我应该如何构建我的 MongoDB 复合索引？

Posted 2023-02-16

【中文标题】我应该如何构建我的 MongoDB 复合索引？

【英文标题】：How should I structure my MongoDB compond index?

【发布时间】：2022-01-17 11:24:35

【问题描述】：

我有一个 mongo 图像元数据集合，包含以下字段：camera_name(str)、photographer_name(str)、resolution(str)、image_size(int in MB, rounded) 和 timestamp(10 digit UNIX timestamp)

我只想运行 2 个查询：

给定 camera_name，返回时间戳 给定 camera_name、photographer_name、resolution、image_size 和 timestamp，我想检索记录，按输入的时间戳降序排列。

我创建了 2 个索引：

"camera_name": 1, "timestamp": -1 "camera_name": 1, "photographer_name": 1, "resolution": 1, "image_size": 1, "timestamp": -1

第一个索引有效，但是当我对第二个索引运行查询时，没有返回任何记录。我确信集合中存在记录，并且我希望在运行第二个查询时至少获得 10 条记录，但它返回一个空列表。

索引的配置方式有问题吗？谢谢

这里是示例数据：

"camera_name": "Nikon", "photographer_name": "Aaron", "resolution": "1920x1080", "image_size": "3", "timestamp": 1397232415
"camera_name": "Nikon", "photographer_name": "Paul", "resolution": "1920x1080", "image_size": "4", "timestamp": 1717286853
"camera_name": "Nikon", "photographer_name": "Beth", "resolution": "720x480", "image_size": "1", "timestamp": 1503582086
"camera_name": "Nikon", "photographer_name": "Aaron", "resolution": "1920x1080", "image_size": "4", "timestamp": 1500628458
"camera_name": "Nikon", "photographer_name": "Paul", "resolution": "1920x1080", "image_size": "6", "timestamp": 1407580951
"camera_name": "Canon", "photographer_name": "Beth", "resolution": "1920x1080", "image_size": "5", "timestamp": 1166049453
"camera_name": "Canon", "photographer_name": "Paul", "resolution": "720x480", "image_size": "2", "timestamp": 1086317569
"camera_name": "Canon", "photographer_name": "Beth", "resolution": "720x480", "image_size": "1", "timestamp": 1400638926
"camera_name": "Canon", "photographer_name": "Aaron", "resolution": "720x480", "image_size": "1", "timestamp": 1345248762
"camera_name": "Canon", "photographer_name": "Paul", "resolution": "1920x1080", "image_size": "5", "timestamp": 1462360853
"camera_name": "Fuji", "photographer_name": "Beth", "resolution": "720x480", "image_size": "2", "timestamp": 1815298047
"camera_name": "Fuji", "photographer_name": "Shane", "resolution": "720x480", "image_size": "3", "timestamp": 1666493455
"camera_name": "Fuji", "photographer_name": "Beth", "resolution": "1920x1080", "image_size": "5", "timestamp": 1846677247
"camera_name": "Fuji", "photographer_name": "Beth", "resolution": "1920x1080", "image_size": "5", "timestamp": 1630996389
"camera_name": "Fuji", "photographer_name": "Shane", "resolution": "720x480", "image_size": "2", "timestamp": 1816829362

我执行的查询：

camera_name=Nikon 和时间戳 camera_name='Fuji' ,photographer_name='Beth', resolution='1920x1080', image_size='5' and timestamp

【问题讨论】：

索引不会阻止结果。索引提高了查询的速度。如果您没有得到任何结果，则要么数据不符合预期，要么查询不符合预期。此条件与索引规范无关。

由于您没有提供示例数据，也没有提供查询，我无法帮助您解决结果问题。

@barrypicker 我已经编辑了问题并添加了示例数据和查询及其预期输出

感谢您提供示例数据。这有点帮助。您尚未提供查询。您已经提供了查询的目的。有时实施是问题所在。我同意您的结果与您的意图不符。请提供实际查询。

【参考方案1】：

索引不会“过滤”结果，它们允许您通过扫描索引树而不是扫描原始文档来更快地访问数据。

这意味着如果第二个查询“不返回任何内容”，则它与您构建的任何索引都无关，但您使用的实际查询与数据库中的任何文档都不匹配。

我还要提到您的第二个索引可能会更小（取决于某些假设，例如规模和数据分布），这可以帮助更新/插入性能，同时额外减少存储大小。但是，从原始数据的外观来看，我认为这些并不是您的紧迫考虑。