您可以在单个查询中完成此操作，您只需要在操作文档方面有点创意，然后基本上执行 两个 $group 阶段，先按日期添加，然后按客户添加.
首先对于当前的 MongoDB 版本 2.6 及更高版本，由于使用了一些运算符：
db.transactions.aggregate([
// Project an additional array, stands for "order", "delivery"
{ "$project": {
"_id": 0,
"customer": 1,
"order_date": 1,
"delivery_date": 1,
"qty": 1,
"type": { "$literal": ["o","d"] }
}},
// Unwind that array, creates two documents by "type"
{ "$unwind": "$type" },
// Group by "customer", "type" and date
{ "$group": {
"_id": {
"customer": "$customer",
"type": "$type",
"year": {
"$year": {
"$cond": [
{ "$eq": [ "$type", "o" ] },
"$order_date",
"$delivery_date"
]
}
},
"month": {
"$month": {
"$cond": [
{ "$eq": [ "$type", "o" ] },
"$order_date",
"$delivery_date"
]
}
}
},
"qty": { "$sum": "$qty" }
}},
// Group on the "customer" selecting which array to add to
{ "$group": {
"_id": "$_id.customer",
"orders": {
"$push": {
"$cond": [
{ "$eq": [ "$_id.type", "o" ] },
{
"year": "$_id.year",
"month": "$_id.month",
"qty": "$qty"
},
false
]
}
},
"deliveries": {
"$push": {
"$cond": [
{ "$eq": [ "$_id.type", "d" ] },
{
"year": "$_id.year",
"month": "$_id.month",
"qty": "$qty"
},
false
]
}
}
}},
// Getting rid of the `false` values in there
{ "$project": {
"orders": { "$setDifference": [ "$orders", [false] ] },
"deliveries": { "$setDifference": [ "$deliveries", [false] ] },
}},
// But "sets" are not considered ordered, so sort them
{ "$unwind": "$orders" },
{ "$sort": { "orders.year": 1, "orders.month": 1 } },
{ "$group": {
"_id": "$_id",
"orders": { "$push": "$orders" },
"deliveries": { "$first": "$deliveries" }
}},
{ "$unwind": "$deliveries" },
{ "$sort": { "deliveries.year": 1, "deliveries.month": 1 } },
{ "$group": {
"_id": "$_id",
"orders": { "$first": "$orders" },
"deliveries": { "$push": "$deliveries" }
}}
)
只是在 2.6 之前的版本中这样做有点不同：
db.transactions.aggregate([
// Project an additional array, stands for "order", "delivery"
{ "$project": {
"_id": 0,
"customer": 1,
"order_date": 1,
"delivery_date": 1,
"qty": 1,
"type": { "$cond": [ 1, ["o","d"], 0 ] }
}},
// Unwind that array, creates two documents by "type"
{ "$unwind": "$type" },
// Group by "customer", "type" and date
{ "$group": {
"_id": {
"customer": "$customer",
"type": "$type",
"year": {
"$year": {
"$cond": [
{ "$eq": [ "$type", "o" ] },
"$order_date",
"$delivery_date"
]
}
},
"month": {
"$month": {
"$cond": [
{ "$eq": [ "$type", "o" ] },
"$order_date",
"$delivery_date"
]
}
}
},
"qty": { "$sum": "$qty" }
}},
// Group on the "customer" selecting which array to add to
{ "$group": {
"_id": "$_id.customer",
"orders": {
"$push": {
"$cond": [
{ "$eq": [ "$_id.type", "o" ] },
{
"year": "$_id.year",
"month": "$_id.month",
"qty": "$qty"
},
false
]
}
},
"deliveries": {
"$push": {
"$cond": [
{ "$eq": [ "$_id.type", "d" ] },
{
"year": "$_id.year",
"month": "$_id.month",
"qty": "$qty"
},
false
]
}
}
}},
// Filter `false` and sort on date
{ "$unwind": "$orders" },
{ "$match": { "orders": { "$ne": false } } },
{ "$sort": { "orders.year": 1, "orders.month": 1 } },
{ "$group": {
"_id": "$_id",
"orders": { "$push": "$orders" },
"deliveries": { "$first": "$deliveries" }
}},
{ "$unwind": "$deliveries" },
{ "$match": { "deliveries": { "$ne": false } } },
{ "$sort": { "deliveries.year": 1, "deliveries.month": 1 } },
{ "$group": {
"_id": "$_id",
"orders": { "$first": "$orders" },
"deliveries": { "$push": "$deliveries" }
}}
])
基本上总结一下这里的方法，你正在做的是复制每个文档并分配一个代表“订单”或“交付”的“类型”。然后，当您按“客户”和“日期”和“类型”分组时，您有条件地根据当前类型决定选择哪个“日期”，并总结该键下的“数量”。
由于结果是每个客户的“订单”和“交付”数组，因此您可以根据当前的“类型”有条件地将该数组$push 指定为文档值或false " 文档的每个数组。
最后，由于这些数组现在包含 false 的值以及所需的文档，因此您可以过滤掉这些值并确保您的数组处于正确的“日期”顺序（如果您确实需要）。
是的，列表有两个以上的$group 阶段，繁重的工作实际上是在两个分组中完成的，如果您需要，其他的只是用于数组操作，但它会为您提供准确和有序的结果。
因此，这可能不是您可能想到的第一种方法，但显示了一些有趣的转换想法，您可以将其与各种 aggregation operators 一起使用以解决问题。这是做什么的:)