MongoDB Property String Starts With Query And Set Group Id - mongodb

Question

Data structure - one document in a big collection:
{
OPERATINGSYSTEM: "Android 6.0"
}
Issue: The operatingsystem can equal e.g. "Android 5.0", "Android 6.0", "Windows Phone", "Windows Phone 8.1"
There is no property which only contains the kind of operating system e.g. only Android.
I need to get the count of windows phones, and android phones.
My temporary solution:
db.getCollection('RB').find(
{OPERATINGSYSTEM: {$regex: "^Android"}}
).count();
I'm doing that query replacing "^Android" by windows phone and so on which takes much time and needs to be done in parallel.
Using the aggregation framework I though about this:
db.RB.aggregate(
{$group: {_id: {OPERATINGSYSTEM:"$OPERATINGSYSTEM"}}},)
But using this I get an entry for each operatingsystem version Android 5.0, Android 6.0 etc...
The solution I'm searching for should return data in this format:
{
"Android": 50,
"Windows Phone": 100
}
How can this be done in a single query?

Answer 1

Provided your strings at least consistently have the numeric version as the last thing in the string, then you could use $split with the aggregation framework to make an array from the "space delimited" content, then remove the last element from the array before reconstructing:
Given data like :
{ "name" : "Android 6.0" }
{ "name" : "Android 7.0" }
{ "name" : "Windows Phone 10" }
You can try:
db.getCollection('phones').aggregate([
{ "$group": {
"_id": {
"$let": {
"vars": { "split": { "$split": [ "$name", " " ] } },
"in": {
"$reduce": {
"input": { "$slice": [ "$$split", 0, { "$subtract": [ { "$size": "$$split" }, 1 ] } ] },
"initialValue": "",
"in": {
"$cond": {
"if": { "$eq": [ "$$value", "" ] },
"then": "$$this",
"else": { "$concat": [ "$$value", " ", "$$this" ] }
}
}
}
}
}
},
"count": { "$sum": 1 }
}},
{ "$replaceRoot": {
"newRoot": {
"$arrayToObject": [[{ "k": "$_id", "v": "$count" }]]
}
}}
])
That's all possible if your MongoDB is at least MongoDB 3.4 to support both $split and $reduce. The $replaceRoot is really about naming the keys, and not really required.
Alternately you can use mapReduce:
db.getCollection('phones').mapReduce(
function() {
var re = /\d+/g;
emit(this.name.substr(0,this.name.search(re)-1),1);
},
function(key,values) { return Array.sum(values) },
{ "out": { "inline": 1 } }
)
Where it's easier to break down the string by the index where a numeric value occurs. In either case, you are not required to "hardcode" anything, and the values of the keys are completely dependent on the strings in context.
Keep in mind though that unless there is an extremely large number of possible values, then running parallel .count() operations "should" be the fastest to process since returning cursor counts is a lot faster than actually counting the aggregated entries.

Answer 2

You can use map reduce, and apply your logic in the map function.
var map = function(){
var name = this.op.includes("android") ? "Android" : ""; // could be a regexp
if(name === ""){
name = this.op.includes("windows") ? "Windows" : "";
}
emit(name, 1);
}
var reduce = function(key, values){
return Array.sum(values)
}
db.operating.mapReduce(map, reduce, {out: "total"})
https://docs.mongodb.com/manual/tutorial/map-reduce-examples/