$ sort后的MongoDB值$ limit会改变聚合结果的顺序

Question

我在将简单的.find().sort().skip().limit()迁移到聚合框架时遇到了问题，导致分页损坏：同一文档出现在多个页面上。最后，我意识到$limit会影响$sort结果的顺序。下面我将介绍如何重现查看问题的步骤（在Mongo 3.4.17和3.6.7上查看）。

> // 1. Create a simple coleection and fill it with test data: 
> for(var i = 0; i < 10; ++i) db.SomeCollection.insert({sortField: "just_some_string"});
> db.SomeCollection.count();
10
>
> // 2. Check, that with .find().sort().limit() the order of results does not depend on the `limit()`'s value:
db.SomeCollection.find().sort({"sortField": 1}).limit(5)
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb531"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb532"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb533"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb534"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb535"), "sortField" : "just_some_string" }
> db.SomeCollection.find().sort({"sortField": 1}).limit(8)
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb531"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb532"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb533"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb534"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb535"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb536"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb537"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb538"), "sortField" : "just_some_string" }
>
> // 3. Logically the same query but using Aggregation Framework
> db.SomeCollection.aggregate([{$sort: {"sortField": 1}}, {$limit: 5}]);
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb534"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb535"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb532"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb531"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb533"), "sortField" : "just_some_string" }
> db.SomeCollection.aggregate([{$sort: {"sortField": 1}}, {$limit: 8}]);
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb538"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb535"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb532"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb534"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb536"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb531"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb533"), "sortField" : "just_some_string" }
{ "_id" : ObjectId("5ba3ddf3a07260b77dddb537"), "sortField" : "just_some_string" }

我的问题是：

1. 为什么带有$ sort后跟$ limit的.aggregate()给出的结果顺序不同于.find().sort().limit()？
2. 为什么.aggregate()的结果顺序取决于$limit的值？

Answer 1

所有这些文档都具有相同的sortField值，因此顺序不是确定性的。在_id上添加辅助排序以确保一致的结果。

db.SomeCollection.aggregate([{$sort: {"sortField": 1, _id: 1}}, {$limit: 5}]);