Grouping over multiple fields in MongoDb - mongodb

Question

How would I go about grouping over multiple fields? I need to get a unique count for case insensitive true over multiple independent documents.
I've looked at both map/reduce and aggregation and I don't quite know what would be the best approach.
Lets say I have the following data in my collection:
/* 0 */
{
"_id" : ObjectId("****"),
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "true",
"A" : "true",
"B" : "true",
"C" : "",
}
}
/* 1 */
{
"_id" : ObjectId("****"),
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "true",
"A" : "true",
"B" : "true",
"C" : "",
"D" : "TRUE"
}
}
/* 2 */
{
"_id" : ObjectId("****"),
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "true",
"A" : "true",
"B" : "TRUE",
"C" : "",
"D" : "false"
}
}
/* 3 */
{
"_id" : ObjectId("****"),
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "false",
"A" : "true",
"B" : "false",
"D" : "true"
}
}
I would like to output the following data, formatting is not important:
isMail : 3
A : 4
B : 3
C : 0
D : 2
Total : 4

Answer 1

Using the conditional operator $cond to map "true" to 1 and anything else to 0, you might achieve the desired result. This is only complicated by the fact your "boolean" values are in fact strings, and that you have case variation on the "true" value -- that's why I use $toLower in the code below:
db.test.sample.aggregate([
{
$group: { _id:null,
isMail: { $sum: { $cond: [{$eq: [{$toLower:"$Data.isMail"}, "true"]}, 1, 0] }},
A: { $sum: { $cond: [{$eq: [{$toLower:"$Data.A"}, "true"]}, 1, 0] }},
B: { $sum: { $cond: [{$eq: [{$toLower:"$Data.B"}, "true"]}, 1, 0] }},
C: { $sum: { $cond: [{$eq: [{$toLower:"$Data.C"}, "true"]}, 1, 0] }},
D: { $sum: { $cond: [{$eq: [{$toLower:"$Data.D"}, "true"]}, 1, 0] }},
total: { $sum: 1 },
}
},
{
$project: {
_id: 0,
A: 1, B: 1, C:1, D:1, total:1, isMail:1,
}
}
])
Producing:
{ "isMail" : 3, "A" : 4, "B" : 3, "C" : 0, "D" : 2, "total" : 4 }

Answer 2

If you could change the schema design so that the data keys become the values, it would go a long way in making it easier for you to do some aggregation operations on the data. A better shchema would look like this:
{
"_id" : ObjectId("5548de01180e84997293903f"),
"IsPartOfBatch" : false,
"Data" : [
{
"key" : "isMail",
"value" : true
},
{
"key" : "A",
"value" : true
},
{
"key" : "B",
"value" : true
},
{
"key" : "C",
"value" : false
},
{
"key" : "D",
"value" : false
}
]
}
Let's use the sample data set you provided in your question:
db.test.insert([
{
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "true",
"A" : "true",
"B" : "true",
"C" : ""
}
},
{
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "true",
"A" : "true",
"B" : "true",
"C" : "",
"D" : "TRUE"
}
},
{
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "true",
"A" : "true",
"B" : "TRUE",
"C" : "",
"D" : "false"
}
},
{
"IsPartOfBatch" : false,
"Data" : {
"isMail" : "false",
"A" : "true",
"B" : "false",
"D" : "true"
}
}
]);
To change the schema so that it follows the above recommended structure, use the following code snippet (performance may be slow over very large datasets):
db.test.find({ "Data.isMail": { $type : 2 } }).forEach(function (doc){
var data = [];
if (doc.Data) {
for(key in doc.Data) {
var isTrueSet = (doc.Data[key] === "true" || doc.Data[key] === "TRUE")
var obj = {};
obj["key"] = key;
obj["value"] = isTrueSet;
data.push(obj);
};
}
doc.Data = data;
db.test.save(doc);
});
A simple db.test.findOne() query will give the result:
{
"_id" : ObjectId("5548de01180e84997293903f"),
"IsPartOfBatch" : false,
"Data" : [
{
"key" : "isMail",
"value" : true
},
{
"key" : "A",
"value" : true
},
{
"key" : "B",
"value" : true
},
{
"key" : "C",
"value" : false
},
{
"key" : "D",
"value" : false
}
]
}
Now you can use aggregation framework to get the counts of the keys with true values:
db.test.aggregate([
{
"$unwind": "$Data"
},
{
"$project": {
"_id": 0,
"key": "$Data.key",
"isTrue": {
"$cond": [{ "$eq": [ "$Data.value", true ] }, 1, 0]
}
}
},
{
"$group": {
"_id": "$key",
"count": {
"$sum": "$isTrue"
}
}
}
])
Output
/* 0 */
{
"result" : [
{
"_id" : "D",
"count" : 2
},
{
"_id" : "C",
"count" : 0
},
{
"_id" : "B",
"count" : 3
},
{
"_id" : "A",
"count" : 4
},
{
"_id" : "isMail",
"count" : 3
}
],
"ok" : 1
}
You can then further modify the result using native JavaScript functions as MongoDB's aggregation framework cannot project the field values as the keys thus you will have to rely on JS to do this:
var pipeline = [
{
"$unwind": "$Data"
},
{
"$project": {
"_id": 0,
"key": "$Data.key",
"isTrue": {
"$cond": [{ "$eq": [ "$Data.value", true ] }, 1, 0]
}
}
},
{
"$group": {
"_id": "$key",
"count": {
"$sum": "$isTrue"
}
}
}],
agg = db.test.aggregate(pipeline),
obj = {},
result = [];
agg.forEach(function (doc){
obj[doc._id] = doc.count;
result.push(obj);
});