I want to split the following array according to the group-value. I know I can do this using $unwind
and $group
. Is there any way to this in a single $project
-stage?
Input
{
"_id": 1,
"some_field": "some_value",
"array": [
{
"group": "a",
"subgroup": "aa",
"value": 1
},
{
"group": "b",
"subgroup": "bb",
"value": 2
},
{
"group": "a",
"subgroup": "ab",
"value": 2
}
]
}
desired output:
{
"_id": 1,
"some_field": "some_value",
"array": [
{
"group": "a",
"values": [
{
"subgroup": "aa",
"value": 1
},
{
"subgroup": "ab",
"value": 2
}
]
},
{
"group": "b",
"values": [
{
"subgroup": "bb",
"value": 2
}
]
}
]
}
CodePudding user response:
Try this: https://mongoplayground.net/p/pFn3tLtAG4D
$set: {
_id: "$_id",
some_field: "$some_field",
array: {
$map: {
input: {
$setUnion: [
"$array.group"
]
},
in: {
group: "$$this",
values: {
$map: {
input: {
$filter: {
input: "$array",
as: "elem",
cond: {
$eq: [
"$$elem.group",
"$$this"
]
}
}
},
as: "vals",
in: {
subgroup: "$$vals.subgroup",
value: "$$vals.value"
}
}
}
}
}
}
}
CodePudding user response:
This is far from a single project
stage, but it does produce the desired output from the given input.
db.collection.aggregate([
{'$match': {'_id': 1}},
{'$unwind': '$array'},
{'$project': {'array': {'group': '$array.group', 'values': '$array'},
'some_field': 1,
'my_id': '$_id'}},
{'$unset': 'array.values.group'},
{'$group': {'_id': '$array.group',
'values': {'$push': '$array.values'},
'some_field': {'$first': '$some_field'},
'my_id': {'$first': '$my_id'}}},
{'$set': {'array': {'group': '$_id', 'values': '$values'}}},
{'$unset': 'values'},
{'$group': {'_id': '$my_id',
'array': {'$push': '$array'},
'some_field': {'$first': '$some_field'}}}
])
Try it on mongoplayground.net.
CodePudding user response:
It is doable, it's definitely not clean or sexy.
My approach is to use $reduce
and $mergeObjects
, we'll iterate over the array and keep reconstructing the result.
The main issue that plagues this approach is this feature that doesn't allow to $concatArrays
expressions, so we have to use some very ugly workarounds.
Anyways here is how you can achieve this:
db.collection.aggregate([
{
$project: {
_id: 1,
some_field: 1,
array: {
$map: {
input: {
"$objectToArray": {
$reduce: {
input: "$array",
initialValue: {},
in: {
"$mergeObjects": [
"$$value",
{
"$arrayToObject": [
[
{
k: "$$this.group",
v: {
$map: {
input: {
"$concatArrays": [
[
"$$this"
],
{
$map: {
input: {
$filter: {
input: {
"$objectToArray": "$$value"
},
as: "filterItem",
cond: {
$eq: [
"$$filterItem.k",
"$$this.group"
]
}
}
},
as: "mapItem",
in: "$$mapItem.v"
}
},
]
},
as: "map2Item",
in: {
$cond: [
{
"$isArray": "$$map2Item"
},
{
$arrayElemAt: [
"$$map2Item",
0
]
},
"$$map2Item"
]
}
}
}
}
]
]
}
]
}
}
}
},
as: "item",
in: {
group: "$$item.k",
values: "$$item.v"
}
}
}
}
}
])