diff --git a/.gitignore b/.gitignore
index 13d1490..dd1d1f0 100644
--- a/.gitignore
+++ b/.gitignore
@@ -128,4 +128,4 @@ dmypy.json
 
 # Pyre type checker
 .pyre/
-
+.idea
diff --git a/Pipfile b/Pipfile
new file mode 100644
index 0000000..8993b7d
--- /dev/null
+++ b/Pipfile
@@ -0,0 +1,13 @@
+[[source]]
+url = "https://pypi.douban.com/simple"
+verify_ssl = false
+name = "pypi"
+
+[packages]
+kafka-python = "*"
+clickhouse-driver = "*"
+
+[dev-packages]
+
+[requires]
+python_version = "3.8"
diff --git a/__init__.py b/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ck/__init__.py b/ck/__init__.py
new file mode 100644
index 0000000..4f98d5a
--- /dev/null
+++ b/ck/__init__.py
@@ -0,0 +1 @@
+from ck import *
diff --git a/ck/ck.py b/ck/ck.py
new file mode 100644
index 0000000..72de08c
--- /dev/null
+++ b/ck/ck.py
@@ -0,0 +1,271 @@
+import time
+from datetime import datetime
+import json
+import re
+
+import arrow
+from clickhouse_driver.client import Client
+
+from settings import settings
+from .struct_cache import StructCacheFile
+from .robot import DDRobot
+
+__all__ = ('CK',)
+
+
+class DataHandler:
+    handler_link = []
+
+    def __init__(self, func):
+        DataHandler.handler_link.append(func)
+
+
+class CK:
+    def __init__(self, client=Client(**settings.CK_CONFIG),
+                 struct_cache=StructCacheFile(),
+                 robot=DDRobot(),
+                 bulk_max=1000):
+
+        self.client = client
+        self.struct_cache = struct_cache
+        self.robot = robot
+        self.type_dict = dict()
+        self.struct_dict = dict()
+        self.bulk_data = dict()
+        self.bulk_max = bulk_max
+        self.__send_ts = int(time.time())
+
+    def up_tb_struct(self, db, tb, data):
+        struct_dict = self.struct_dict.get(f'{db}_{tb}', {})
+
+        # 用于类型转换
+        type_dict = self.type_dict.setdefault(f'{db}_{tb}', {})
+        for k, v in struct_dict.items():
+            type_dict.setdefault(v, set())
+            type_dict[v].add(k)
+        # 更新结构记录
+        self.struct_cache.update(db, tb, data)
+
+    @DataHandler
+    def date2utc(self, db, tb, data: dict):
+        """
+        日期置为utc时间
+        :param data:
+        :return:
+        """
+        for k in self.type_dict[f'{db}_{tb}'].get('Nullable(DateTime(\'UTC\'))', set()) | self.type_dict[
+            f'{db}_{tb}'].get(
+            'DateTime(\'UTC\')', set()):
+            try:
+                data[k] = arrow.get(data[k]).shift(hours=-data.get('zone_offset', 8)).format('YYYY-MM-DD HH:mm:ss')
+            except:
+                pass
+
+    @DataHandler
+    def array2str(self, db, tb, data: dict):
+        """
+        数组里统一存字符串
+        :param data:
+        :return:
+        """
+        for k in self.type_dict[f'{db}_{tb}'].get('Array(String)', set()):
+            try:
+                data[k] = [str(v) for v in data[k]]
+            except:
+                pass
+
+    @DataHandler
+    def up_user(self, db, tb, data: dict):
+        """
+        从视图中查出最新user info 与当前user 合并
+        :param db:
+        :param tb:
+        :param data:
+        :return:
+        """
+        if tb != 'user':
+            return
+        sql = f"SELECT * from {db}.user_view where account_id='{data['account_id']}'"
+        d, col = self.client.execute(sql, with_column_types=True)
+        if not d:
+            return
+        for v, c in zip(d[0], col):
+            c = c[0]
+            if c not in data:
+                data[c] = v.strftime('%Y-%m-%d %H:%M:%S') if isinstance(v, datetime) else v
+
+    def __notify(self, *args, **kwargs):
+        """
+        预留机器人通知
+        :param args:
+        :param kwargs:
+        :return:
+        """
+        if self.robot:
+            self.robot.send(*args, **kwargs)
+        else:
+            pass
+
+    def get_tb_struct_cache(self, db, tb):
+        """
+        查一条记录 取字段 和类型
+        :param db:
+        :param tb:
+        :return:
+        """
+        if self.struct_dict.get(f'{db}_{tb}'):
+            return self.struct_dict.get(f'{db}_{tb}')
+
+        sql = f'select * from {db}.{tb} limit 1'
+        _, columns = self.client.execute(sql, with_column_types=True)
+        res = {item[0]: item[1] for item in columns}
+        self.struct_dict[f'{db}_{tb}'] = res
+
+        # s = self.client.execute(f'show create {db}.{tb}')[0][0]
+        # s = re.match(r'.*?\((.*?)\)\n?ENGINE', s, re.S).group(1)
+        # res = dict()
+        # for row in s.split('\n'):
+        #     row = row.strip()
+        #     if not row:
+        #         continue
+        #     row = re.sub("[,`]", '', row)
+        #     k, t = row.split(' ')[:2]
+        #     res[k] = t
+        #
+        # self.struct_dict[f'{db}_{tb}'] = res
+        self.up_tb_struct(db, tb, res)
+        return res
+
+    @staticmethod
+    def is_valid_date(date: str):
+        try:
+            res = arrow.get(date)
+            return res
+        except:
+            return False
+
+    def update_user_view(self, db, tb):
+        """
+        更新视图
+        :param db:
+        :param tb:
+        :return:
+        """
+        if tb != 'user':
+            return
+        sql = f"""drop table {db}.user_view;create view {db}.user_view as select *
+                 from {db}.user
+                 order by role_create_time desc
+                 LIMIT 1 by account_id"""
+        self.client.execute(sql)
+
+    def alter_table(self, db, tb, data):
+        """
+        数据库字段检查
+        添加新字段为第一次出现类型
+
+        如果要修改字段类型  存在类型转换问题。停止程序，删除列
+        :param db:
+        :param tb:
+        :param data:
+        :return:
+        """
+        default_field = self.get_tb_struct_cache(db, tb)
+        keys = set(default_field)
+        for k, v in data.items():
+            if k not in default_field:
+                if isinstance(v, str):
+                    if self.is_valid_date(v):
+                        default_field[k] = "Nullable(DateTime('UTC'))"
+                    else:
+                        default_field[k] = 'Nullable(String)'
+
+                if isinstance(v, int):
+                    default_field[k] = 'Nullable(UInt64)'
+
+                if isinstance(v, float):
+                    default_field[k] = 'Nullable(Float32)'
+
+                if isinstance(v, list):
+                    default_field[k] = 'Array(String)'
+
+                if isinstance(v, bool):
+                    default_field[k] = 'Nullable(UInt8)'
+
+                sql = f'alter table {db}.{tb} add column `{k}` {default_field[k]}'
+                print(sql)
+                try:
+                    self.client.execute(sql)
+                except Exception as e:
+                    print(f'添加字段 {k} 失败')
+                    default_field.pop(k)
+                else:
+                    self.update_user_view(db, tb)
+
+        if set(default_field) - keys:
+            self.up_tb_struct(db, tb, default_field)
+
+    def __send(self, db, tb, data):
+        """
+        sql写入需要字段名或全字段
+        这里 json 格式写入
+        超过错误允许率去掉那条记录再次递归
+        一般都是 类型不匹配错误
+
+        :param db:
+        :param tb:
+        :param data:
+        :return:
+        """
+        if not data:
+            return
+        sql = f'INSERT INTO {db}.{tb} FORMAT JSONEachRow '
+        sql = sql + '\n'.join(data)
+        try:
+            # 允许20%错误率
+            self.client.execute('set input_format_allow_errors_ratio=0.2')
+            self.client.execute(sql)
+        except Exception as e:
+            # 丢弃错误行 再次发送
+            if e.code == 26:
+                m = re.match('(.*)?Stack trace', e.message)
+                if m:
+                    error_msg = m.group(1)
+                    error_row = re.match('.*?errors out of (\d+) rows', error_msg)
+                    if error_row:
+                        error_row = int(error_row.group(1)) - 1
+                        error_data = data.pop(error_row)
+                        self.__notify(error_msg, error_data)
+                        self.__send(db, tb, data)
+        else:
+            print(f'{db}.{tb}插入{len(data)}条')
+
+        finally:
+            data.clear()
+
+    def __add(self, db, tb, msg):
+        """
+        列表缓存池
+        :param db:
+        :param tb:
+        :param msg:
+        :return:
+        """
+        bulk_data = self.bulk_data.setdefault(f'{db}', {'event': [], 'user': []})
+        bulk_data[tb].append(json.dumps(msg))
+        ts = int(time.time())
+        # 满足其一条件 写入ck
+        if len(bulk_data[tb]) >= self.bulk_max or self.__send_ts + 60 <= ts:
+            self.__send_ts = ts
+            self.__send(db, tb, bulk_data[tb])
+
+    def send(self, db, tb, msg):
+        params = (db, tb, msg)
+        self.alter_table(*params)
+        # 数据加工链
+        [f(self, *params) for f in DataHandler.handler_link]
+        self.__add(*params)
+
+
+if __name__ == '__main__':
+    ck_client = CK()
diff --git a/ck/robot.py b/ck/robot.py
new file mode 100644
index 0000000..3c663ef
--- /dev/null
+++ b/ck/robot.py
@@ -0,0 +1,9 @@
+class DDRobot:
+    def __init__(self):
+        pass
+
+    def send(self, *args, **kwargs):
+        if args:
+            print(args)
+        if kwargs:
+            print(kwargs)
diff --git a/ck/struct_cache.py b/ck/struct_cache.py
new file mode 100644
index 0000000..17f06ec
--- /dev/null
+++ b/ck/struct_cache.py
@@ -0,0 +1,11 @@
+import json
+
+
+class StructCacheFile:
+    def __init__(self):
+        pass
+
+    @staticmethod
+    def update(db, tb, data):
+        with open(f'{db}_{tb}.json', 'w') as f:
+            json.dump(data, f)
diff --git a/common.py b/common.py
new file mode 100644
index 0000000..77f753f
--- /dev/null
+++ b/common.py
@@ -0,0 +1,13 @@
+from kafka import KafkaConsumer
+from settings import settings
+
+__all__ = 'consumer',
+
+
+def consumer():
+    c = KafkaConsumer(**settings.KAFKA_CONSUMER_CONF)
+    c.subscribe(settings.SUBSCRIBE_TOPIC)
+    for msg in c:
+        topic = msg.topic
+        val = msg.value
+        yield topic, val
diff --git a/main.py b/main.py
new file mode 100644
index 0000000..018e06e
--- /dev/null
+++ b/main.py
@@ -0,0 +1,29 @@
+from ck.ck import CK
+from common import *
+
+from settings import settings
+import traceback
+
+ck_client = CK()
+
+
+def run():
+    for topic, msg in consumer():
+        # print(msg)
+        try:
+            db = settings.APPID_TO_CKDB.get(msg['app_id'])
+            if 'user' in msg['type']:
+                table = 'user'
+            elif 'track' in msg['type']:
+                table = 'event'
+            else:
+                continue
+            del msg['type']
+
+            ck_client.send(db, table, msg)
+        except Exception as e:
+            print(traceback.print_exc())
+
+
+if __name__ == '__main__':
+    run()
diff --git a/settings.py b/settings.py
new file mode 100644
index 0000000..35bd256
--- /dev/null
+++ b/settings.py
@@ -0,0 +1,38 @@
+import json
+
+
+class Config:
+    CK_CONFIG = {'host': '119.29.176.224',
+                 'send_receive_timeout': 3}
+
+    SUBSCRIBE_TOPIC = ['legu_test']
+
+    KAFKA_CONSUMER_CONF = {
+        'bootstrap_servers': ["192.168.0.30:9092", "192.168.0.71:9092", "192.168.0.229:9092"],
+        'value_deserializer': json.loads,
+        'group_id': 'legu_group'
+    }
+
+    TOPIC_TO_LEGU = {
+        'a77703e24e6643d08b74a4163a14f74c': 'legu_test',
+        'c3e0409ac18341149877b08f087db640': 'legu_test'
+    }
+
+    APPID_TO_CKDB = {
+        'a77703e24e6643d08b74a4163a14f74c': 'shjy',
+        'c3e0409ac18341149877b08f087db640': 'shjy'
+    }
+
+    REDIS_CONF = {
+        'host': '192.168.0.161',
+        'port': 6379,
+        'password': 'd1Gh*zp5',
+        'ck': 10  # ck
+    }
+
+
+class Debug(Config):
+    pass
+
+
+settings = Debug