Mongoose (MongoDB): Exclude Properties in a Dictionary Like Schema Type - node.js

Question

I have a Schema of the following structure:
var schema = mongoose.Schema({
answers: {type: mongoose.Schema.Types.Mixed}
});
I use the answers field as an object (associative array to implement something like a dictionary). Here is an example:
{
"__v": 0,
"_id": {
"$oid": "53a0251c50d0536c1bfc6006"
},
"answers": {
"fea": {
"viewed": false
},
"3d2": {
"viewed": true,
"value": true
},
"4fr": {
"viewed": true,
"value": true
},
"84h": {
"viewed": false
},
...
}
}
In a query I want to select only the "value" field of each entry. How is that possible through the select syntax? This of course doesn't work:
XY.find(...)
.select({'answers': true, 'answers.*.value': false})
.exec(...);
Maybe I have to design the data in another fashion?
Best regards,
Kersten

Answer 1

You should never model with "explicit values" as the "key" names. This is very bad practice. Consider what you would do in a SQL database. Would you create "fields/columns" for the different "names" of the things you want?
No you would not. You have a generic field that specifies a "type" and then you have others that hold the data. Nothing changes here:
{
"_id": {
"$oid": "53a0251c50d0536c1bfc6006"
},
"answers": [
{ "type": "fea", "viewed": false },
{ "type": "3d2", "viewed": true, "value": true },
{ "type": "4fr", "viewed": true, "value": true },
{ "type": "84h", "viewed": false },
...
]
}
Now this is easy to use something like the aggregation framework to make the projection of the content like you want:
With Modern MongoDB 2.6 and onwards you can use $map and $setDifference to filter the array without using $unwind:
Model.aggregate(
[
{ "$project": {
"answers": {
"$setDifference": [
{
"$map": {
"input": "$answers",
"as": "el",
"in": {
"$cond": [
1,
{
"type": "$$el.type",
"value": { "$ifNull": [ "$$el.value", false ] }
},
false
]
}
}
},
[false]
]
}
}}
],
function(err,result) {
}
);
Or with older versions pre 2.6:
Model.aggregate(
[
{ "$unwind": "$answers" },
{ "$group": {
"_id": "$_id",
"answers": {
"$push": {
"type": "$answers.type",
"value": { "$ifNull": [ "$answers.value", false ] }
}
}
}}
],
function(err,result) {
}
);
Of course you can "filter" the array results to certain conditions by either adding a logical evaluation as the first argument to the $cond operator in the $map implementation. Or by using a $match pipeline stage in between the $unwind and $group stages.
Either form allows you to re-shape the result without problem and is the fastest way to process this, which is a strong advantage of using arrays as opposed to embedded objects whose keys are actually a "data" item.
If you are stuck with this then you need to process with JavaScript evaluation like mapReduce. This runs much slower than the aggregation framework due to the need to invoke and run in a JavaScript interpreter process:
Model.mapReduce(
{
"map": function() {
for ( var k in this.answers ) {
this.answers[k] = this.answers[k].hasOwnProperty("value")
? this.answers[k].value : false;
}
var id = this._id;
delete this._id;
emit( id, this );
},
"reduce": function(){}
},
function(err,docs) {
}
)
But really, consider changing your structure as it makes things much more flexible for queries and other operations.