MongoDB: Transform array of objects to array of arrays - mongodb

Question

I have a collection named "records" that contains documents in the following form:
{
"name": "a"
"items": [
{
"a": "5",
"b": "1",
"c": "2"
},
{
"a": "6",
"b": "3",
"c": "7"
}
]
}
I want to keep the data just as it is in the database (to make the data easy to read and interpret). But I'd like to run a query that returns the data in the following form:
{
"name": "a"
"items": [
["5", "1", "2"],
["6", "3", "7"],
]
}
Is this possible with pymongo? I know I can run a query and translate the documents using Python, but I'd like to avoid iterating over the query result if possible.

Answer 1

I have a table named "records"
Collection
Is this possible with pymongo?
Yes
Any pointers on how to approach this would be super helpful!
I'd suggest you to use a view to transform your data during a query in MongoDB.
In this way, you can get transformed data and apply find to already transformed data if you need.
db.createCollection(
"view_name",
{"viewOn": "original_collection_name",
"pipeline": [{$unwind: "$items"},
{$project: {name: 1, items: {$objectToArray: "$items"}}},
{$project: {name: 1, items: {$concatArrays: ["$items.v"]}}},
{$group: {_id: "$_id", name: {$first: "$name"},
items: {$push: "$items"}}}]
}
)
> db.view_name.find({name: "a"})
{ "_id" : ObjectId("5fc3dbb69cb76f866582620f"), "name" : "a", "items" : [ [ "5", "1", "2" ], [ "6", "3", "7" ] ] }
> db.view_name.find({"items": {$in: [["5", "1", "2"]]}})
{ "_id" : ObjectId("5fc3dbb69cb76f866582620f"), "name" : "a", "items" : [ [ "5", "1", "2" ], [ "6", "3", "7" ] ] }
> db.view_name.find()
{ "_id" : ObjectId("5fc3dbb69cb76f866582620f"), "name" : "a", "items" : [ [ "5", "1", "2" ], [ "6", "3", "7" ] ] }
Query:
db.original_collection_name.aggregate([
{$unwind: "$items"},
{$project: {name: 1, items: {$objectToArray: "$items"}}},
{$project: {name: 1, items: {$concatArrays: ["$items.v"]}}},
{$group: {_id: "$_id", name: {$first: "$name"}, items: {$push: "$items"}}}])

Answer 2

Using $objectToArray and $map transformations:
// { name: "a", items: [ { a: "5", b: "1", c: "2" }, { a: "6", b: "3", c: "7" } ] }
db.collection.aggregate([
{ $set: { items: { $map: { input: "$items", as: "x", in: { $objectToArray: "$$x" } } } } },
// {
// name: "a",
// items: [
// [ { k: "a", v: "5" }, { k: "b", v: "1" }, { k: "c", v: "2" } ],
// [ { k: "a", v: "6" }, { k: "b", v: "3" }, { k: "c", v: "7" } ]
// ]
// }
{ $set: { items: { $map: { input: "$items", as: "x", in: "$$x.v" } } } }
])
// { name: "a", items: [["5", "1", "2"], ["6", "3", "7"]] }
This maps items' elements as key/value arrays such that { field: "value" } becomes [ { k: "field", v: "value" } ]. This way whatever the field name, we can easily access the value using v, which is the role of the second $set stage: "$$x.v".
This has the benefit of avoiding heavy stages such as unwind/group.
Note that you can also imbricate the second $map within the first; but that's probably less readable.