SQL 'UNION ALL' like implementation in MongoDB

问题

There are two collections:

Sales

{
  "_id" : ObjectId("5ba0bfb8d1acdc0de716e839"),
  "invoiceNumber" : 1,
  "saleDate" : ISODate("2018-09-01T00:00:00.000Z"),
  "totalTaxAmount" : 613,
  "subTotalAmount" : 2000,
  "totalAmount" : 2613,
  "balance" : 2613,
  "financialYear" : "2018-2019",
  "modeOfPayment" : "Digital Transfer",
  "customerName": "Acme Inc"
}

Transactions

{
  "_id" : ObjectId("5bbb4e131fb8af0dc645212d"),
  "transactionNumber" : 1    
  "transactionDate" : ISODate("2018-09-03T00:00:00.000Z"),
  "transactionType" : "Income",
  "partyName" : "Acme Inc",
  "transactionMode" : "Digital Transfer",
  "amount" : 2613,
  "paidItems" : [ 
      {
          "orderId" : "5b90a7d62bb5a21be4ff97e3",
          "invoiceNumber" : "1",
          "orderType" : "sale",
          "totalAmount" : 2613,
          "balance" : 613,
          "payingAmount" : 2000
      }
   ]
}

I need to retrieve sales and transactions as 'heading' for the specific party (i.e. customerName, partyName) between two dates (i.e. saleDate, transactionDate) ordered by date; as follows:

[
  {
    "date": ISODate("2018-09-01T00:00:00.000Z"),
    "heading": "Sale",
    "particulars": "Invoice # 1",
    "amount": 2613
  },
  {
    "date": ISODate("2018-09-03T00:00:00.000Z"),
    "heading": "Payment by Digital Transfer",
    "particulars": "Transaction # 1",
    "amount": 2000
  }
]

I researched and tried with aggregation, $lookup but it's not returning what's desired.

Switching from SQL to MongoDB. In SQL following query works fine:

select sale_date as dated, 'Sale' as heading, 'Invoice # ' + 
convert(varchar(12),invoice_number) as particulars, 
convert(varchar(12), total) as amount, 
from sales where sale_date between @from_date AND @to_date AND 
customer_name=@customer_name
UNION ALL
select transaction_date as dated, 'Payment by ' + transaction_mode as 
heading, 'Transaction # ' + convert(varchar(12), transaction_id) as 
particulars, convert(varchar(12), amount) as amount from transactions 
where transaction_date between @from_date AND @to_date AND 
party_name=@customer_name
order by dated DESC

There's a feature request filed in the MongoDB community and it's 'unresolved'.

I would like to know is there any way for this within mongoShell or MongoDB driver (mongoose / JS). Using current stable versions of MongoDB, nodejs, express, and mongoose. Thanks!

回答1:

You can try below aggregation

db.sales.aggregate([
  { "$limit": 1 },
  { "$facet": {
    "collection1": [
      { "$limit": 1 },
      { "$lookup": {
        "from": "sales",
        "pipeline": [
          { "$match": {
            "date": { "$gte": ISODate("2018-09-01"), "$lte": ISODate("2018-09-10") },
            "customer.name": customerName
          }},
          { "$project": {
            "_id":0, "dated": "$saleDate", "heading": "Sale", "particulars": "$invoiceNumber",
            "amount": "$totalAmount", "modeOfPayment": null
          }}
        ],
        "as": "collection1"
      }}
    ],
    "collection2": [
      { "$limit": 1 },
      { "$lookup": {
        "from": "transactions",
        "pipeline": [
          { "$match": {
            "transactionDate": { "$gte": ISODate("2018-09-01"), "$lte": ISODate("2018-09-10") },
            "userId": userId, "partyName": customerName
          }},
          { "$project": {
            "_id":0, "dated": "$transactionDate", "heading": "Payment","particulars": "$transactionNumber",
            "amount": "$amount", "paymentMode": "$transactionMode"
          }}
        ],
        "as": "collection2"
      }}
    ]
  }},
  { "$project": {
    "data": {
      "$concatArrays": [
        { "$arrayElemAt": ["$collection1.collection1", 0] },
        { "$arrayElemAt": ["$collection2.collection2", 0] },
      ]
    }
  }},
  { "$unwind": "$data" },
  { "$replaceRoot": { "newRoot": "$data" } },
  { "$sort": { "dated": -1 }}
])

回答2:

I dont think so that you can perform a union on 2 different collections in mongodb.

However you can get data from both collections using q.all and afterwards make a union from them using your own function or may be third party module like lodash.

The only way you can perform union in mongodb is as mentioned here. https://docs.mongodb.com/manual/reference/operator/aggregation/setUnion/

回答3:

Disclaimer: The technique presented below is not exactly advisable. ;) This is true in particular when dealing with big collections. However, it can be used to achieve the identical effect as a SQL UNION ALL from MongoDB v3.6 onwards.

Given a collection first and a collection second:

db.first.aggregate([{
    $group: { // create an array to hold all documents of the first collection
        "_id": null,
        "first": {
            $push: "$$ROOT"
        }
    }
}, {
    $lookup: { // perform some kind of ridiculous lookup which will return all documents from the second collection in an array
        from: "second",
        let: { /* we do not need any variables */ },
        pipeline: [ { $match: { /* this filter will match every document */ } } ],
        as: "second"
    }
}, {
    $project: {
        "all": { $concatArrays: [ "$first", "$second" ] } // merge the two collections
    }
}, {
    $unwind: "$all" // flatten the resulting array
}, {
    $replaceRoot: { "newRoot": "$all" } // move content of "all" field all the way up 
}], {
    allowDiskUse: true // make sure we do not run into memory issues
})

来源：https://stackoverflow.com/questions/52715308/sql-union-all-like-implementation-in-mongodb

标签

node.js

mongodb

aggregation