Datasets:

dicemy
/

DataClawEval

	#!/usr/bin/env python3
	# -- coding: utf-8 --
	"""
	mysql_003 ground truth: 离线推理任务特征宽表构建

	Task:
	Build an offline inference task feature wide table by joining 7 input tables,
	aggregating pipeline timing and wait time stats, extracting model info via JSON,
	and writing to the output table.

	WHERE conditions match the original Hive SQL:
	- a.dt = '2026050700'
	- a.base = 'JOB' AND a.model_type = 'STATIC_MODEL'
	- a.create_time < '2026-05-07 01:00:00' (dt parsed + 1 hour)
	- status NOT IN ('FINISH','KILL','FAILED') OR update_time >= '2026-05-06 23:00:00' (dt parsed - 1 hour)
	"""
	import pymysql
	import sys

	DB_NAME = "internal_platform_db"
	OUTPUT_TABLE = "dwd_aide_offline_inference_feature_cand_mysql_003"

	MYSQL_CONFIG = {
	"host": "localhost",
	"port": 3306,
	"user": "root",
	"password": "root123",
	"charset": "utf8mb4",
	}

	gt_sql = f"""
	INSERT INTO {DB_NAME}.{OUTPUT_TABLE}
	(dt, create_time, ct, id, name, servingName, base, status, submit_type,
	mould_id, mould_name, time_cost_in_second, dataset_act_num, dataset_fin_num,
	dataset_err_num, max_instance_first_wait_time, avg_instance_first_wait_time,
	task_dispatch_time, offline_inference_time, inference_config,
	submit_operator, submit_time, image_tag)

	WITH
	-- Step 1: Aggregate pipeline timing by task_id
	pipeline_time_agg AS (
	SELECT
	task_id,
	SUM(time_cost_in_second) AS time_cost_in_second,
	CAST(SUM(CASE WHEN step_desc = '任务下发' THEN time_cost_in_second ELSE 0 END) AS SIGNED) AS task_dispatch_time,
	CAST(SUM(CASE WHEN step_desc = '离线推理' THEN time_cost_in_second ELSE 0 END) AS SIGNED) AS offline_inference_time
	FROM {DB_NAME}.aide_offline_inference_pipeline_time_mysql_003
	WHERE end_time IS NOT NULL
	GROUP BY task_id
	),

	-- Step 2: Aggregate wait time stats by service_name (latest partition)
	service_wait_time_stats AS (
	SELECT
	service_name,
	CAST(MAX(first_wait_time) AS DOUBLE) AS max_instance_first_wait_time,
	CAST(AVG(first_wait_time) AS DOUBLE) AS avg_instance_first_wait_time
	FROM {DB_NAME}.task_instance_wait_time_stats_mysql_003
	WHERE dt = (SELECT MAX(dt) FROM {DB_NAME}.task_instance_wait_time_stats_mysql_003)
	AND first_wait_time IS NOT NULL
	GROUP BY service_name
	)

	-- Step 3: Join all tables and extract features
	SELECT
	'2026050700' AS dt,
	a.create_time,
	1 AS ct,
	a.id,
	a.name AS `name`,
	CONCAT(a.name, '_', c.id, '_OFFLINE') AS servingName,
	a.base,
	a.status,
	a.submit_type,
	JSON_UNQUOTE(JSON_EXTRACT(a.model_ids, '$.mould_id')) AS mould_id,
	JSON_UNQUOTE(JSON_EXTRACT(a.model_ids, '$.mould_name')) AS mould_name,
	d.time_cost_in_second,
	b.dataset_act_num,
	b.dataset_fin_num,
	b.dataset_err_num,
	w.max_instance_first_wait_time,
	w.avg_instance_first_wait_time,
	d.task_dispatch_time,
	d.offline_inference_time,
	a.inference_config,
	e.operator AS submit_operator,
	e.create_time AS submit_time,
	f.image_tag
	FROM {DB_NAME}.aide_offline_inference_info_fixed_mysql_003 a
	LEFT JOIN {DB_NAME}.aide_offline_inference_dataset_info_h_mysql_003 b
	ON a.id = b.offline_inference_id
	LEFT JOIN {DB_NAME}.aide_offline_inference_stage_info_h_mysql_003 c
	ON a.id = c.offline_inference_id
	LEFT JOIN pipeline_time_agg d
	ON a.id = d.task_id
	LEFT JOIN service_wait_time_stats w
	ON CONCAT(a.name, '_', c.id, '_OFFLINE') = w.service_name
	LEFT JOIN (
	SELECT
	id,
	operator,
	create_time,
	ROW_NUMBER() OVER(PARTITION BY id ORDER BY update_time DESC) AS rn
	FROM {DB_NAME}.aide_offline_inference_file_info_h_mysql_003
	WHERE dt = '2026050700'
	) e ON a.file_id = e.id AND e.rn = 1
	LEFT JOIN {DB_NAME}.aide_mould_h_mysql_003 f
	ON CAST(JSON_UNQUOTE(JSON_EXTRACT(a.model_ids, '$.mould_id')) AS SIGNED) = f.id
	WHERE a.dt = '2026050700'
	AND a.base = 'JOB'
	AND a.model_type = 'STATIC_MODEL'
	AND a.create_time < '2026-05-07 01:00:00'
	AND (
	a.status NOT IN ('FINISH', 'KILL', 'FAILED')
	OR
	a.update_time >= '2026-05-06 23:00:00'
	)
	"""


	def main():
	conn = pymysql.connect(**MYSQL_CONFIG)

	try:
	with conn.cursor() as cur:
	# Ensure output table exists
	cur.execute(f"""
	CREATE TABLE IF NOT EXISTS {DB_NAME}.{OUTPUT_TABLE} (
	dt VARCHAR(256) COMMENT '分区字段',
	create_time VARCHAR(256) COMMENT '创建时间',
	ct INT COMMENT '计数标识',
	id BIGINT COMMENT '离线推理任务ID',
	name VARCHAR(256) COMMENT '任务名称',
	servingName VARCHAR(256) COMMENT '服务名称',
	base VARCHAR(256) COMMENT '基础类型',
	status VARCHAR(256) COMMENT '状态',
	submit_type VARCHAR(256) COMMENT '提交类型',
	mould_id VARCHAR(256) COMMENT '模型ID',
	mould_name VARCHAR(256) COMMENT '模型名称',
	time_cost_in_second BIGINT COMMENT '总耗时(秒)',
	dataset_act_num BIGINT COMMENT '数据集激活数量',
	dataset_fin_num BIGINT COMMENT '数据集完成数量',
	dataset_err_num BIGINT COMMENT '数据集错误数量',
	max_instance_first_wait_time DOUBLE COMMENT '最大首次等待时长(秒)',
	avg_instance_first_wait_time DOUBLE COMMENT '平均首次等待时长(秒)',
	task_dispatch_time BIGINT COMMENT '任务下发耗时(秒)',
	offline_inference_time BIGINT COMMENT '离线推理耗时(秒)',
	inference_config VARCHAR(256) COMMENT '推理配置',
	submit_operator VARCHAR(256) COMMENT '提交人',
	submit_time VARCHAR(256) COMMENT '提交时间',
	image_tag VARCHAR(256) COMMENT '镜像标签'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4
	""")

	# Truncate + insert
	cur.execute(f"TRUNCATE TABLE {DB_NAME}.{OUTPUT_TABLE}")
	cur.execute(gt_sql)

	conn.commit()
	print("mysql_003 ground_truth done: rows written to output table")
	except Exception as e:
	print(f"ground_truth error: {e}", file=sys.stderr)
	conn.rollback()
	sys.exit(1)
	finally:
	conn.close()


	if __name__ == "__main__":
	main()