fix: add deploy bias on demo and docs (#3520)

demo/byzer-taxi/openmldb_byzer_taxi.bznb

@@ -64,7 +64,7 @@
     "job_id" : null
   }, {
     "id" : "240",
-    "content" : "run command as FeatureStoreExt.`` where\r\nzkAddress=\"127.0.0.1:2181\"\r\nand zkPath=\"/openmldb\"\r\nand `sql-0`='''\r\nSET @@execute_mode='online';\r\n'''\r\nand `sql-1`='''\r\nDEPLOY d1 SELECT trip_duration, passenger_count,\r\nsum(pickup_latitude) OVER w AS vendor_sum_pl,\r\nmax(pickup_latitude) OVER w AS vendor_max_pl,\r\nmin(pickup_latitude) OVER w AS vendor_min_pl,\r\navg(pickup_latitude) OVER w AS vendor_avg_pl,\r\nsum(pickup_latitude) OVER w2 AS pc_sum_pl,\r\nmax(pickup_latitude) OVER w2 AS pc_max_pl,\r\nmin(pickup_latitude) OVER w2 AS pc_min_pl,\r\navg(pickup_latitude) OVER w2 AS pc_avg_pl ,\r\ncount(vendor_id) OVER w2 AS pc_cnt,\r\ncount(vendor_id) OVER w AS vendor_cnt\r\nFROM t1 \r\nWINDOW w AS (PARTITION BY vendor_id ORDER BY pickup_datetime ROWS_RANGE BETWEEN 1d PRECEDING AND CURRENT ROW),\r\nw2 AS (PARTITION BY passenger_count ORDER BY pickup_datetime ROWS_RANGE BETWEEN 1d PRECEDING AND CURRENT ROW);\r\n'''\r\nand db=\"db1\"\r\nand action=\"ddl\";",
+    "content" : "run command as FeatureStoreExt.`` where\r\nzkAddress=\"127.0.0.1:2181\"\r\nand zkPath=\"/openmldb\"\r\nand `sql-0`='''\r\nSET @@execute_mode='online';\r\n'''\r\nand `sql-1`='''\r\nDEPLOY d1 OPTIONS(RANGE_BIAS="inf", ROWS_BIAS="inf") SELECT trip_duration, passenger_count,\r\nsum(pickup_latitude) OVER w AS vendor_sum_pl,\r\nmax(pickup_latitude) OVER w AS vendor_max_pl,\r\nmin(pickup_latitude) OVER w AS vendor_min_pl,\r\navg(pickup_latitude) OVER w AS vendor_avg_pl,\r\nsum(pickup_latitude) OVER w2 AS pc_sum_pl,\r\nmax(pickup_latitude) OVER w2 AS pc_max_pl,\r\nmin(pickup_latitude) OVER w2 AS pc_min_pl,\r\navg(pickup_latitude) OVER w2 AS pc_avg_pl ,\r\ncount(vendor_id) OVER w2 AS pc_cnt,\r\ncount(vendor_id) OVER w AS vendor_cnt\r\nFROM t1 \r\nWINDOW w AS (PARTITION BY vendor_id ORDER BY pickup_datetime ROWS_RANGE BETWEEN 1d PRECEDING AND CURRENT ROW),\r\nw2 AS (PARTITION BY passenger_count ORDER BY pickup_datetime ROWS_RANGE BETWEEN 1d PRECEDING AND CURRENT ROW);\r\n'''\r\nand db=\"db1\"\r\nand action=\"ddl\";",
     "job_id" : null
   }, {
     "id" : "241",

demo/java_quickstart/demo/src/main/java/com/openmldb/demo/App.java

@@ -146,7 +146,7 @@ private void createDeployment() {
                     "(PARTITION BY %s.c1 ORDER BY %s.c7 ROWS_RANGE BETWEEN 2d PRECEDING AND CURRENT ROW);", table,
                     table, table);
             // 上线一个Deployment
-            String deploySql = String.format("DEPLOY %s %s", deploymentName, selectSql);
+            String deploySql = String.format("DEPLOY %s OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') %s", deploymentName, selectSql);
             // set return null rs, don't check the returned value, it's false
             state.execute(deploySql);
         } catch (Exception e) {

demo/jd-recommendation/sql_scripts/deploy.sql

@@ -1,6 +1,6 @@
 USE JD_db;
 SET @@execute_mode='online';
-DEPLOY demo select * from
+DEPLOY demo OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') select * from
 (
 select
 `reqId` as reqId_1,

demo/predict-taxi-trip-duration/README.md

@@ -85,7 +85,7 @@ python3 train.py /tmp/feature_data /tmp/model.txt
 # The below commands are executed in the CLI
 > USE demo_db;
 > SET @@execute_mode='online';
-> DEPLOY demo SELECT trip_duration, passenger_count,
+> DEPLOY demo OPTIONS(RANGE_BIAS="inf", ROWS_BIAS="inf") SELECT trip_duration, passenger_count,
 sum(pickup_latitude) OVER w AS vendor_sum_pl,
 max(pickup_latitude) OVER w AS vendor_max_pl,
 min(pickup_latitude) OVER w AS vendor_min_pl,
@@ -193,7 +193,7 @@ python3 train.py /tmp/feature.csv /tmp/model.txt
 ```sql
 # The below commands are executed in the CLI
 > USE demo_db;
-> DEPLOY demo SELECT trip_duration, passenger_count,
+> DEPLOY demo OPTIONS(RANGE_BIAS="inf", ROWS_BIAS="inf") SELECT trip_duration, passenger_count,
 sum(pickup_latitude) OVER w AS vendor_sum_pl,
 max(pickup_latitude) OVER w AS vendor_max_pl,
 min(pickup_latitude) OVER w AS vendor_min_pl,

demo/predict-taxi-trip-duration/script/taxi.sql

@@ -22,7 +22,7 @@ w2 AS (PARTITION BY passenger_count ORDER BY pickup_datetime ROWS_RANGE BETWEEN
 OPTIONS(mode='overwrite');
 
 SET @@execute_mode='online';
-DEPLOY demo SELECT trip_duration, passenger_count,
+DEPLOY demo OPTIONS(RANGE_BIAS="inf", ROWS_BIAS="inf") SELECT trip_duration, passenger_count,
 sum(pickup_latitude) OVER w AS vendor_sum_pl,
 max(pickup_latitude) OVER w AS vendor_max_pl,
 min(pickup_latitude) OVER w AS vendor_min_pl,

demo/talkingdata-adtracking-fraud-detection/train_and_serve.ipynb

@@ -187,7 +187,7 @@
    "outputs": [],
    "source": [
     "deploy_name='d1'\n",
-    "%sql DEPLOY $deploy_name $sql_part;"
+    "%sql DEPLOY $deploy_name OPTIONS(RANGE_BIAS=\"inf\", ROWS_BIAS=\"inf\") $sql_part;"
    ]
   },
   {

demo/talkingdata-adtracking-fraud-detection/train_and_serve.py

@@ -166,7 +166,8 @@ def nothrow_execute(sql):
 connection.execute("SET @@execute_mode='online';")
 connection.execute(f'USE {DB_NAME}')
 nothrow_execute(f'DROP DEPLOYMENT {DEPLOY_NAME}')
-deploy_sql = f"""DEPLOY {DEPLOY_NAME} {sql_part}"""
+# to avoid data expired by abs ttl, set inf
+deploy_sql = f"""DEPLOY {DEPLOY_NAME} OPTIONS(RANGE_BIAS="inf", ROWS_BIAS="inf") {sql_part}"""
 print(deploy_sql)
 connection.execute(deploy_sql)
 print('Import data to online')

docs/en/use_case/lightgbm_demo.md

@@ -152,7 +152,7 @@ Assuming that the model produced by the features designed in Section 2.3 in the
 ```sql
 > USE demo_db;
 > SET @@execute_mode='online';
-> DEPLOY demo SELECT trip_duration, passenger_count,
+> DEPLOY demo OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') SELECT trip_duration, passenger_count,
 sum(pickup_latitude) OVER w AS vendor_sum_pl,
 max(pickup_latitude) OVER w AS vendor_max_pl,
 min(pickup_latitude) OVER w AS vendor_min_pl,

docs/zh/integration/deploy_integration/OpenMLDB_Byzer_taxi.md

@@ -232,7 +232,7 @@ and `sql-0`='''
 SET @@execute_mode='online';
 '''
 and `sql-1`='''
-DEPLOY d1 SELECT trip_duration, passenger_count,
+DEPLOY d1 OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') SELECT trip_duration, passenger_count,
 sum(pickup_latitude) OVER w AS vendor_sum_pl,
 max(pickup_latitude) OVER w AS vendor_max_pl,
 min(pickup_latitude) OVER w AS vendor_min_pl,

docs/zh/openmldb_sql/deployment_manage/DEPLOY_STATEMENT.md

@@ -175,16 +175,22 @@ deploy demo options(SYNC="false") SELECT t1.col1, t2.col2, sum(col4) OVER w1 as
     WINDOW w1 AS (PARTITION BY t1.col2 ORDER BY t1.col3 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW);
 ```
 
-#### 设置偏移
+#### 设置偏移BIAS
 
-如果你并不希望数据根据deploy的索引淘汰，或者希望晚一点淘汰，可以在deploy时设置偏移，常用于数据时间戳并不实时的情况、测试等情况。如果deploy后的索引ttl为abs 3h，但是数据的时间戳是3h前的(以系统时间为基准)，那么这条数据就会被淘汰，无法参与计算。设置一定时间或永久的偏移，则可以让数据更久的停留在在线表中。
+如果你并不希望数据根据deploy的索引淘汰，或者希望晚一点淘汰，可以在deploy时设置偏移BIAS，常用于数据时间戳并不实时的情况、测试等情况。如果deploy后的索引ttl为abs 3h，但是数据的时间戳是3h前的(以系统时间为基准)，那么这条数据就会被淘汰，无法参与计算。设置一定时间或永久的偏移，则可以让数据更久的停留在在线表中。
 
 时间偏移，单位可以是`s`、`m`、`h`、`d`，也可以是整数，单位为`ms`，也可以是`inf`，表示永不淘汰；如果是行数偏移，可以是整数，单位是`row`，也可以是`inf`，表示永不淘汰。两种偏移中，0均表示不偏移。
 
 注意，我们只将偏移加在deploy的解析索引中，也就是新索引，它们并不是最终索引。最终索引的计算方式是，如果是创建索引，最终索引是`解析索引 + 偏移`；如果是更新索引，最终索引是`merge(旧索引, 新索引 + 偏移)`。
 
 而时间偏移的单位是`min`，我们会在内部将其转换为`min`，并且取上界。比如，新索引ttl是abs 2min，加上偏移20s，结果是`2min + ub(20s) = 3min`，然后和旧索引1min取上界，最终索引ttl是`max(1min, 3min) = 3min`。
 
+**Example**
+```sql
+DEPLOY demo OPTIONS(RANGE_BIAS="inf", ROWS_BIAS="inf") SELECT t1.col1, t2.col2, sum(col4) OVER w1 as w1_col4_sum FROM t1 LAST JOIN t2 ORDER BY t2.col3 ON t1.col2 = t2.col2
+    WINDOW w1 AS (PARTITION BY t1.col2 ORDER BY t1.col3 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW);
+```
+
 ## 相关SQL
 
 [USE DATABASE](../ddl/USE_DATABASE_STATEMENT.md)

docs/zh/quickstart/sdk/java_sdk.md

@@ -403,7 +403,7 @@ try {
             "(PARTITION BY %s.c1 ORDER BY %s.c7 ROWS_RANGE BETWEEN 2d PRECEDING AND CURRENT ROW);", table,
             table, table);
     // 上线一个Deployment
-    String deploySql = String.format("DEPLOY %s %s", deploymentName, selectSql);
+    String deploySql = String.format("DEPLOY %s OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') %s", deploymentName, selectSql);
     // set return null rs, don't check the returned value, it's false
     state.execute(deploySql);
 } catch (Exception e) {

docs/zh/use_case/JD_recommendation.md

@@ -393,7 +393,7 @@ bash train_deepfm.sh $demodir/feature_preprocess/out
     ```sql
     -- OpenMLDB CLI
     USE JD_db;
-    DEPLOY demo <SQL>;
+    DEPLOY demo OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') <SQL>;
     ```
 
     也可以在 Docker 容器内直接运行部署脚本：

docs/zh/use_case/taxi_tour_duration_prediction.md

@@ -151,7 +151,7 @@ w2 AS (PARTITION BY passenger_count ORDER BY pickup_datetime ROWS_RANGE BETWEEN
     --OpenMLDB CLI
     USE demo_db;
     SET @@execute_mode='online';
-    DEPLOY demo SELECT trip_duration, passenger_count,
+    DEPLOY demo OPTIONS(RANGE_BIAS='inf', ROWS_BIAS='inf') SELECT trip_duration, passenger_count,
     sum(pickup_latitude) OVER w AS vendor_sum_pl,
     max(pickup_latitude) OVER w AS vendor_max_pl,
     min(pickup_latitude) OVER w AS vendor_min_pl,
@@ -167,6 +167,10 @@ w2 AS (PARTITION BY passenger_count ORDER BY pickup_datetime ROWS_RANGE BETWEEN
     w2 AS (PARTITION BY passenger_count ORDER BY pickup_datetime ROWS_RANGE BETWEEN 1d PRECEDING AND CURRENT ROW);
     ```
 
+```{note}
+此处DEPLOY包含BIAS OPTIONS，是因为导入在线存储的数据文件不会更新，对于当前时间来讲，可能会超过DEPLOY后的表索引的时间TTL，导致表淘汰掉这些数据。时间淘汰，只看每个索引的ts列和ttl，只要数据中该列的值<(当前时间-abs_ttl)，在该索引上就会被淘汰，与其他因素无关，各个索引也互相不影响。如果你的数据不是实时产生的新timestamp，也需要考虑带上BIAS OPTIONS。
+```
+
 ### 步骤 7：导入在线数据
 
 首先，请切换到**在线**执行模式。接着在在线模式下，导入样例数据 `/work/taxi-trip/data/taxi_tour_table_train_simple.csv` 作为在线数据，用于在线特征计算。