---
id: offline-compute_HiveSQL_hivesql_008
name: "Count tasks matching the shuffle_split tuning rule. Input table: internal_platform_db"
category: offline-compute/HiveSQL
timeout_seconds: 600
modality: pure-text
engine: hivesql
---
## Prompt
Task Objective: Count tasks that match the `shuffle_split` tuning rule, producing task-level tuning rule hit results.

Inputs:
- `internal_platform_db.nextgen_platform_dsl_spark_props_fht0_query_engine_021`
- `internal_platform_db.ods_spark_props_extra_query_engine_021`
- `internal_platform_db.deep_tuning_rule_hit_app_info_query_engine_021`
- `internal_platform_db.nextgen_platform_dsl_us_task_detail_fdt0_query_engine_021`
- `internal_platform_db.nextgen_platform_dsl_gputj_task_detail_fdt0_query_engine_021`
- `internal_platform_db.app_group_product_info_history_query_engine_021`

Processing Rules: Join the Spark configuration table, task detail tables (from both the `us` and `gputj` sources), and the application group product mapping table. Based on `deep_tuning_rule_hit_app_info`, filter records that match the `shuffle_split` tuning rule, and aggregate at the task granularity to produce deduplicated results. The join conditions and aggregation logic must be consistent with the historical logic, without introducing additional filtering or business rules.

Output Requirements: The output field order must match the target table schema, retaining task-level fields related to tuning rule hits, deduplicated at the task granularity.

Write Requirements: Write to table `internal_platform_db.shuffle_split_tuning_rule_hit_task_info_cand_query_engine_021`.

Please write the final HiveSQL to `result.sql` and execute it.