[Bug] date_bin_gapfill 末尾窗口生成不稳定：受 device_id 过滤与上界运算符影响 #18988
Description
Search before asking
- I searched in the issues and found nothing similar.
Version
IoTDB 版本：[2.0.11]
部署方式：单机
操作入口：CLI / JDBC（表模型）
Describe the bug and provide the minimal reproduce step
问题描述
使用 date_bin_gapfill 配合 FILL METHOD PREVIOUS 时，查询范围末尾的时间窗口生成不稳定，存在两个异常：
是否加 device_id 过滤会改变结果：不加过滤时某设备有末尾窗口，加上该设备过滤后末尾窗口消失。
上界运算符（< / <=）影响末尾窗口：< 2025-01-01 时缺 2024-12 窗口；<= 2025-01-01 时不仅补出 2024-12，还多出 2025-01 窗口；<= 2024-12-31T23:59:59.999 时又缺失 2024-12 窗口。
复现步骤
- 建表与测试数据
CREATE TABLE "table1" ("time" TIMESTAMP TIME, "region" STRING TAG, "plant_id" STRING TAG, "device_id" STRING TAG, "model_id" STRING ATTRIBUTE, "maintenance" STRING ATTRIBUTE, "temperature" FLOAT FIELD, "humidity" FLOAT FIELD, "status" BOOLEAN FIELD, "arrival_time" TIMESTAMP FIELD) WITH (ttl = 'INF');
-- 插入测试数据
INSERT INTO table1(time, region, plant_id, device_id, model_id, maintenance, temperature, humidity, status, arrival_time) VALUES (2024-11-30T09:30:00.000+08:00, '上海', '3002', '101', 'F', '360', 35.2, 90, true, NULL), (2024-11-30T14:30:00.000+08:00, '上海', '3002', '101', 'F', '360', 34.8, 90, true, 2024-11-30T14:30:17.000+08:00), (2024-11-29T10:00:00.000+08:00, '上海', '3001', '101', 'D', '360', NULL, 85, NULL, 2024-11-29T10:00:13.000+08:00), (2024-11-27T16:38:00.000+08:00, '北京', '1001', '101', 'B', '180', 35.1, NULL, true, 2024-11-27T16:37:01.000+08:00), (2024-11-27T16:39:00.000+08:00, '北京', '1001', '101', 'B', '180', 35.3, 85, NULL, NULL), (2024-11-27T16:40:00.000+08:00, '北京', '1001', '101', 'B', '180', NULL, 85, NULL, 2024-11-27T16:37:03.000+08:00), (2024-11-27T16:41:00.000+08:00, '北京', '1001', '101', 'B', '180', NULL, 85, NULL, 2024-11-27T16:37:04.000+08:00), (2024-11-27T16:42:00.000+08:00, '北京', '1001', '101', 'B', '180', 35.2, NULL, false, NULL), (2024-11-27T16:43:00.000+08:00, '北京', '1001', '101', 'B', '180', NULL, NULL, false, NULL), (2024-11-27T16:44:00.000+08:00, '北京', '1001', '101', 'B', '180', NULL, NULL, false, 2024-11-27T16:37:08.000+08:00), (2024-11-29T11:00:00.000+08:00, '上海', '3002', '100', 'E', '180', 45.1, NULL, true, NULL), (2024-11-29T18:30:00.000+08:00, '上海', '3002', '100', 'E', '180', 35.4, 90, true, 2024-11-29T18:30:15.000+08:00), (2024-11-28T08:00:00.000+08:00, '上海', '3001', '100', 'C', '90', NULL, 85, NULL, 2024-11-28T08:00:09.000+08:00), (2024-11-28T09:00:00.000+08:00, '上海', '3001', '100', 'C', '90', 40.9, NULL, true, NULL), (2024-11-28T10:00:00.000+08:00, '上海', '3001', '100', 'C', '90', 35.2, 85, NULL, 2024-11-28T10:00:11.000+08:00), (2024-11-28T11:00:00.000+08:00, '上海', '3001', '100', 'C', '90', 45.1, 88, true, 2024-11-28T11:00:12.000+08:00), (2024-11-26T13:37:00.000+08:00, '北京', '1001', '100', 'A', '180', 35.1, 90, true, 2024-11-26T13:37:34.000+08:00), (2024-11-26T13:38:00.000+08:00, '北京', '1001', '100', 'A', '180', 35.1, 90, true, 2024-11-26T13:38:25.000+08:00);
- 查询用例与现象
用例 1：不加 device_id 过滤
SELECT date_bin_gapfill(1mo, time, 2001-01-01T00:00:00+08:00) AS window_time, device_id, AVG(temperature) AS avg_temp FROM table1 WHERE time >= 2024-01-01T00:00:00+08:00 AND time < 2025-01-01T00:00:00+08:00 GROUP BY 1, device_id fill method PREVIOUS FILL_GROUP 2;
现象： 设备 100 有 2024-12-01 窗口，设备 101 没有。
用例 2：加 device_id = '100'，上界用 <
-- SELECT 与 GROUP BY 同上
WHERE time >= 2024-01-01T00:00:00+08:00
AND time < 2025-01-01T00:00:00+08:00
AND device_id = '100'
现象： 2024-12-01 窗口消失，结果只到 2024-11-01。而不加过滤时该设备明明有这个窗口。
用例 3：加 device_id = '101'，上界改为 <= 2025-01-01
-- SELECT 与 GROUP BY 同上
WHERE time >= 2024-01-01T00:00:00+08:00
AND time <= 2025-01-01T00:00:00+08:00
AND device_id = '101'
现象： 结果不仅补出了 2024-12-01，还多出 2025-01-01 窗口。
用例 4：加 device_id = '101'，上界用 <= 2024-12-31T23:59:59.999
-- SELECT 与 GROUP BY 同上
WHERE time >= 2024-01-01T00:00:00+08:00
AND time <= 2024-12-31T23:59:59.999+08:00
AND device_id = '101'
现象： 2024-12-01 窗口再次消失。
What did you expect to see?
对于固定查询范围 [2024-01-01, 2025-01-01)、步长 1mo，date_bin_gapfill 应稳定生成 2024-01 到 2024-12 共 12 个窗口，不因以下因素变化：
WHERE 中是否加了具体 device_id 过滤；
上界写作 < 2025-01-01 还是 <= 2024-12-31T23:59:59.999。
末尾窗口不应整行丢失，也不应在边界处凭空多出下一窗口。
补充说明
FILL_GROUP 2 能正确按设备隔离填充，但窗口生成本身（gapfill）似乎在上界处处理不一致。
期望得到的结果
2024-01-01 00:00:00.000 101
2024-02-01 00:00:00.000 101
2024-03-01 00:00:00.000 101
2024-04-01 00:00:00.000 101
2024-05-01 00:00:00.000 101
2024-06-01 00:00:00.000 101
2024-07-01 00:00:00.000 101
2024-08-01 00:00:00.000 101
2024-09-01 00:00:00.000 101
2024-10-01 00:00:00.000 101
2024-11-01 00:00:00.000 101 35.119999694824216
2024-12-01 00:00:00.000 101 35.119999694824216
2024-01-01 00:00:00.000 100
2024-02-01 00:00:00.000 100
2024-03-01 00:00:00.000 100
2024-04-01 00:00:00.000 100
2024-05-01 00:00:00.000 100
2024-06-01 00:00:00.000 100
2024-07-01 00:00:00.000 100
2024-08-01 00:00:00.000 100
2024-09-01 00:00:00.000 100
2024-10-01 00:00:00.000 100
2024-11-01 00:00:00.000 100 38.84285681588309
2024-12-01 00:00:00.000 100 38.84285681588309
What did you see instead?
见：复现步骤
Anything else?
No response
Are you willing to submit a PR?
- I'm willing to submit a PR!
Activity