MongoDB's Aggregation Framework: project only matching element of an array

Question

I have a "class" document as:

{
className: "AAA",
students: [
   {name:"An", age:"13"},
   {name:"Hao", age:"13"},
   {name:"John", age:"14"},
   {name:"Hung", age:"12"}
   ]
}

And i want to get the student who has name is "An", get only matching element in array "students". I can do that with function find() as:

>db.class.find({"students.name":"An"}, {"students.$":true})
{
"_id" : ObjectId("548b01815a06570735b946c1"),
"students" : [ 
    {
        "name" : "An",
        "age" : "13"
    }
]}

It's fine, but when i do the same with Aggregation as following, it get error:

db.class.aggregate([
   {$match:{"students.name":'An'}},
   {$project:{"students.$":true}}
])

Error is:

uncaught exception: aggregate failed: {
    "errmsg" : "exception: FieldPath field names may not start with '$'.",
    "code" : 16410,
    "ok" : 0
}

Why? I can't use "$" for array in $project operator of aggregate() while can use this one in project operator of find().

BatScream · Accepted Answer · 2014-12-12 20:01:33Z

3

From the docs:

Use $ in the projection document of the find() method or the findOne() method when you only need one particular array element in selected documents.

The positional operator $ cannot be used in an aggregation pipeline projection stage. It is not recognized there.

This makes sense, because, when you execute a projection along with a find query, the input to the projection part of the query is a single document that has matched the query.The context of the match is known even during projection. So for each document that matches the query, the projection operator is applied then and there before the next match is found.

db.class.find({"students.name":"An"}, {"students.$":true})

In case of:

db.class.aggregate([
   {$match:{"students.name":'An'}},
   {$project:{"students.$":true}}
])

The aggregation pipeline is a set of stages. Each stage is completely unaware and independent of its previous or next stages. A set of documents pass a stage completely before being passed on to the next stage in the pipeline. The first stage in this case being the $match stage, all the documents are filtered based on the match condition. The input to the projection stage is now a set of documents that have been filtered as part of the match stage.

So a positional operator in the projection stage makes no sense, since in the current stage it doesn't know on what basis the fields had been filtered. Therefore, $ operators are not allowed as part of the field paths.

Why does the below work?

db.class.aggregate([
     { $match: { "students.name": "An" },
     { $unwind: "$students" },
     { $project: { "students": 1 } }
])

As you see, the projection stage gets a set of documents as input, and projects the required fields. It is independent of its previous and next stages.

answered Dec 12, 2014 at 20:01

BatScream

19.7k4 gold badges54 silver badges68 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Deka Over a year ago

Thank! You have helped me understanding why "we can't use "$" for array in $project operator of aggregate() while can use this one in project operator of find()". I will use this solution for my project, but i still find another way that don't unwind array. If there are so many elements in array, performance will be low. Anyway...thank you again!

BatScream Over a year ago

Yes, you are correct. If you know and expect only one matching sub document for your query, then you should blindly go with the find() query as in your question. But unfortunately, in aggregation, you simply just can't without unwinding.

atyagi · Accepted Answer · 2014-12-12 15:38:30Z

1

Try using the unwind operator in the pipeline: http://docs.mongodb.org/manual/reference/operator/aggregation/unwind/#pipe._S_unwind

Your aggregation would look like

db.class.aggregate([
     { $match: { "students.name": "An" },
     { $unwind: "$students" },
     { $project: { "students": 1 } }
])

edited Dec 12, 2014 at 15:38

answered Dec 12, 2014 at 15:28

atyagi

112 bronze badges

3 Comments

Deka Over a year ago

I have tried it before, but i think that this solution is not good. When we use $unwind, the document will be clone so many time (as number of element in array), may affect performance, i want project it rather than unwind it.

atyagi Over a year ago

That makes sense. What you could do then is switch the order of unwind and match in the pipeline so that you're filtering your documents first and then unwinding for easier projection. See the updated answer.

Deka Over a year ago

It would be acceptable solution, i just don't know why "I can't use "$" for array in $project operator of aggregate() while can use this one in project operator of find()"

Jitendra · Accepted Answer · 2021-11-23 10:55:05Z

1

You can use $filter to selects a subset of an array to return based on the specified condition.

db.class.aggregate([
   {
       $match:{
          "className": "AAA"
       }
   },
   {
       $project: {
          $filter: {
             input: "$students",
             as: "stu",
             cond: { $eq: [ "$$stu.name", "An" ] }
          }
   }
])

The following example filters the Students array to only include documents that have a name equal to "An".

edited Nov 23, 2021 at 10:55

answered Dec 3, 2018 at 9:51

Jitendra

3,1953 gold badges31 silver badges44 bronze badges

1 Comment

mosid Over a year ago

This should be the accepted answer IMHO, just a note @jitendra className key misses a closing quotation mark

Collectives™ on Stack Overflow

MongoDB's Aggregation Framework: project only matching element of an array

3 Answers 3

2 Comments

3 Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

3 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related