How to select Specific attributes in an embedded mongo document - mongodb

Question

I have a mongo document similar to following structure:
{
id: '111eef8b94d3e91f4c7d22a37deb4aad',
description: 'Secret Project',
title: 'secret project',
students: [
{ _id: '123', name: 'Alex', primary_subject: 'Math', address: 'xxxxx', dob: '1989-10-10', gender: 'F', nationality: 'German' },
{ _id: '124', name: 'Emanuel', primary_subject: 'Physics', address: 'yyyyyy', dob: '1988-05-07', gender: 'M', nationality: 'French' },
{ _id: '242', name: 'Mike', primary_subject: 'Chemistry', address: 'zzzz', dob: '1990-02-02', gender: 'M', nationality: 'English' }
]
}
I need to fetch specific attributes. For example want to fetch only name, primary_subject, nationality attributes.
Using the below mongo query, I am able to fetch all attributes.
db.student_projects.aggregate({
$project: {
"students": {
$filter: {
input: "$students",
as: "st",
cond: {
$eq: [ "$$st._id", "242" ]
}
},
}
}
},
{ $unwind: { path: "$students", preserveNullAndEmptyArrays: false } }
).pretty();
Above query fetches all attributes of the matching student. But in my case I need just 3 attributes.

Answer 1

Use $map to reshape the output array:
db.student_projects.aggregate({
$project: {
"students": {
$map: {
input: {
$filter: {
input: "$students",
as: "st",
cond: {
$eq: [ "$$st._id", "242" ]
}
}
},
in: {
name: "$$this.name",
primary_subject: "$$this.primary_subject",
nationality: "$$this.nationality"
}
}
}
}
},
{ $unwind: { path: "$students", preserveNullAndEmptyArrays: false } }
).pretty();
Just like it's other language counterparts, "reshaphing" arrays is what $map does.
If you want to get "fancy" with a longer list of "included" fields than "excluded", then there are some modern operators from later releases of MongoDB 3.6 and above which can help here:
db.student_projects.aggregate({
$project: {
"students": {
$map: {
input: {
$filter: {
input: "$students",
as: "st",
cond: {
$eq: [ "$$st._id", "242" ]
}
}
},
in: {
$arrayToObject: {
$filter: {
input: { $objectToArray: "$$this" },
cond: {
"$not": {
"$in": [ "$$this.k", [ "_id", "address", "dob" ] ]
}
}
}
}
}
}
}
}
},
{ $unwind: "$students" }
).pretty();
The $objectToArray transforms to "key/value" pairs of k and v representing the object keys and values. From this "array" you can $filter the results on the k values for those you don't want. The $in allows comparison to a "list", and the $not negates the comparison value.
Finally you can transform the "array" back into the object form via $arrayToObject.
And of course you could always simply $project after the $unwind:
db.student_projects.aggregate({
$project: {
"students": {
$filter: {
input: "$students",
as: "st",
cond: {
$eq: [ "$$st._id", "242" ]
}
},
}
}
},
{ $unwind: "$students" },
{ $project: {
"students": {
"name": "$students.name",
"primary_subject": "$students.primary_subject",
"nationality": "$students.nationality"
}
}
).pretty();
And if you don't want the "students" key, then just remove it:
{ $project: {
"name": "$students.name",
"primary_subject": "$students.primary_subject",
"nationality": "$students.nationality"
}
Or use $replaceRoot from the original $map version:
db.student_projects.aggregate({
$project: {
"students": {
$map: {
input: {
$filter: {
input: "$students",
as: "st",
cond: {
$eq: [ "$$st._id", "242" ]
}
}
},
in: {
name: "$$this.name",
primary_subject: "$$this.primary_subject",
nationality: "$$this.nationality"
}
}
}
}
},
{ $unwind: "$students" },
{ $replaceRoot: { newRoot: "$students" } }
).pretty();
But for that matter you could have also done a $match after the $unwind instead of even using $filter. It's generally more efficient to work with arrays in place though, and a lot of the time you don't need the $unwind at all, so it's good practice to get used to the methods that manipulate arrays.
Of course if the result with $replaceRoot or similar was really what you were looking for, then it would be far more advisable to not use embedded documents in an array at all. If your intended access pattern uses those embedded documents "separately" to the main document most of the time, then you really should consider keeping them in their own collection instead. This avoids the "aggregation overhead" and is a simple query and projection to return the data.
N.B The $unwind operator "defaults" to preserveNullAndEmptyArrays: false. So the original form does not need that specified, nor the path key. It's shorter to write this way unless you specifically intend to preserve those null and empty results.