I have this data structure:
"_id" : "121212",
"terms" : [
{
"term" : "hi",
"tf" : 2
},
{
"term" : "you",
"tf" : 1
}
]
}
and making this query:
db.foo.aggregate( [
{
$match : { _id : "121212" }
},
{
$project:{ terms:1 }
},
{
$unwind: "$terms"
}
]).pretty();
I have come to get this result in my db:
{
"_id" : "121212",
"terms" : {
"term" : "hi",
"tf" : 2
}
}
{
"_id" : "121212",
"terms" : {
"term" : "you",
"tf" : 1
}
}
but is there any way to get a result like this?:
{
"_id" : "121212",
"term" : "hi",
"tf" : 2
}
{
"_id" : "121212",
"term" : "you",
"tf" : 1
}
I have tried to build the query with $ replaceRoot: {newRoot: "$ terms"}, but after I can't select the _id field anymore.
Well, you can use the $map and $mergeObjects to do this beautifully.
[
{ "$match":{"_id":"121212"}},
{
"$addFields":{
"terms":{
"$map":{
"input":"$terms",
"in":{
"$mergeObjects":[
"$$this",
{
"_id":"$_id"
}
]
}
}
}
}
}
]
If you really need to deconstruct the "terms" array, then add the $unwind: "$terms" to the pipeline.
You can achieve by using $project stage at the end of the pipeline
db.foo.aggregate([
{ "$match" : { "_id": "121212" } },
{ "$unwind": "$terms" },
{ "$project": { "term": "$terms.term", "tf": "$terms.tf" }}
])
Output
[
{
"_id": "121212",
"term": "hi",
"tf": 2
},
{
"_id": "121212",
"term": "you",
"tf": 1
}
]
Check it here
You need to use $mergeObjects inside $replaceRoot:
db.foo.aggregate( [
{
$match : { _id : "121212" }
},
{
$project:{ terms:1 }
},
{
$unwind: "$terms"
},
{
$replaceRoot: {
newRoot: {
$mergeObjects: [ { _id: "$_id" }, "$terms" ]
}
}
}
]).pretty();
Just to complete the range of options:
db.foo.aggregate([
{ "$match" : { "_id": "121212" } }, // filter by "_id"
{ "$addFields": { "terms._id": "$_id" } }, // copy "_id" field into terms
{ "$unwind": "$terms" }, // flatten the "terms" array
{ "$replaceRoot": { "newRoot": "$terms" } } // move the contents of the "terms" field up to the root level
])
Related
Given the following document data in collection called 'blah'...
[
{
"_id" : ObjectId("60913f55987438922d5f0db6"),
"procedureCode" : "code1",
"description" : "Description 1",
"coding" : [
{
"system" : "ABC",
"code" : "L111"
},
{
"system" : "DEFG",
"code" : "S222"
}
]
},
{
"_id" : ObjectId("60913f55987438922d5f0dbc"),
"procedureCode" : "code2",
"description" : "Description 2",
"coding" : [
{
"system" : "ABC",
"code" : "L999"
},
{
"system" : "DEFG",
"code" : "X3333"
}
]
}
]
What I want to get is all of the coding elements where system is ABC for all parents, and an array of codes like so.
[
{ "code": "L111" },
{ "code": "L999" },
]
If I use db.getCollection('blah').find({"coding.system": "ABC"}) I get the parent document with any child in the coding array of ICD.
If I use...
db.getCollection("blah")
.find({ "coding.system": "ABC" })
.projection({ "coding.code": 1 })
I do get the parent documents which have a child with a system of "ABC", but the coding for "DEFG" seems to come along for the ride too.
{
"_id" : ObjectId("60913f55987438922d5f0db6"),
"coding" : [
{
"code" : "L989"
},
{
"code" : "S102"
}
]
},
{
"_id" : ObjectId("60913f55987438922d5f0dbc"),
"coding" : [
{
"code" : "L989"
},
{
"code" : "X382"
}
]
}
I have also tried experimenting with:
db.getCollection("blah").aggregate(
{ $unwind: "$coding" },
{ $match: { "system": "ICD" } }
);
.. as per this page: mongoDB query to find the document in nested array
... but go no where fast with that approach. i.e. no records at all.
What query do I need, please, to achieve something like this..?
[
{ "code": "L111" },
{ "code": "L999" },
...
]
or even better, this..?
[
"L111",
"L999",
...
]
db.collection.aggregate([
{
$match: { "coding.system": "ABC" }
},
{
$unwind: "$coding"
},
{
$match: { "coding.system": "ABC" }
},
{
$project: { code: "$coding.code" }
}
])
mongoplayground
db.collection.aggregate([
{
$match: { "coding.system": "ABC" }
},
{
$unwind: "$coding"
},
{
$match: { "coding.system": "ABC" }
},
{
$group: {
_id: null,
coding: { $push: "$coding.code" }
}
}
])
mongoplayground
Instead of $unwind, $match you can also use $filter:
db.collection.aggregate([
{ $match: { "coding.system": "ABC" } },
{
$project: {
coding: {
$filter: {
input: "$coding",
cond: { $eq: [ "$$this.system", "ABC" ] }
}
}
}
}
])
I'm writing an aggregate query for the following records and output.
Data:
[
{
"_id" : ObjectId("5f3b2626927b18001db86884"),
"collections" : [
Art, Craft
]
},{
"_id" : ObjectId("5f3b2626927b18001db86885"),
"collections" : [
Craft
]
},{
"_id" : ObjectId("5f3b2626927b18001db86886"),
"collections" : [
Apex, Art
]
},
...
]
Expected Output:
count of collections id
{
Art : 2,
Craft : 2,
Apex : 1
}
Right now, we are looping through the collection to calculate count for each collections as the desired output, but it is low in performance because this collection is consists of 10,000 of records.
So, I was thinking to build an aggregate query and if someone can help me to start or point towards a right direction that would be really appreciated. Thank you.
$unwind
$group
$group
$replaceRoot
db.collection.aggregate([
{
$unwind: "$collections"
},
{
"$group": {
"_id": "$collections",
"v": {
"$sum": 1
}
}
},
{
"$group": {
"_id": null,
"collections": {
"$push": {
$arrayToObject: [
[ { "k": "$$ROOT._id", "v": "$$ROOT.v" } ]
]
}
}
}
},
{
$replaceRoot: {
newRoot: {
$mergeObjects: "$collections"
}
}
}
])
mongoplayground
I have figured a solution after checking for a while.
db.getCollection("collectionName").aggregate(
[
// get all the records with at least one collection name
{
$match: {
"collections.0": { $exists: true }
}
},
// populate the collection record
{
$lookup: {
from: "from_collection",
localField: "localField",
foreignField: "foreignField",
as: "collections"
}
},
// unwind
{ $unwind: "$collections" },
// group by the collections._id
{ $group: { _id: "$collections._id", collections: { $push: "$$ROOT.ID" } } },
// project with collection contains _id, and count
{
$project : {
collections: "$collections",
count: { $size: "$collections" }
}
}
]
).toArray();
output:
[
{
"_id" : ObjectId("61c4c42d68579f00311dd3e1"),
"collections" : [
"015151",
"015152",
"015153"
],
"count" : 3.0
},
{
"_id" : ObjectId("615f38016f40710033699939"),
"collections" : [
"014871"
],
"count" : 1.0
},
{
"_id" : ObjectId("611fed5ee0d12c00337cb009"),
"collections" : [
"014788",
"014786",
"014789",
"014787",
"014884",
"014893",
"014967",
"014968",
"015016",
"015017"
],
"count" : 10.0
}
...
]
I've tried many answers to similar problems using $lookup, $unwind, and $match, but I can't get this to work for my sub-sub-subdocument situation.
I have this collection, Things:
{
"_id" : ObjectId("5a7241f7912cfc256468cb27"),
"name" : "Fortress of Solitude",
"alias" : "fortress_of_solitude",
},
{
"_id" : ObjectId("5a7247ec548c9ad042f579e2"),
"name" : "Batcave",
"alias" : "batcave",
},
{
"_id" : ObjectId("6a7247bc548c9ad042f579e8"),
"name" : "Oz",
"alias" : "oz",
},
and this one-document collection, Venues:
{
"_id" : ObjectId("5b9acabbbf71f39223f8de6e"),
"name" : "The Office",
"floors" : [
{
"name" : "1st Floor",
"places" : [
{
"name" : "Front Entrance",
"alias" : "front_entrance"
}
]
},
{
"name" : "2nd Floor",
"places" : [
{
"name" : "Batcave",
"alias" : "batcave"
},
{
"name" : "Oz",
"alias" : "oz"
}
]
}
]
}
I want to return all the Things, but with the Venue's floors.places.name aggregated with each Thing if it exists if the aliases match between Things and Venues. So, I want to return:
{
"_id" : ObjectId("5a7241f7912cfc256468cb27"),
"name" : "Fortress of Solitude",
"alias" : "fortress_of_solitude",
<-- nothing added here because
<-- it's not found in Venues
},
{
"_id" : ObjectId("5a7247ec548c9ad042f579e2"),
"name" : "Batcave",
"alias" : "batcave",
"floors" : [ <-- this should be
{ <-- returned
"places" : [ <-- because
{ <-- the alias
name" : "Batcave" <-- matches
} <-- in Venues
] <--
} <--
] <--
},
{
"_id" : ObjectId("6a7247bc548c9ad042f579e8"),
"name" : "Oz",
"alias" : "oz",
"floors" : [ <-- this should be
{ <-- returned
"places" : [ <-- because
{ <-- the alias
name" : "Oz" <-- matches
} <-- in Venues
] <--
} <--
] <--
}
I've gotten as far as the following query, but it only returns the entire Venues.floors array as an aggregate onto each Thing, which is way too much extraneous data aggregated. I just want to merge each relevant floor.place sub-subsubdocument from Venues into its corresponding Thing if it exists in Venues.
db.getCollection('things').aggregate([
{$lookup: {from: "venues",localField: "alias",foreignField: "floors.places.alias",as: "matches"}},
{
$replaceRoot: { newRoot: { $mergeObjects: [ { $arrayElemAt: [ "$matches", 0 ] }, "$$ROOT" ] } }
},
{ $project: { matches: 0 } }
])
I'm struggling with existing answers, which seem to change at MongoDB version 3.2, 3.4, 3.6, or 4.2 to include or not include $unwind, $pipeline, and other terms. Can someone explain how to get a sub-sub-subdocument aggregated like this? Thanks!
You can try this :
db.things.aggregate([
{
$lookup:
{
from: "venues",
let: { alias: "$alias" },
pipeline: [
{ $unwind: { path: "$floors", preserveNullAndEmptyArrays: true } },
{ $match: { $expr: { $in: ['$$alias', '$floors.places.alias'] } } },
/** Below stages are only if you've docs like doc 2 in Venues */
{ $addFields: { 'floors.places': { $filter: { input: '$floors.places', cond: { $eq: ['$$this.alias', '$$alias'] } } } } },
{ $group: { _id: '$_id', name: { $first: '$name' }, floors: { $push: '$floors' } } },
{$project : {'floors.places.alias': 1, _id :0}} // Optional
],
as: "matches"
}
}
])
Test : MongoDB-Playground
Since MongoDB v3.6, we may perform uncorrelated sub-queries which gives us more flexibility to join two collections.
Try this:
db.things.aggregate([
{
$lookup: {
from: "venues",
let: {
"alias": "$alias"
},
pipeline: [
{
$unwind: "$floors"
},
{
$project: {
_id: 0,
places: {
$filter: {
input: "$floors.places",
cond: {
$eq: [
"$$alias",
"$$this.alias"
]
}
}
}
}
},
{
$match: {
"places.0": {
$exists: true
}
}
},
{
$unset: "places.name"
}
],
as: "floors"
}
}
])
MongoPlayground
Mongo query generated out of java code:
{
"pipeline": [{
"$match": {
"Id": "09cd9a5a-85c5-4948-808b-20a52d92381a"
}
},
{
"$group": {
"_id": "$result",
"id": {
"$first": "$result"
},
"labelKey": {
"$first": {
"$ifNull": ["$result",
"$result"]
}
},
"value": {
"$sum": 1
}
}
}]
}
Field 'result' can have values like Approved, Rejected, null and "" (empty string). What I am trying to achieve is combining the count of both null and empty together.
So that the empty string Id will have the count of both null and "", which is equal to 4
I'm sure theres a more "proper" way but this is what i could quickly come up with:
[
{
"$group" : {
"_id" : "$result",
"id" : {
"$first" : "$result"
},
"labelKey" : {
"$first" : {
"$ifNull" : [
"$result",
"$result"
]
}
},
"value" : {
"$sum" : 1.0
}
}
},
{
"$group" : {
"_id" : {
"$cond" : [{
$or: [
{"$eq": ["$_id", "Approved"]},
{"$eq": ["$_id", "Rejected"]},
]}},
"$_id",
""
]
},
"temp" : {
"$push" : {
"_id" : "$_id",
"labelKey" : "$labelKey"
}
},
"count" : {
"$sum" : "$value"
}
}
},
{
"$unwind" : "$temp"
},
{
"$project" : {
"_id" : "$temp._id",
"labelKey": "$temp.labelKey",
"count" : "$count"
}
}
],
);
Due to the fact the second group is only on 4 documents tops i don't feel too bad about doing this.
I have used $facet.
The MongoDB stage $facet lets you run several independent pipelines within the stage of a pipeline, all using the same data. This means that you can run several aggregations with the same preliminary stages, and successive stages.
var queries = [{
"$match": {
"Id": "09cd9a5a-85c5-4948-808b-20a52d92381a"
}
},{
$facet: {//
"empty": [
{
$match : {
result : { $in : ['',null]}
}
},{
"$group" : {
"_id" : null,
value : { $sum : 1}
}
}
],
"non_empty": [
{
$match : {
result : { $nin : ['',null]}
}
},{
"$group" : {
"_id" : '$result',
value : { $sum : 1}
}
}
]
}
},
{
$project: {
results: {
$concatArrays: [ "$empty", "$non_empty" ]
}
}
}];
Output :
{
"results": [{
"_id": null,
"value": 52 // count of both '' and null.
}, {
"_id": "Approved",
"value": 83
}, {
"_id": "Rejected",
"value": 3661
}]
}
Changing the group by like below solved the problem
{
"$group": {
"_id": {
"$ifNull": ["$result", ""]
},
"id": {
"$first": "$result"
},
"labelKey": {
"$first": {
"$ifNull": ["$result",
"$result"]
}
},
"value": {
"$sum": 1
}
}
}
Basically the structure is :
{
"_id" : ObjectId("123123"),
"stores" : [
{
"messages" : [
{
"updated_time" : "2018-05-15T05:12:25+0000",
"message_count" : 4,
"thread_id" : "123",
"messages" : [
{
"message" : "Hi User ",
"created_time" : "2018-05-15T05:12:25+0000",
"message_id" : "111",
},
{
"message" : "This is tes",
"created_time" : "2018-05-15T05:12:21+0000",
"message_id" : "222",
}
]
},
],
"store_id" : "123"
}
]
}
I have these values to get message_id object : 111. So how to get this object, any idea or help will be appreciated. THanks
store_id: 123,
thread_id:123,
message_id:111
The simplest way would be to $unwind all the nested arrays and then use $match to get single document. You can also add $replaceRoot to get only nested document. Try:
db.collection.aggregate([
{ $unwind: "$stores" },
{ $unwind: "$stores.messages" },
{ $unwind: "$stores.messages.messages" },
{ $match: { "stores.store_id": "123", "stores.messages.thread_id": "123", "stores.messages.messages.message_id": "111" } },
{ $replaceRoot: { newRoot: "$stores.messages.messages" } }
])
Prints:
{
"created_time": "2018-05-15T05:12:25+0000",
"message": "Hi User ",
"message_id": "111"
}
To improve the performance you can use $match after every $unwind to filter out unnecessary data as soon as possible, try:
db.collection.aggregate([
{ $unwind: "$stores" },
{ $match: { "stores.store_id": "123" } },
{ $unwind: "$stores.messages" },
{ $match: { "stores.messages.thread_id": "123" } },
{ $unwind: "$stores.messages.messages" },
{ $match: { "stores.messages.messages.message_id": "111" } },
{ $replaceRoot: { newRoot: "$stores.messages.messages" } }
])