tfduck-bsd 0.16.4__py3-none-any.whl → 0.17.0__py3-none-any.whl

tfduck/__init__.py

@@ -1 +1 @@
-__version__="0.16.4"
+__version__="0.17.0"

tfduck/tga/base_tga.py

@@ -9,6 +9,7 @@ class BaseTga(object):
 
     def tran_dt_by_zone(self, field_name, field_zone_offset, target_offset=0):
         """
+        #zone_offset是#event_time当对于utc时区的偏移量
         @des: 调用
         tran_zone_sql = cobj.tran_zone_sql(a."#event_time", a."#zone_offset", 0)  # 统一转为utc
         例如：

tfduck/tga/predict_sql_ltv.py

@@ -61,8 +61,10 @@ def call_method(self, ctx=None, **kwargs):
                         ) as %(sub_field_name)s
                         """
     }       
-    # 额外的用户属性比如(可选参数)
-    addon_attrs = ["#screen_width", "#screen_height"]         
+    # 额外的event表属性和预查询的用户属性base_user_table_fields里面选，比如(可选参数)
+    addon_attrs = ["#screen_width", "#screen_height"]    
+    # 是否等待用户注册后经过max(gconf.level_durs)的时间才开始计算特征值(可选参数)---默认是不等待
+    wait_dur_get_feature     
     """
     创建调用实例---最好通过dict创建实例
     """         

tfduck/tga/predict_sql_retain.py

@@ -61,8 +61,10 @@ def call_method(self, ctx=None, **kwargs):
                         ) as %(sub_field_name)s
                         """
     }      
-    # 额外的用户属性比如(可选参数)
-    addon_attrs = ["#screen_width", "#screen_height"]        
+    # 额外的event表属性和预查询的用户属性base_user_table_fields里面选，比如(可选参数)
+    addon_attrs = ["#screen_width", "#screen_height"]       
+    # 是否等待用户注册后经过max(gconf.level_durs)的时间才开始计算特征值(可选参数)---默认是不等待
+    wait_dur_get_feature
     """
     创建调用实例---最好通过dict创建实例
     注意： 针对一个tga项目属于多个项目的打点情况,例如3dfish， 可以适时启用user_cols1,user_cols2或者event_cols1
@@ -106,6 +108,7 @@ class PredictFeatureSql(BaseTga):
                  user_cols2='1=1',
                  event_cols1='1=1',
                  addon_attrs=[], 
+                 wait_dur_get_feature=False,
                  **kwargs):
         """
         @des: 参数说明看上面的文档说明
@@ -135,6 +138,8 @@ class PredictFeatureSql(BaseTga):
         self.user_cols2 = user_cols2
         self.event_cols1 = event_cols1
         self.addon_attrs = addon_attrs if isinstance(addon_attrs, list) else []
+        #
+        self.wait_dur_get_feature = wait_dur_get_feature
         # 其他属性
         for k, v in kwargs.items():
             setattr(self, k, v)
@@ -223,6 +228,12 @@ class PredictFeatureSql(BaseTga):
         gconf = self.gconf
         comm_cc = self.get_sql_config()
         #
+        if self.wait_dur_get_feature:
+            # 这里需要转为真正的utc时间和当前时间的差值, 否则会读本地时区，不准确
+            wait_dur_get_feature_sql = f"to_unixtime(current_timestamp)-to_unixtime(try_cast (concat(to_char(event_time_utc, 'yyyy-mm-dd hh24:mi:ss'),' UTC') as timestamp with time zone)) >= {max(gconf.level_durs)}"
+        else:
+            wait_dur_get_feature_sql = "1=1"
+        #
         sql = f"""
     -- des: add by yuanxiao for machine learn predict
     with new_user as (
@@ -273,6 +284,7 @@ class PredictFeatureSql(BaseTga):
                         WHERE
                             event_time_utc >= timestamp '{comm_cc["new_device_start"]}'
                             AND event_time_utc < timestamp '{comm_cc["new_device_end"]}'
+                            AND {wait_dur_get_feature_sql}
                     ) b
                     ON a."#user_id" = b."#user_id"
                 )
@@ -289,9 +301,11 @@ class PredictFeatureSql(BaseTga):
                 a5.user_register_time,
                 a5.user_user_id,
                 floor(to_unixtime(b5.event_time_utc))-floor(to_unixtime(a5.user_register_time)) as rt_dur,
+                -- floor(to_unixtime(try_cast (concat(to_char(b5.event_time_utc, 'yyyy-mm-dd hh24:mi:ss'),' UTC') as timestamp with time zone)))-floor(to_unixtime(try_cast (concat(to_char(a5.user_register_time, 'yyyy-mm-dd hh24:mi:ss'),' UTC') as timestamp with time zone))) as rt_dur,
                 c5."afrawip__meida_source" as "afrawip__meida_source_ikfdssausercommonend",
                 c5."#account_id" as "#account_id_ikfdssausercommonend",
                 c5."#distinct_id" as "#distinct_id_ikfdssausercommonend",
+                c5."afrawip__campaign" as "afrawip__campaign_ikfdssausercommonend",
                 b5.*,
                 b5.event_time_utc as "#event_time"
             from 
@@ -322,7 +336,7 @@ class PredictFeatureSql(BaseTga):
             INNER JOIN
             (
                 SELECT
-                    "#user_id","afrawip__meida_source","#account_id","#distinct_id"
+                    "#user_id","afrawip__meida_source","#account_id","#distinct_id","afrawip__campaign"
                 FROM
                     {gconf.tga_user_table}
             )

tfduck/tga/predict_sql_yh.py

@@ -61,8 +61,10 @@ def call_method(self, ctx=None, **kwargs):
                         ) as %(sub_field_name)s
                         """
     }       
-    # 额外的用户属性比如(可选参数)
-    addon_attrs = ["#screen_width", "#screen_height"]         
+    # 额外的event表属性和预查询的用户属性base_user_table_fields里面选，比如(可选参数)
+    addon_attrs = ["#screen_width", "#screen_height"]     
+    # 是否等待用户注册后经过max(gconf.level_durs)的时间才开始计算特征值(可选参数)---默认是不等待
+    wait_dur_get_feature    
     """
     创建调用实例---最好通过dict创建实例
     """         

tfduck/tga/train_sql_ltv.py

@@ -62,7 +62,7 @@ def call_method(self, ctx=None, **kwargs):
                         ) as %(sub_field_name)s
                         """
     }  
-    # 额外的用户属性比如(可选参数)
+    # 额外的event表属性和预查询的用户属性base_user_table_fields里面选，比如(可选参数)
     addon_attrs = ["#screen_width", "#screen_height"] 
     # line_value_rd  ltv的线性值大于多少算正样本，小于多少算负样本，这样可以控制正负样本比例     
     line_value_rd = 0.5
@@ -243,6 +243,7 @@ class TrainFeatureSql(BaseTga):
                 SELECT
                     a1.user_register_time as user_register_time,
                     a1."#user_id" as user_user_id,
+                    a1."#distinct_id" as user_distinct_id,
                     b1."#user_id" as event_user_id,
                     CASE WHEN b1."yiap__itemrevenue" is NULL THEN 0 ELSE b1."yiap__itemrevenue" END yiap__itemrevenue,    
                     CASE WHEN b1."sdk_ad_price" is NULL THEN 0 ELSE b1."sdk_ad_price" END sdk_ad_price,
@@ -260,11 +261,11 @@ class TrainFeatureSql(BaseTga):
                         FROM
                         (
                             SELECT
-                                b.event_time_utc as user_register_time, a."#user_id"
+                                b.event_time_utc as user_register_time, a."#user_id", a."#distinct_id"
                             FROM
                             (
                                 SELECT
-                                    "#user_id"
+                                    "#user_id","#distinct_id"
                                 FROM
                                     {gconf.tga_user_table}
                                 WHERE
@@ -329,13 +330,19 @@ class TrainFeatureSql(BaseTga):
             a3.revenue_price as is_rd, -- 将指定mode的ltv值作为标签
             a3.user_user_id,
             a3.user_register_time,
-            -- 计算固定随机值，打乱顺序
-            (abs(from_ieee754_64(xxhash64(cast(cast(user_user_id as varchar) as varbinary)))) % 100) / 100. as tt_stable_rand 
+            -- 计算固定随机值，打乱顺序(废弃)--计算一个固定的采样值--现在也失效了，因为tga上云到k8s，不同节点计算，这个值也会变, 而且不同的id生成的数字可能一样，这样就不能达到目的了
+            -- (abs(from_ieee754_64(xxhash64(cast(cast(user_user_id as varchar) as varbinary)))) % 100) / 100. as tt_stable_rand 
+            -- 直接用user_user_id作为排序值，排序的必须是唯一的，否则下面会对不上，会产生很多null的数据
+            -- user_user_id as tt_stable_rand  -- 这种方式最保险，但是不能乱序，这样采样的数据就不是随机分布在每天的 
+            -- bitwise_xor(user_user_id, 906867964886667264) as tt_stable_rand  -- 这种方式可能会产生left null的情况，但是是少数，过滤掉就行，不影响结果，但支持乱序采样
+            a3.user_distinct_id as tt_stable_rand  -- 这种方式最保险，即是乱序也是唯一
+
         FROM
         (
             -- 获取指定日期的注册用户 连接 这些用户在7-10天后的触发的事件 的 数量
             SELECT
                a2_1.user_user_id,
+               a2_1.user_distinct_id,
                a2_1.user_register_time,
                {
                 [
@@ -347,6 +354,7 @@ class TrainFeatureSql(BaseTga):
             FROM (
                 SELECT
                     a2.user_user_id as user_user_id,
+                    a2.user_distinct_id as user_distinct_id,
                     a2.user_register_time as user_register_time,
                     SUM(
                         CASE
@@ -366,7 +374,7 @@ class TrainFeatureSql(BaseTga):
                 (
                     new_user
                 ) a2
-                GROUP BY a2.user_user_id, a2.user_register_time
+                GROUP BY a2.user_user_id, a2.user_distinct_id, a2.user_register_time
             ) a2_1
         ) a3
     )
@@ -427,6 +435,7 @@ class TrainFeatureSql(BaseTga):
                 c5."afrawip__meida_source" as "afrawip__meida_source_ikfdssausercommonend",
                 c5."#account_id" as "#account_id_ikfdssausercommonend",
                 c5."#distinct_id" as "#distinct_id_ikfdssausercommonend",
+                c5."afrawip__campaign" as "afrawip__campaign_ikfdssausercommonend",
                 b5.*,
                 b5.event_time_utc as "#event_time"
             from 
@@ -457,7 +466,7 @@ class TrainFeatureSql(BaseTga):
             INNER JOIN
             (
                 SELECT
-                    "#user_id","afrawip__meida_source","#account_id","#distinct_id"
+                    "#user_id","afrawip__meida_source","#account_id","#distinct_id","afrawip__campaign"
                 FROM
                     {gconf.tga_user_table}
             )
@@ -588,6 +597,7 @@ class TrainFeatureSql(BaseTga):
         on {base_user_table}.user_user_id={base_feature_table}.user_user_id
     )
     -- with结束没有逗号, 过滤左连接没有特征值的行，调试的时候取消where条件
+    -- 出现null过滤掉就行，少量的不管，因为tt_stable_rand的构建方法会有较小影响s
     select * from user_tzz where {self.get_first_col_name()} is not NULL
     """
         return sql

tfduck/tga/train_sql_retain.py

@@ -62,7 +62,7 @@ def call_method(self, ctx=None, **kwargs):
                         ) as %(sub_field_name)s
                         """
     }  
-    # 额外的用户属性比如(可选参数)
+    # 额外的event表属性和预查询的用户属性base_user_table_fields里面选，比如(可选参数)
     addon_attrs = ["#screen_width", "#screen_height"]      
     """
     创建调用实例---最好通过dict创建实例
@@ -232,6 +232,7 @@ class TrainFeatureSql(BaseTga):
                 SELECT
                     a1.user_register_time as user_register_time,
                     a1."#user_id" as user_user_id,
+                    a1."#distinct_id" as user_distinct_id,
                     b1."#user_id" as event_user_id,
                     -- b1."event_time_utc" as pd_event_time_utc,
                     floor(to_unixtime(b1.event_time_utc))-floor(to_unixtime(a1.user_register_time)) as pd_rt_dur
@@ -248,11 +249,12 @@ class TrainFeatureSql(BaseTga):
                         FROM
                         (
                             SELECT
-                                b.event_time_utc as user_register_time, a."#user_id"
+                                b.event_time_utc as user_register_time, a."#user_id", a."#distinct_id"
                             FROM
                             (
                                 SELECT
-                                    "#user_id"
+                                    "#user_id",
+                                    "#distinct_id"
                                 FROM
                                     {gconf.tga_user_table}
                                 WHERE
@@ -314,13 +316,18 @@ class TrainFeatureSql(BaseTga):
             CASE WHEN a3.event_count>0 THEN '1' ELSE '0' END as is_rd,
             a3.user_user_id,
             a3.user_register_time,
-            -- 计算固定随机值，打乱顺序
-            (abs(from_ieee754_64(xxhash64(cast(cast(user_user_id as varchar) as varbinary)))) % 100) / 100. as tt_stable_rand 
+            -- 计算固定随机值，打乱顺序(废弃)--计算一个固定的采样值--现在也失效了，因为tga上云到k8s，不同节点计算，这个值也会变, 而且不同的id生成的数字可能一样，这样就不能达到目的了
+            -- (abs(from_ieee754_64(xxhash64(cast(cast(user_user_id as varchar) as varbinary)))) % 100) / 100. as tt_stable_rand 
+            -- 直接用user_user_id作为排序值，排序的必须是唯一的，否则下面会对不上，会产生很多null的数据
+            -- user_user_id as tt_stable_rand  -- 这种方式最保险，但是不能乱序，这样采样的数据就不是随机分布在每天的 
+            -- bitwise_xor(user_user_id, 906867964886667264) as tt_stable_rand  -- 这种方式也不太行，数据会几种在某一天
+            a3.user_distinct_id as tt_stable_rand  -- 这种方式最保险，即是乱序也是唯一
         FROM
         (
             -- 获取指定日期的注册用户 连接 这些用户在7-10天后的触发的事件 的 数量
             SELECT
                 a2.user_user_id as user_user_id,
+                a2.user_distinct_id as user_distinct_id,
                 a2.user_register_time as user_register_time,
                 SUM(
                     CASE
@@ -333,7 +340,7 @@ class TrainFeatureSql(BaseTga):
             (
                 new_user
             ) a2
-            GROUP BY a2.user_user_id, a2.user_register_time
+            GROUP BY a2.user_user_id, a2.user_distinct_id, a2.user_register_time
         ) a3
     )
     , nav_table as (
@@ -393,6 +400,7 @@ class TrainFeatureSql(BaseTga):
                 c5."afrawip__meida_source" as "afrawip__meida_source_ikfdssausercommonend",
                 c5."#account_id" as "#account_id_ikfdssausercommonend",
                 c5."#distinct_id" as "#distinct_id_ikfdssausercommonend",
+                c5."afrawip__campaign" as "afrawip__campaign_ikfdssausercommonend",
                 b5.*,
                 b5.event_time_utc as "#event_time"
             from 
@@ -423,7 +431,7 @@ class TrainFeatureSql(BaseTga):
             INNER JOIN
             (
                 SELECT
-                    "#user_id","afrawip__meida_source","#account_id","#distinct_id"
+                    "#user_id","afrawip__meida_source","#account_id","#distinct_id","afrawip__campaign"
                 FROM
                     {gconf.tga_user_table}
             )
@@ -554,6 +562,7 @@ class TrainFeatureSql(BaseTga):
         on {base_user_table}.user_user_id={base_feature_table}.user_user_id
     )
     -- with结束没有逗号, 过滤左连接没有特征值的行，调试的时候取消where条件
+    -- 出现null过滤掉就行，少量的不管，因为tt_stable_rand的构建方法会有较小影响
     select * from user_tzz where {self.get_first_col_name()} is not NULL
     """
         return sql

tfduck/tga/train_sql_yh.py

@@ -62,7 +62,7 @@ def call_method(self, ctx=None, **kwargs):
                         ) as %(sub_field_name)s
                         """
     }    
-    # 额外的用户属性比如(可选参数)
+    # 额外的event表属性和预查询的用户属性base_user_table_fields里面选，比如(可选参数)
     addon_attrs = ["#screen_width", "#screen_height"]        
     """
     创建调用实例---最好通过dict创建实例
@@ -232,6 +232,7 @@ class TrainFeatureSql(BaseTga):
                 SELECT
                     a1.user_register_time as user_register_time,
                     a1."#user_id" as user_user_id,
+                    a1."#distinct_id" as user_distinct_id,
                     b1."#user_id" as event_user_id,
                     -- b1."event_time_utc" as pd_event_time_utc,
                     floor(to_unixtime(b1.event_time_utc))-floor(to_unixtime(a1.user_register_time)) as pd_rt_dur
@@ -248,11 +249,11 @@ class TrainFeatureSql(BaseTga):
                         FROM
                         (
                             SELECT
-                                b.event_time_utc as user_register_time, a."#user_id"
+                                b.event_time_utc as user_register_time, a."#user_id", a."#distinct_id"
                             FROM
                             (
                                 SELECT
-                                    "#user_id"
+                                    "#user_id","#distinct_id"
                                 FROM
                                     {gconf.tga_user_table}
                                 WHERE
@@ -318,20 +319,25 @@ class TrainFeatureSql(BaseTga):
             CASE WHEN a3.event_count>0 THEN '1' ELSE '0' END as is_rd,
             a3.user_user_id,
             a3.user_register_time,
-            -- 计算固定随机值，打乱顺序
-            (abs(from_ieee754_64(xxhash64(cast(cast(user_user_id as varchar) as varbinary)))) % 100) / 100. as tt_stable_rand 
+            -- 计算固定随机值，打乱顺序(废弃)--计算一个固定的采样值--现在也失效了，因为tga上云到k8s，不同节点计算，这个值也会变, 而且不同的id生成的数字可能一样，这样就不能达到目的了
+            -- (abs(from_ieee754_64(xxhash64(cast(cast(user_user_id as varchar) as varbinary)))) % 100) / 100. as tt_stable_rand 
+            -- 直接用user_user_id作为排序值，排序的必须是唯一的，否则下面会对不上，会产生很多null的数据
+            -- user_user_id as tt_stable_rand  -- 这种方式最保险，但是不能乱序，这样采样的数据就不是随机分布在每天的 
+            --bitwise_xor(user_user_id, 906867964886667264) as tt_stable_rand  -- 这种方式可能会产生left null的情况，但是是少数，过滤掉就行，不影响结果，但支持乱序采样
+            a3.user_distinct_id as tt_stable_rand  -- 这种方式最保险，即是乱序也是唯一
         FROM
         (
             -- 获取指定日期的注册用户 连接 这些用户在7-10天后的触发的事件 的 数量
             SELECT
                 a2.user_user_id as user_user_id,
+                a2.user_distinct_id as user_distinct_id,
                 a2.user_register_time as user_register_time,
                 SUM(CASE WHEN a2.event_user_id IS NULL THEN 0 ELSE 1 END) AS event_count
             FROM
             (
                 new_user
             ) a2
-            GROUP BY a2.user_user_id, a2.user_register_time
+            GROUP BY a2.user_user_id, a2.user_distinct_id, a2.user_register_time
         ) a3
     )
     , nav_table as (
@@ -391,6 +397,7 @@ class TrainFeatureSql(BaseTga):
                 c5."afrawip__meida_source" as "afrawip__meida_source_ikfdssausercommonend",
                 c5."#account_id" as "#account_id_ikfdssausercommonend",
                 c5."#distinct_id" as "#distinct_id_ikfdssausercommonend",
+                c5."afrawip__campaign" as "afrawip__campaign_ikfdssausercommonend",
                 b5.*,
                 b5.event_time_utc as "#event_time"
             from 
@@ -421,7 +428,7 @@ class TrainFeatureSql(BaseTga):
             INNER JOIN
             (
                 SELECT
-                    "#user_id","afrawip__meida_source","#account_id","#distinct_id"
+                    "#user_id","afrawip__meida_source","#account_id","#distinct_id","afrawip__campaign"
                 FROM
                     {gconf.tga_user_table}
             )
@@ -552,6 +559,7 @@ class TrainFeatureSql(BaseTga):
         on {base_user_table}.user_user_id={base_feature_table}.user_user_id
     )
     -- with结束没有逗号, 过滤左连接没有特征值的行，调试的时候取消where条件
+    -- 出现null过滤掉就行，少量的不管，因为tt_stable_rand的构建方法会有较小影响
     select * from user_tzz where {self.get_first_col_name()} is not NULL
     """
         return sql

{tfduck_bsd-0.16.4.dist-info → tfduck_bsd-0.17.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tfduck-bsd
-Version: 0.16.4
+Version: 0.17.0
 Summary: A small example package
 Home-page: UNKNOWN
 Author: yuanxiao

{tfduck_bsd-0.16.4.dist-info → tfduck_bsd-0.17.0.dist-info}/RECORD

@@ -1,4 +1,4 @@
-tfduck/__init__.py,sha256=nHVzJ8KVjyvWjqN4txczg0vMUrfR6_FZdqAPhUCmrYQ,20
+tfduck/__init__.py,sha256=UOdb1TyDAAWyMDClpACUUPZRxaht77qoX9AuhbBhsAs,20
 tfduck/main.py,sha256=zNTC16wkwGJ0QX1-i8vzlGophOxmFuO4SLsF1tkjsbE,14670
 tfduck/bdp_sdk_py/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 tfduck/bdp_sdk_py/example.py,sha256=Xq1_gcSyu0zho_iMmMYfYgMblcgF8a-GwRBWPTw0FuU,2879
@@ -21,20 +21,20 @@ tfduck/s3/s3oper.py,sha256=U1kYqNlVUX6SqTLnVRGqYbP0uRgWrW16N8pruQsxhW0,17761
 tfduck/sagemaker/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 tfduck/sagemaker/saoper.py,sha256=oVdxoUeXrM4sGXrTg8F-ZXG0lF6VXfjZ9gW3Q7ubjkU,11991
 tfduck/tga/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-tfduck/tga/base_tga.py,sha256=8G2DI86vr1zSa5NVFHLde2Ln42GmAgEns6dUNFmUAGY,2119
-tfduck/tga/predict_sql_ltv.py,sha256=P738DD6n70Yz8PaVR0gyHyXOxXN-COCfoP56wvAOLwE,3372
-tfduck/tga/predict_sql_retain.py,sha256=dZ-H5bBVmeYxv3ebzsohhU5sdEsj2n--8XgoJyAzujU,18559
-tfduck/tga/predict_sql_yh.py,sha256=MzBkt5BQx8NBK_zsNJvtiTQIrY0gTlw82BOFmuJn6bk,3371
+tfduck/tga/base_tga.py,sha256=rg1BHaIKGSwUVbOlWSA3Y1arB8J_caemjTjH9jh8GYM,2184
+tfduck/tga/predict_sql_ltv.py,sha256=25rpOZdHyMcEU3O8u67oUpLsTiZburEPsvXR38hTUJ0,3589
+tfduck/tga/predict_sql_retain.py,sha256=Nsl0lSZ_CC8j_GB4jLrJgkoqqDRAAPYwzo7vAOYZ764,19772
+tfduck/tga/predict_sql_yh.py,sha256=uYeuCZX2btxO-pvjrhlmuG35PquJbLvVtQEuEZHu8Cs,3588
 tfduck/tga/tga.py,sha256=bAW_RFV0Xclgol-MdP9NOWHwNeGLaVSVLqE9Ji-f_Hw,12496
 tfduck/tga/tga_test.py,sha256=A3n2LdvgQWlkX6E54K6cnsIUeYrgHlzG3QYbA7ZKgHk,2750
-tfduck/tga/train_sql_ltv.py,sha256=CR0G4EFCLzstDlsxqbQLtd1KKZRhAQEjBokt9SqWn18,25266
-tfduck/tga/train_sql_retain.py,sha256=2y8Q-fiBtevpYYPkkHcjwbwE6V5cCtMH8w8d3hwh1os,23105
-tfduck/tga/train_sql_yh.py,sha256=GuD_B6FfbX5uiypELAckpBhcQVvWLAFRZmtnL6oyKrg,23115
+tfduck/tga/train_sql_ltv.py,sha256=iJwKzc_u5o_KZ5ZyFsHyQN0cUnPAYX96NXuwPCtK27A,26563
+tfduck/tga/train_sql_retain.py,sha256=q385ZRzg9cymWqd_7qi-c5vX87mzL-pKC5vYyfCxS98,24326
+tfduck/tga/train_sql_yh.py,sha256=IJNr7U__dOqsPNpF0Y47_EdW8D3ScTZ_YR8cde-VViM,24367
 tfduck/thinkdata/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 tfduck/thinkdata/query.py,sha256=DsfcxjZrc0ZFTwN2pI5fKdM1Bwr6ageoPcA2MP3r2bE,1314
-tfduck_bsd-0.16.4.data/scripts/tfduck,sha256=UsuoAs4peJW4I-e6Gn91gEToP_YyuUp-rUUg3ObKneY,192
-tfduck_bsd-0.16.4.dist-info/LICENSE,sha256=2bm9uFabQZ3Ykb_SaSU_uUbAj2-htc6WJQmS_65qD00,1073
-tfduck_bsd-0.16.4.dist-info/METADATA,sha256=vAHtPnrXwaowG1JH0JomgsXA7GrOkLmESQ8w8LCrNWM,971
-tfduck_bsd-0.16.4.dist-info/WHEEL,sha256=g4nMs7d-Xl9-xC9XovUrsDHGXt-FT0E17Yqo92DEfvY,92
-tfduck_bsd-0.16.4.dist-info/top_level.txt,sha256=503etRkoyeI1VYcAwe5KpD5Bamhx0R0y2ofkE8HpRDA,7
-tfduck_bsd-0.16.4.dist-info/RECORD,,
+tfduck_bsd-0.17.0.data/scripts/tfduck,sha256=UsuoAs4peJW4I-e6Gn91gEToP_YyuUp-rUUg3ObKneY,192
+tfduck_bsd-0.17.0.dist-info/LICENSE,sha256=2bm9uFabQZ3Ykb_SaSU_uUbAj2-htc6WJQmS_65qD00,1073
+tfduck_bsd-0.17.0.dist-info/METADATA,sha256=P6YmdRS7t2nyfnzlei7Mmq9SNfBLHCfFA1Xpb794spU,971
+tfduck_bsd-0.17.0.dist-info/WHEEL,sha256=g4nMs7d-Xl9-xC9XovUrsDHGXt-FT0E17Yqo92DEfvY,92
+tfduck_bsd-0.17.0.dist-info/top_level.txt,sha256=503etRkoyeI1VYcAwe5KpD5Bamhx0R0y2ofkE8HpRDA,7
+tfduck_bsd-0.17.0.dist-info/RECORD,,