Develop Reference

How to paginate terms aggregation results in Elasticsearch

I've been trying to figure out a way to paginate the results of a terms aggregation in Elasticsearch and so far I have not been able to achieve the desired result.
Here's the problem I am trying to solve. In my index, I have a bunch of documents that have a score (separate to the ES _score) that is calculated based on the values of the other fields in the document. Each document "belongs" to a customer, referenced by the customer_id field. The document also has an id, referenced by the doc_id field, and is the same as the ES meta-field _id. Here is an example.
{
'_id': '1',
'doc_id': '1',
'doc_score': '85',
'customer_id': '123'
}
For each customer_id there are multiple documents, all with different document ids and different scores. What I want to be able to do is, given a list of customer ids, return the top document for each customer_id (only 1 per customer) and be able to paginate those results similar to the size, from method in the regular ES search API. The field that I want to use for the document score is the doc_score field.
So far in my current Python script, I've tried is a nested aggs with a "top hits" aggregation to only get the top document for each customer.
{
"size": 0,
"query:": {
"bool": {
"must": [
{
"match_all": {}
},
{
"terms": {
"customer_id": customer_ids # a list of the customer ids I want documents for
}
},
{
"exists": {
"field": "score" # sometimes it's possible a document does not have a score
}
}
]
}
}
"aggs": {
"customers": {
"terms" : {
{"field": "customer_id", "min_doc_count": 1},
"aggs": {
"top_documents": {
"top_hits": {
"sort": [
{"score": {"order": "desc"}}
],
"size": 1
}
}
}
}
}
}
}
I then "paginate" by going through each customer bucket, appending the top document blob to a list and then sorting the list based on the value of the score field and finally taking a slice documents_list[from:from+size].
The issue with this is that, say I have 500 customers in the list but I only want the 2nd 20 documents, i.e. size = 20, from=20. So each time I call the function I have to first get the list for each of the 500 customers and then slice. This sounds very inefficient and is also a speed issue, since I need that function to be as fast as I can possibly make it.
Ideally, I could just get the 2nd 20 directly from ES without having to do any slicing in my function.
I have looked into Composite aggregations that ES offers, but it looks to me like I would not be able to use it in my case, since I need to get the entire doc, i.e. everything in the _source field in the regular search API response.
I would greatly appreciate any suggestions.

The best way to do this would be to use partitions
According to documentation:
GET /_search
{
"size": 0,
"aggs": {
"expired_sessions": {
"terms": {
"field": "account_id",
"include": {
"partition": 1,
"num_partitions": 25
},
"size": 20,
"order": {
"last_access": "asc"
}
},
"aggs": {
"last_access": {
"max": {
"field": "access_date"
}
}
}
}
}
}
https://www.elastic.co/guide/en/elasticsearch/reference/6.8/search-aggregations-bucket-terms-aggregation.html#_filtering_values_with_partitions

Sort Elastic search result with best _score in Aggregation

I am trying to do an aggregation and then sort the best scored result in top. Even though aggregation is successful, yet its not score sorted. How to make it possible?
"aggs": {
"group": {
"terms": {
"field": "description.keyword",
"order": {
"description_score": "desc"
}
},
"aggs": {
"group_docs": {
"top_hits": {
"size": 1
}
}
}
},
}

As #Lupanoide mentioned scores are not calculated in aggregation so you cannot sort your bucket on document score. Instead field collapsing can be used
Allows to collapse search results based on field values. The
collapsing is done by selecting only the top sorted document per
collapse key
{
"query": {
"match_all": {}
},
"collapse": { --> group by on given field
"field": "sources.keyword"
}
}
By default search results are sorted on "score" so you don't need to add "sort"

Getting linked documents in single lookup query in Elastic Search

To provide some context :
I want to write a bulk update query(possibly affecting 0.5 - 1M docs). The update would be in the aspects field (shown below) which are mostly duplicated.
My thinking was if I normalised it into another entity (aspect_label), the amount of docs updated would be reduced drastically (say 500-1000 max).
Query : I want to find out if there is a way to get linked documents via id in Elastic Search.
Eg. if I have documents in index my_db according to the mapping below.
Just to point out : processed_reviews is a child of aspect_label
{
"my_db":{
"mappings":{
"processed_reviews":{
"_all":{
"enabled":false
},
"_parent":{
"type":"aspect_label"
},
"_routing":{
"required":true
},
"properties":{
"data":{
"properties":{
"insights":{
"type":"nested",
"properties":{
"aspects":{
"type":"nested",
"properties":{
"aspect_label_id":{
"type":"keyword"
},
"aspect_term_frequency":{
"type":"long"
}
}
}
}
},
"preprocessed_text":{
"type":"text"
},
"preprocessed_title":{
"type":"text"
}
}
}
}
}
}
}
}
And another entity aspect_label :
{
"my_db": {
"mappings": {
"aspect_label": {
"_all": {
"enabled": false
},
"properties": {
"aspect": {
"type": "keyword"
},
"aspect_label_new": {
"type": "keyword"
},
"aspect_label_old": {
"type": "text"
}
}
}
}
}
}
Now, I want to write a search query on the processed_reviews type such that the aspect_label_id entity is replaced with the the value of aspect_label_new in the doc or the entire doc in aspect_label matching the id.
{
"_index":"my_db",
"_type":"processed_reviews",
"_id":"191b3bff-4915-4404-a05a-10e6bd2b19d4",
"_score":1,
"_routing":"5",
"_parent":"5",
"_source":{
"data":{
"preprocessed_text":"Good product I really like so comfortable and so light wait and looks good",
"preprocessed_title":"Good choice",
"insights":[
{
"aspects":[
{
"aspect_label":"color",
"aspect_term_frequency":1
}
]
}
]
}
}
}
Also, if there is a better way to approach this problem/ something wrong with my approach or if this is possible or not. Please inform me of the same as well.

Get elements from array between two dates [duplicate]

Suppose you have the following documents in my collection:
{
"_id":ObjectId("562e7c594c12942f08fe4192"),
"shapes":[
{
"shape":"square",
"color":"blue"
},
{
"shape":"circle",
"color":"red"
}
]
},
{
"_id":ObjectId("562e7c594c12942f08fe4193"),
"shapes":[
{
"shape":"square",
"color":"black"
},
{
"shape":"circle",
"color":"green"
}
]
}
Do query:
db.test.find({"shapes.color": "red"}, {"shapes.color": 1})
Or
db.test.find({shapes: {"$elemMatch": {color: "red"}}}, {"shapes.color": 1})
Returns matched document (Document 1), but always with ALL array items in shapes:
{ "shapes":
[
{"shape": "square", "color": "blue"},
{"shape": "circle", "color": "red"}
]
}
However, I'd like to get the document (Document 1) only with the array that contains color=red:
{ "shapes":
[
{"shape": "circle", "color": "red"}
]
}
How can I do this?

MongoDB 2.2's new $elemMatch projection operator provides another way to alter the returned document to contain only the first matched shapes element:
db.test.find(
{"shapes.color": "red"},
{_id: 0, shapes: {$elemMatch: {color: "red"}}});
Returns:
{"shapes" : [{"shape": "circle", "color": "red"}]}
In 2.2 you can also do this using the $ projection operator, where the $ in a projection object field name represents the index of the field's first matching array element from the query. The following returns the same results as above:
db.test.find({"shapes.color": "red"}, {_id: 0, 'shapes.$': 1});
MongoDB 3.2 Update
Starting with the 3.2 release, you can use the new $filter aggregation operator to filter an array during projection, which has the benefit of including all matches, instead of just the first one.
db.test.aggregate([
// Get just the docs that contain a shapes element where color is 'red'
{$match: {'shapes.color': 'red'}},
{$project: {
shapes: {$filter: {
input: '$shapes',
as: 'shape',
cond: {$eq: ['$$shape.color', 'red']}
}},
_id: 0
}}
])
Results:
[
{
"shapes" : [
{
"shape" : "circle",
"color" : "red"
}
]
}
]

The new Aggregation Framework in MongoDB 2.2+ provides an alternative to Map/Reduce. The $unwind operator can be used to separate your shapes array into a stream of documents that can be matched:
db.test.aggregate(
// Start with a $match pipeline which can take advantage of an index and limit documents processed
{ $match : {
"shapes.color": "red"
}},
{ $unwind : "$shapes" },
{ $match : {
"shapes.color": "red"
}}
)
Results in:
{
"result" : [
{
"_id" : ObjectId("504425059b7c9fa7ec92beec"),
"shapes" : {
"shape" : "circle",
"color" : "red"
}
}
],
"ok" : 1
}

Caution: This answer provides a solution that was relevant at that time, before the new features of MongoDB 2.2 and up were introduced. See the other answers if you are using a more recent version of MongoDB.
The field selector parameter is limited to complete properties. It cannot be used to select part of an array, only the entire array. I tried using the $ positional operator, but that didn't work.
The easiest way is to just filter the shapes in the client.
If you really need the correct output directly from MongoDB, you can use a map-reduce to filter the shapes.
function map() {
filteredShapes = [];
this.shapes.forEach(function (s) {
if (s.color === "red") {
filteredShapes.push(s);
}
});
emit(this._id, { shapes: filteredShapes });
}
function reduce(key, values) {
return values[0];
}
res = db.test.mapReduce(map, reduce, { query: { "shapes.color": "red" } })
db[res.result].find()

Another interesing way is to use $redact, which is one of the new aggregation features of MongoDB 2.6. If you are using 2.6, you don't need an $unwind which might cause you performance problems if you have large arrays.
db.test.aggregate([
{ $match: {
shapes: { $elemMatch: {color: "red"} }
}},
{ $redact : {
$cond: {
if: { $or : [{ $eq: ["$color","red"] }, { $not : "$color" }]},
then: "$$DESCEND",
else: "$$PRUNE"
}
}}]);
$redact "restricts the contents of the documents based on information stored in the documents themselves". So it will run only inside of the document. It basically scans your document top to the bottom, and checks if it matches with your if condition which is in $cond, if there is match it will either keep the content($$DESCEND) or remove($$PRUNE).
In the example above, first $match returns the whole shapes array, and $redact strips it down to the expected result.
Note that {$not:"$color"} is necessary, because it will scan the top document as well, and if $redact does not find a color field on the top level this will return false that might strip the whole document which we don't want.

Better you can query in matching array element using $slice is it helpful to returning the significant object in an array.
db.test.find({"shapes.color" : "blue"}, {"shapes.$" : 1})
$slice is helpful when you know the index of the element, but sometimes you want
whichever array element matched your criteria. You can return the matching element
with the $ operator.

db.getCollection('aj').find({"shapes.color":"red"},{"shapes.$":1})
OUTPUTS
{
"shapes" : [
{
"shape" : "circle",
"color" : "red"
}
]
}

The syntax for find in mongodb is
db.<collection name>.find(query, projection);
and the second query that you have written, that is
db.test.find(
{shapes: {"$elemMatch": {color: "red"}}},
{"shapes.color":1})
in this you have used the $elemMatch operator in query part, whereas if you use this operator in the projection part then you will get the desired result. You can write down your query as
db.users.find(
{"shapes.color":"red"},
{_id:0, shapes: {$elemMatch : {color: "red"}}})
This will give you the desired result.

Thanks to JohnnyHK.
Here I just want to add some more complex usage.
// Document
{
"_id" : 1
"shapes" : [
{"shape" : "square", "color" : "red"},
{"shape" : "circle", "color" : "green"}
]
}
{
"_id" : 2
"shapes" : [
{"shape" : "square", "color" : "red"},
{"shape" : "circle", "color" : "green"}
]
}
// The Query
db.contents.find({
"_id" : ObjectId(1),
"shapes.color":"red"
},{
"_id": 0,
"shapes" :{
"$elemMatch":{
"color" : "red"
}
}
})
//And the Result
{"shapes":[
{
"shape" : "square",
"color" : "red"
}
]}

You just need to run query
db.test.find(
{"shapes.color": "red"},
{shapes: {$elemMatch: {color: "red"}}});
output of this query is
{
"_id" : ObjectId("562e7c594c12942f08fe4192"),
"shapes" : [
{"shape" : "circle", "color" : "red"}
]
}
as you expected it'll gives the exact field from array that matches color:'red'.

Along with $project it will be more appropriate other wise matching elements will be clubbed together with other elements in document.
db.test.aggregate(
{ "$unwind" : "$shapes" },
{ "$match" : { "shapes.color": "red" } },
{
"$project": {
"_id":1,
"item":1
}
}
)

Likewise you can find for the multiple
db.getCollection('localData').aggregate([
// Get just the docs that contain a shapes element where color is 'red'
{$match: {'shapes.color': {$in : ['red','yellow'] } }},
{$project: {
shapes: {$filter: {
input: '$shapes',
as: 'shape',
cond: {$in: ['$$shape.color', ['red', 'yellow']]}
}}
}}
])

db.test.find( {"shapes.color": "red"}, {_id: 0})

Use aggregation function and $project to get specific object field in document
db.getCollection('geolocations').aggregate([ { $project : { geolocation : 1} } ])
result:
{
"_id" : ObjectId("5e3ee15968879c0d5942464b"),
"geolocation" : [
{
"_id" : ObjectId("5e3ee3ee68879c0d5942465e"),
"latitude" : 12.9718313,
"longitude" : 77.593551,
"country" : "India",
"city" : "Chennai",
"zipcode" : "560001",
"streetName" : "Sidney Road",
"countryCode" : "in",
"ip" : "116.75.115.248",
"date" : ISODate("2020-02-08T16:38:06.584Z")
}
]
}

Although the question was asked 9.6 years ago, this has been of immense help to numerous people, me being one of them. Thank you everyone for all your queries, hints and answers. Picking up from one of the answers here.. I found that the following method can also be used to project other fields in the parent document.This may be helpful to someone.
For the following document, the need was to find out if an employee (emp #7839) has his leave history set for the year 2020. Leave history is implemented as an embedded document within the parent Employee document.
db.employees.find( {"leave_history.calendar_year": 2020},
{leave_history: {$elemMatch: {calendar_year: 2020}},empno:true,ename:true}).pretty()
{
"_id" : ObjectId("5e907ad23997181dde06e8fc"),
"empno" : 7839,
"ename" : "KING",
"mgrno" : 0,
"hiredate" : "1990-05-09",
"sal" : 100000,
"deptno" : {
"_id" : ObjectId("5e9065f53997181dde06e8f8")
},
"username" : "none",
"password" : "none",
"is_admin" : "N",
"is_approver" : "Y",
"is_manager" : "Y",
"user_role" : "AP",
"admin_approval_received" : "Y",
"active" : "Y",
"created_date" : "2020-04-10",
"updated_date" : "2020-04-10",
"application_usage_log" : [
{
"logged_in_as" : "AP",
"log_in_date" : "2020-04-10"
},
{
"logged_in_as" : "EM",
"log_in_date" : ISODate("2020-04-16T07:28:11.959Z")
}
],
"leave_history" : [
{
"calendar_year" : 2020,
"pl_used" : 0,
"cl_used" : 0,
"sl_used" : 0
},
{
"calendar_year" : 2021,
"pl_used" : 0,
"cl_used" : 0,
"sl_used" : 0
}
]
}

if you want to do filter, set and find at the same time.
let post = await Post.findOneAndUpdate(
{
_id: req.params.id,
tasks: {
$elemMatch: {
id: req.params.jobId,
date,
},
},
},
{
$set: {
'jobs.$[i].performer': performer,
'jobs.$[i].status': status,
'jobs.$[i].type': type,
},
},
{
arrayFilters: [
{
'i.id': req.params.jobId,
},
],
new: true,
}
);

This answer does not fully answer the question but it's related and I'm writing it down because someone decided to close another question marking this one as duplicate (which is not).
In my case I only wanted to filter the array elements but still return the full elements of the array. All previous answers (including the solution given in the question) gave me headaches when applying them to my particular case because:
I needed my solution to be able to return multiple results of the subarray elements.
Using $unwind + $match + $group resulted in losing root documents without matching array elements, which I didn't want to in my case because in fact I was only looking to filter out unwanted elements.
Using $project > $filter resulted in loosing the rest of the fields or the root documents or forced me to specify all of them in the projection as well which was not desirable.
So at the end I fixed all of this problems with an $addFields > $filter like this:
db.test.aggregate([
{ $match: { 'shapes.color': 'red' } },
{ $addFields: { 'shapes': { $filter: {
input: '$shapes',
as: 'shape',
cond: { $eq: ['$$shape.color', 'red'] }
} } } },
])
Explanation:
First match documents with a red coloured shape.
For those documents, add a field called shapes, which in this case will replace the original field called the same way.
To calculate the new value of shapes, $filter the elements of the original $shapes array, temporarily naming each of the array elements as shape so that later we can check if the $$shape.color is red.
Now the new shapes array only contains the desired elements.

for more details refer =
mongo db official referance
suppose you have document like this (you can have multiple document too) -
{
"_id": {
"$oid": "63b5cfbfbcc3196a2a23c44b"
},
"results": [
{
"yearOfRelease": "2022",
"imagePath": "https://upload.wikimedia.org/wikipedia/en/d/d4/The_Kashmir_Files_poster.jpg",
"title": "The Kashmir Files",
"overview": "Krishna endeavours to uncover the reason behind his parents' brutal killings in Kashmir. He is shocked to uncover a web of lies and conspiracies in connection with the massive genocide.",
"originalLanguage": "hi",
"imdbRating": "8.3",
"isbookMark": null,
"originCountry": "india",
"productionHouse": [
"Zee Studios"
],
"_id": {
"$oid": "63b5cfbfbcc3196a2a23c44c"
}
},
{
"yearOfRelease": "2022",
"imagePath": "https://upload.wikimedia.org/wikipedia/en/a/a9/Black_Adam_%28film%29_poster.jpg",
"title": "Black Adam",
"overview": "In ancient Kahndaq, Teth Adam was bestowed the almighty powers of the gods. After using these powers for vengeance, he was imprisoned, becoming Black Adam. Nearly 5,000 years have passed, and Black Adam has gone from man to myth to legend. Now free, his unique form of justice, born out of rage, is challenged by modern-day heroes who form the Justice Society: Hawkman, Dr. Fate, Atom Smasher and Cyclone",
"originalLanguage": "en",
"imdbRating": "8.3",
"isbookMark": null,
"originCountry": "United States of America",
"productionHouse": [
"DC Comics"
],
"_id": {
"$oid": "63b5cfbfbcc3196a2a23c44d"
}
},
{
"yearOfRelease": "2022",
"imagePath": "https://upload.wikimedia.org/wikipedia/en/0/09/The_Sea_Beast_film_poster.png",
"title": "The Sea Beast",
"overview": "A young girl stows away on the ship of a legendary sea monster hunter, turning his life upside down as they venture into uncharted waters.",
"originalLanguage": "en",
"imdbRating": "7.1",
"isbookMark": null,
"originCountry": "United States Canada",
"productionHouse": [
"Netflix Animation"
],
"_id": {
"$oid": "63b5cfbfbcc3196a2a23c44e"
}
},
{
"yearOfRelease": "2021",
"imagePath": "https://upload.wikimedia.org/wikipedia/en/7/7d/Hum_Do_Hamare_Do_poster.jpg",
"title": "Hum Do Hamare Do",
"overview": "Dhruv, who grew up an orphan, is in love with a woman who wishes to marry someone with a family. In order to fulfil his lover's wish, he hires two older individuals to pose as his parents.",
"originalLanguage": "hi",
"imdbRating": "6.0",
"isbookMark": null,
"originCountry": "india",
"productionHouse": [
"Maddock Films"
],
"_id": {
"$oid": "63b5cfbfbcc3196a2a23c44f"
}
},
{
"yearOfRelease": "2021",
"imagePath": "https://upload.wikimedia.org/wikipedia/en/7/74/Shang-Chi_and_the_Legend_of_the_Ten_Rings_poster.jpeg",
"title": "Shang-Chi and the Legend of the Ten Rings",
"overview": "Shang-Chi, a martial artist, lives a quiet life after he leaves his father and the shadowy Ten Rings organisation behind. Years later, he is forced to confront his past when the Ten Rings attack him.",
"originalLanguage": "en",
"imdbRating": "7.4",
"isbookMark": null,
"originCountry": "United States of America",
"productionHouse": [
"Marvel Entertainment"
],
"_id": {
"$oid": "63b5cfbfbcc3196a2a23c450"
}
}
],
"__v": 0
}
=======
mongo db query by aggregate command -
mongomodels.movieMainPageSchema.aggregate(
[
{
$project: {
_id:0, // to supress id
results: {
$filter: {
input: "$results",
as: "result",
cond: { $eq: [ "$$result.yearOfRelease", "2022" ] }
}
}
}
}
]
)

For the new version of MongoDB, it's slightly different.
For db.collection.find you can use the second parameter of find with the key being projection
db.collection.find({}, {projection: {name: 1, email: 0}});
You can also use the .project() method.
However, it is not a native MongoDB method, it's a method provided by most MongoDB driver like Mongoose, MongoDB Node.js driver etc.
db.collection.find({}).project({name: 1, email: 0});
And if you want to use findOne, it's the same that with find
db.collection.findOne({}, {projection: {name: 1, email: 0}});
But findOne doesn't have a .project() method.

Query DSL not working in pyes search

I am trying to use a custom query DSL to get results using the pyes library. I have query DSL that works when I use the command line
curl -XGET localhost:9200/test_index/_search -d '{
"query": {
"function_score": {
"query": {
"match_all": {}
},
"field_value_factor": {
"field": "starred",
"modifier": "none",
"factor": 2
}
}
},
"aggs" : {
"types" : {
"filters" : {
"filters" : {
"category1" : { "type" : { "value" : "category1"}},
"category2" : { "type" : { "value" : "category2"}},
"category3" : { "type" : { "value" : "category3"}},
"category4": { "type" : { "value" : "category4"}},
"category5" : { "type" : { "value" : "category5"}}
}
},
"aggs": {
"topFoundHits": {
"top_hits": {
"size": 5
}
}
}
}
}
}'
The idea here is to search across many categorized documents for all documents matching a particular string query. Then using aggregations I want to find the top five resulting documents by category. Starred items are boosted so that they show up above other search results.
This works great when I enter the command as listed above directly in terminal but it doesn't work when I try to put it in pyes. I'm not sure what the best way is to do it. The documentation for the pyes library is really confusing for me to translate this totally into pyes objects.
I'm trying to do the following:
query_dsl = self.get_text_index_query_dsl()
resulting_docs = conn.search(query=query_dsl)
(where self.get_test_index_query_dsl returns the query dsl dict above)
Searching as is gives me a:
ElasticSearchException: QueryParsingException[[test_index] No query registered for [query]]; }]
If I remove the parent "query" mapping and try:
query_dsl = {
"function_score": {
"query": {
"match_all": {}
},
"field_value_factor": {
"field": "starred",
"modifier": "none",
"factor": 2
}
},
"aggs" : {
"types" : {
"filters" : {
"filters" : {
"category1" : { "type" : { "value" : "category1"}},
"category2" : { "type" : { "value" : "category2"}},
"category3" : { "type" : { "value" : "category3"}},
"category4": { "type" : { "value" : "category4"}},
"category5" : { "type" : { "value" : "category5"}}
}
},
"aggs": {
"topFoundHits": {
"top_hits": {
"size": 5
}
}
}
}
}
}
This also errors out with: ElasticSearchException: ElasticsearchParseException[Expected field name but got START_OBJECT "aggs"]; }]
These errors in addition to the fact that pyes doesn't seem to have a 'topFoundHits' functionality yet (I think) are holding me up.
Any ideas why this is happening and how to fix it?
Thank you so much!

I got this working using this library where you can just use your regular query dsl JSON syntax : http://elasticsearch-dsl.readthedocs.org/en/latest/.