Filter sub-document array using substring as criteria - mongodb

Question

My collection:
{
title: 'Computers',
maincategories:[
{
title: 'Monitors',
subcategories:[
{
title: '24 inch',
code: 'AFG'
}
]
}
]
}
I want query the code. The code is just the first part so I want to have all subcategories that contains the given search. So AFG101 would return this subcategories.
My query:
module.exports = (req, res) => {
var q = {
'maincategories.subcategories': {
$elemMatch: {
code: 'AFG101'
}
}
};
var query = mongoose.model('TypeCategory').find(q, {'maincategories.$': 1, 'title': 1});
query.exec((err, docs) => {
res.status(200).send(docs);
});
};
My problem:
How do I search for a part of a string? AFG101 should return all subcategories with property code containing any part of the string. So in this case, AFG would be a hit. Same as in this sql question: MySQL: What is a reverse version of LIKE?
How do I project the subcategories. Current query returns all subcategories. I only want to returns those hitting.

Answer 1

The best way to do this is in MongoDB 3.4 using the $indexOfCP string aggregation operator.
let code = "afg101";
db.collection.aggregate([
{ "$project": {
"title": 1,
"maincategories": {
"$map": {
"input": "$maincategories",
"as": "mc",
"in": {
"$filter": {
"input": "$$mc.subcategories",
"as": "subcat",
"cond": {
"$gt": [
{
"$indexOfCP": [
code,
{ "$toLower": "$$subcat.code" }
]
},
-1
]
}
}
}
}
}
}}
])
which returns:
{
"_id" : ObjectId("582cba57e6f570d40d77b3a8"),
"title" : "Computers",
"maincategories" : [
[
{
"title" : "24 inch",
"code" : "AFG"
}
]
]
}
You can read my other answers to similar question 1, 2 and 3.
From 3.2 backward, the only way to do this is with mapReduce.
db.collection.mapReduce(
function() {
var code = 'AFG101';
var maincategories = this.maincategories.map(function(sdoc) {
return {
"title": sdoc.title,
"subcategories": sdoc.subcategories.filter(function(scat) {
return code.indexOf(scat.code) != -1;
}
)};
});
emit(this._id, maincategories);
},
function(key, value) {},
{ "out": { "inline": 1 }
})
which yields something like this:
{
"results" : [
{
"_id" : ObjectId("582c9a1aa358615b6352c45a"),
"value" : [
{
"title" : "Monitors",
"subcategories" : [
{
"title" : "24 inch",
"code" : "AFG"
}
]
}
]
}
],
"timeMillis" : 15,
"counts" : {
"input" : 1,
"emit" : 1,
"reduce" : 0,
"output" : 1
},
"ok" : 1
}

Answer 2

Well, just like your question has two parts, I could think of two separate solution, however I don't see a way to join them together.
For first part $where can be used to do a reverse regex, but it's dirty, it's an overkill and it can't use any indexes, since $where runs on each documents.
db.TypeCategory.find({$where:function(){for(var i in this.maincategories)
{for(var j in this.maincategories[i].subcategories)
{if("AFG101".indexOf(this.maincategories[i].subcategories[j].code)>=0)
{return true}}}}},{"maincategories.subcategories.code":1})
Even if you use this option, it would need couple of boundary check and it cannot project two level of nested array. MongoDB doesn't support such projection (yet).
For that purpose we might go for aggregation
db.TypeCategory.aggregate([{$unwind:"$maincategories"},
{$unwind:"$maincategories.subcategories"},
{$match:{"maincategories.subcategories.code":"AFG"}},
{$group:{_id:"$_id","maincategories":{$push:"$maincategories"}}}
])
However I don't think there is a way to do reverse regex check in aggregation, but I might be wrong too. Also this aggregation is costly since there are two unwinds which can lead to overflow the memory limit for aggregation for a really large collection.

Answer 3

You can use $substr and do it
db.getCollection('cat').aggregate([
{"$unwind" : "$maincategories"},
{"$unwind" : "$maincategories.subcategories"},
{"$project" :
{"maincategories" : 1,
"title":1,"sub" : {"$substr" :["$maincategories.subcategories.code",0,3]}}},
{"$match" : {"sub" : "AFG"}},
{"$project" :
{"maincategories" : 1,
"title":1}
}
])