Datasets:
id: offline-compute_HiveSQL_hivesql_006
name: Input Table internal_platform_db.ods_t_databus_access_topic
category: offline-compute/HiveSQL
timeout_seconds: 600
modality: pure-text
engine: hivesql
Prompt
Task Objective: Cleanse the 数据总线 topic access cost raw data and aggregate it by topic dimension, excluding test topics, then write the results to a detail table.
Input: internal_platform_db.ods_t_databus_access_topic_cost_date_d_query_engine_015, partitioned by field dt (STRING, YYYYMMDD). Key fields: systemname, dwproductname, dwappgroup, cityid, iset, topic, data_size (BIGINT), total_cost (DOUBLE), in_charge.
Processing Rules: No joins, single-table processing. Filter condition: dt = '20260507' AND topic <> 'test'. Aggregation: Group by topic, applying MAX to systemname, dwproductname, dwappgroup, cityid, iset, data_size, total_cost, and in_charge. Derived column: dt directly takes the partition value '20260507'.
Output Requirements: The output field order is dt, system_belong, category_name, dw_appgroup, city_id, cluster_set, topic, total_data_size_d, total_cost, in_charge. Field mappings: dt=dt, system_belong=MAX(systemname), category_name=MAX(dwproductname), dw_appgroup=MAX(dwappgroup), city_id=MAX(cityid), cluster_set=MAX(iset), topic=topic, total_data_size_d=MAX(data_size), total_cost=MAX(total_cost), in_charge=MAX(in_charge). Data types: dt STRING, system_belong STRING, category_name STRING, dw_appgroup STRING, city_id STRING, cluster_set STRING, topic STRING, total_data_size_d BIGINT, total_cost DOUBLE, in_charge STRING.
Write Requirements: Target table internal_platform_db.dwd_databus_topic_cost_detail_d_cand_query_engine_015. Table comment: "数据总线 topic access cost detail table, including access cost and dimension information, performing data cleansing on the ODS layer." Partition field dt (STRING, YYYYMMDD). Write method: INSERT OVERWRITE, overwriting the partition dt='20260507'. If the target table does not exist, first create the table using standard Hive ORC storage format partitioned by dt, then write the data.
Please write the final HiveSQL to result.sql and execute it.