MongoDB结合组聚合和strLenBytes

Question

我有一个 Mongo 集合，其中包含这样的文档：

{
    "_id" : ObjectId("5a9d0d44c3a1ce5f14c6940a"),
    "topic_id" : "5a7af30613b79405643e7da1",
    "value" : "VMware Virtual Platform",
    "timestamp" : "2018-03-05 09:26:25.136546",
    "insert_ts" : "2018-03-05 09:26:25.136682",
    "inserted_by" : 1
},
{
    "_id" : ObjectId("5a9d0d44c3a1ce5f14c69409"),
    "topic_id" : "5a7af30713b79479f82b4b84",
    "value" : "VMware, Inc.",
    "timestamp" : "2018-03-05 09:26:25.118931",
    "insert_ts" : "2018-03-05 09:26:25.119081",
    "inserted_by" : 1
},
{
    "_id" : ObjectId("5a9d0d44c3a1ce5f14c69408"),
    "topic_id" : "5a7af30713b7946d6d0a8772",
    "value" : "Phoenix Technologies LTD 6.00 09/21/2015",
    "timestamp" : "2018-03-05 09:26:25.101624",
    "insert_ts" : "2018-03-05 09:26:25.101972",
    "inserted_by" : 1
}

我想从这个集合中获取一些聚合数据。我想知道最早的时间戳、文档数和所有值的总strlen，但按topic_id分组，其中document-id大于x。

在 mysql 中，我会像这样构建一个 sql：

SELECT 
    MAX(_id) as max_id,
    COUNT(*) as message_count,
    MIN(timestamp) as min_timestamp,
    LENGTH(GROUP_CONCAT(value)) as size
FROM `dev_topic_data_numeric` 
WHERE _id > 22000
GROUP BY topic_id

我如何在 MongoDB 中实现这一点？我已经尝试过构建它，如下所示：

db.getCollection('topic_data_text').aggregate(
[
    {
        "$match": 
        {
            "_id": {"$gte": ObjectId("5a9d0aefc3a1ce5f14c68c81") }
        }
    },
    {
        "$group": 
        {
            "_id": "$topic_id",
            "max_id": {"$max":"$_id"},
            "min_timestamp": {"$min": "$timestamp"},
            "message_count": {"$sum": 1},
            /*"size": {"$strLenBytes": "$value" }*/
        }
    } 
]
);

然后我取消注释 $strLenBytes 它崩溃说 strLenBytes 不是组运算符。 API of MongoDB 在这里对我没有帮助。怎么写才能得到strlen？

我的预期结果应该是这样的：

{
    "_id" : "5a7af30613b79405643e7da1",
    "max_id" : ObjectId("5a9d0d44c3a1ce5f14c6940a"),
    "min_timestamp" : "2018-03-05 09:26:25.136546",
    "message_count" : 1,
    "size" : 23,
}

我的 MongoDB 版本是 3.4.4。

Answer 1

这是因为$strLenBytes 不是累加器，不像$sum 或$max。 $group 阶段累加值，因此在 $group 阶段有效的任何运算符通常都是累加器。

$strLenBytes 以 1-1 的方式将一个值转换为另一个值。这通常是 $project 阶段的运算符。

在聚合中添加 $project 阶段应该会给您所需的结果。请注意，您还需要稍微修改 $group 阶段以传递所需的值：

> db.test.aggregate([
     {
         "$match":
         {
             "_id": {"$gte": ObjectId("5a9d0aefc3a1ce5f14c68c81") }
         }
     },
     {
         "$group":
         {
             "_id": {"topic_id": "$topic_id", value: "$value"},
             "max_id": {"$max":"$_id"},
             "min_timestamp": {"$min": "$timestamp"},
             "message_count": {"$sum": 1}
         }
     },
     {
         "$project":
         {
             "_id": "$_id.topic_id",
             "max_id": "$max_id",
             "min_timestamp": "$min_timestamp",
             "message_count": "$message_count",
             size: {"$strLenBytes": "$_id.value" }
         }
     }
 ])

使用您的示例文档输出：

{
  "_id": "5a7af30613b79405643e7da1",
  "max_id": ObjectId("5a9d0d44c3a1ce5f14c6940a"),
  "min_timestamp": "2018-03-05 09:26:25.136546",
  "message_count": 1,
  "size": 23
}
{
  "_id": "5a7af30713b79479f82b4b84",
  "max_id": ObjectId("5a9d0d44c3a1ce5f14c69409"),
  "min_timestamp": "2018-03-05 09:26:25.118931",
  "message_count": 1,
  "size": 12
}
{
  "_id": "5a7af30713b7946d6d0a8772",
  "max_id": ObjectId("5a9d0d44c3a1ce5f14c69408"),
  "min_timestamp": "2018-03-05 09:26:25.101624",
  "message_count": 1,
  "size": 40
}

Answer 2

在测试了@kevin-adistambha 的答案并进行了一些进一步的实验之后，我找到了另一种方法来实现我想要的结果——也许它有更好的性能——但这需要更多的测试来确定。

db.getCollection('topic_data_text').aggregate(
[
    {
        "$match": 
        {
            "_id":  {"$gt": ObjectId("5a9f9d8bd5de3ac75f8cc269") }
        }
    },
    {
        "$group": 
        {
            "_id": "$topic_id",
            "max_id": {"$max":"$_id"},
            "min_timestamp": {"$min": "$timestamp"},
            "message_count": {"$sum": 1},
            "size": {"$sum": {"$strLenBytes": "$value"}}
        }
    }
]
);