新时间'}
,{'tb_field_name': 'field_mulselect','feishu_field_name': '多选'}
,{'tb_field_name': 'field_singleunion','feishu_field_name': '单向关联'}
,{'tb_field_name': 'field_doubleunion','feishu_field_name': '双向关联'}
,{'tb_field_name': 'field_file','feishu_field_name': '附件'}
]
main(mc_table_name, bitable_url, fields_map)
最终执行的结果和上一小节一致，参考如下：
四、小结
本文探讨了怎么通过 PyODPS 的 DataFrame 将飞书数据入库，主要涉及四点：新增 PyODPS 的 DataFrame 的数据类型映射、定义 astype、将飞书数据转为 PyODPS 的 DataFrame 并入库和定制化中新增 astype 的修改。
PyODPS 的 DataFrame 更多用于数据科学计算，方便将分析的结果数据保存到表中，此处仅用它作为一个中间桥梁，将 pandas 的 DataFrame 和数仓表连接起来。
