Filter subdocument by datetime - mongodb

Question

I've the following model
var messageSchema = new Schema({
creationDate: { type: Date, default: Date.now },
comment: { type: String },
author: { type: Schema.Types.ObjectId }
});
var conversationSchema = new Schema({
title: { type: String },
author: { type : Schema.Types.ObjectId },
members: [ { type: Schema.Types.ObjectId } ],
creationDate: { type: Date, default: Date.now },
lastUpdate: { type: Date, default: Date.now },
comments: [ messageSchema ]
});
I want to create two methods to get the comments generated after a date by user or by conversationId.
By User
I tried with the following method
var query = {
members : { $all : [ userId, otherUserId ], "$size" : 2 }
, comments : { $elemMatch : { creationDate : { $gte: from } } }
};
When there are no comments after the specified date (at from) the method returns [] or null
By conversationId
The same happen when I try to get by user id
var query = { _id : conversationId
, comments : { $elemMatch : { creationDate : { $gte: from } } }
};
Is there any way to make the method returns the conversation information with an empty comments?
Thank you!

Answer 1

Sounds like a couple of problems here, but stepping through them all
In order to get more than a single match "or" none from an array to need the aggregation framework of mapReduce to do this. You could try "projecting" with $elemMatch but this can only return the "first" match. i.e:
{ "a": [1,2,3] }
db.collection.find({ },{ "$elemMatch": { "$gte": 2 } })
{ "a": [2] }
So standard projection does not work for this. It can return an "empty" array but it an also only return the "first" that is matched.
Moving along, you also have this in your code:
{ $all : [ userId, otherUserId ], "$site" : 2 }
Where $site is not a valid operator. I think you mean $size but there are actuall "two" operators with that name and your intent may not be clear here.
If you mean that the array you are testing must have "only two" elements, then this is the operator for you. If you meant that the matched conversation between the two people had to be equal to both in the match, then $all does this anyway so the $size becomes redundant in either case unless you don't want anyone else in the conversation.
On to the aggregation problem. You need to "filter" the content of the array in a "non-destructive way" in order to get more than one match or an empty array.
The best approach for this is with modern MongoDB features available from 2.6, which allows the array content to be filtered without processing $unwind:
Model.aggregate(
[
{ "$match": {
"members": { "$all": [userId,otherUserId] }
}},
{ "$project": {
"title": 1,
"author": 1,
"members": 1,
"creationDate": 1,
"lastUpdate": 1,
"comments": {
"$setDifference": [
{ "$map": {
"input": "$comments",
"as": "c",
"in": { "$cond": [
{ "$gte": [ "$$c.creationDate", from ] },
"$$c",
false
]}
}},
[false]
]
}
}}
],
function(err,result) {
}
);
That uses $map which can process an expression against each array element. In this case the vallues are tested under the $cond ternary to either return the array element where the condition is true or otherwise return false as the element.
These are then "filtered" by the $setDifference operator which essentially compares the resulting array of $map to the other array [false]. This removes any false values from the result array and only leaves matched elements or no elements at all.
An alternate may have been $redact but since your document contains "creationDate" at multiple levels, then this messes with the logic used with it's $$DESCEND operator. This rules that action out.
In earlier versions "not destroying" the array needs to be treated with care. So you need to do much the same "filter" of results in order to get the "empty" array you want:
Model.aggregate(
[
{ "$match": {
"$and": [
{ "members": userId },
{ "members": otherUserId }
}},
{ "$unwind": "$comments" },
{ "$group": {
"_id": "$_id",
"title": { "$first": "$title" },
"author": { "$first": "$author" },
"members": { "$first": "$members" },
"creationDate": { "$first": "$creationDate" },
"lastUpdate": { "$first": "$lastUpdate" },
"comments": {
"$addToSet": {
"$cond": [
{ "$gte": [ "$comments.creationDate", from ] },
"$comments",
false
]
}
},
"matchedSize": {
"$sum": {
"$cond": [
{ "$gte": [ "$comments.creationDate", from ] },
1,
0
]
}
}
}},
{ "$unwind": "$comments" },
{ "$match": {
"$or": [
{ "comments": { "$ne": false } },
{ "matchedSize": 0 }
]
}},
{ "$group": {
"_id": "$_id",
"title": { "$first": "$title" },
"author": { "$first": "$author" },
"members": { "$first": "$members" },
"creationDate": { "$first": "$creationDate" },
"lastUpdate": { "$first": "$lastUpdate" },
"comments": { "$push": "$comments" }
}},
{ "$project": {
"title": 1,
"author": 1,
"members": 1,
"creationDate": 1,
"lastUpdate": 1,
"comments": {
"$cond": [
{ "$eq": [ "$comments", [false] ] },
{ "$const": [] },
"$comments"
]
}
}}
],
function(err,result) {
}
)
This does much of the same things, but longer. In order to look at the array content you need to $unwind the content. When you $group back, you look at each element to see if it matches the condition to decide what to return, also keeping a count of the matches.
This is going to put some ( one with $addToSet ) false results in the array or only an array with the entry false where there are no matches. So yo filter these out with $match but also testing on the matched "count" to see if no matches were found. If no match was found then you don't throw away that item.
Instead you replace the [false] arrays with empty arrays in a final $project.
So depending on your MongoDB version this is either "fast/easy" or "slow/hard" to process. Compelling reasons to update a version already many years old.
Working example
var async = require('async'),
mongoose = require('mongoose'),
Schema = mongoose.Schema;
mongoose.connect('mongodb://localhost/aggtest');
var memberSchema = new Schema({
name: { type: String }
});
var messageSchema = new Schema({
creationDate: { type: Date, default: Date.now },
comment: { type: String },
});
var conversationSchema = new Schema({
members: [ { type: Schema.Types.ObjectId } ],
comments: [messageSchema]
});
var Member = mongoose.model( 'Member', memberSchema );
var Conversation = mongoose.model( 'Conversation', conversationSchema );
async.waterfall(
[
// Clean
function(callback) {
async.each([Member,Conversation],function(model,callback) {
model.remove({},callback);
},
function(err) {
callback(err);
});
},
// add some people
function(callback) {
async.map(["bill","ted","fred"],function(name,callback) {
Member.create({ "name": name },callback);
},callback);
},
// Create a conversation
function(names,callback) {
var conv = new Conversation();
names.forEach(function(el) {
conv.members.push(el._id);
});
conv.save(function(err,conv) {
callback(err,conv,names)
});
},
// add some comments
function(conv,names,callback) {
async.eachSeries(names,function(name,callback) {
Conversation.update(
{ "_id": conv._id },
{ "$push": { "comments": { "comment": name.name } } },
callback
);
},function(err) {
callback(err,names);
});
},
function(names,callback) {
Conversation.findOne({},function(err,conv) {
callback(err,names,conv.comments[1].creationDate);
});
},
function(names,from,callback) {
var ids = names.map(function(el) {
return el._id
});
var pipeline = [
{ "$match": {
"$and": [
{ "members": ids[0] },
{ "members": ids[1] }
]
}},
{ "$project": {
"members": 1,
"comments": {
"$setDifference": [
{ "$map": {
"input": "$comments",
"as": "c",
"in": { "$cond": [
{ "$gte": [ "$$c.creationDate", from ] },
"$$c",
false
]}
}},
[false]
]
}
}}
];
//console.log(JSON.stringify(pipeline, undefined, 2 ));
Conversation.aggregate(
pipeline,
function(err,result) {
if(err) throw err;
console.log(JSON.stringify(result, undefined, 2 ));
callback(err);
}
)
}
],
function(err) {
if (err) throw err;
process.exit();
}
);
Which produces this output:
[
{
"_id": "55a63133dcbf671918b51a93",
"comments": [
{
"comment": "ted",
"_id": "55a63133dcbf671918b51a95",
"creationDate": "2015-07-15T10:08:51.217Z"
},
{
"comment": "fred",
"_id": "55a63133dcbf671918b51a96",
"creationDate": "2015-07-15T10:08:51.220Z"
}
],
"members": [
"55a63133dcbf671918b51a90",
"55a63133dcbf671918b51a91",
"55a63133dcbf671918b51a92"
]
}
]
Note the "comments" only contain the last two entries which are "greater than or equal" to the date which was used as input ( being the date from the second comment ).