Order by date in sub-document and then by document - mongodb

Question

I have a simple "Event" mongo schema. Two sample documents are below :
Event Document #1
{
"event_name": "Some nice event",
"venues": [
{
"venue_name": "venue #1",
"shows": [
{
"show_time": "2014-06-18T07:46:02.415Z",
"capacity": 20
},
{
"show_time": "2014-06-20T07:46:02.415Z",
"capacity": 40
}
]
},
{
"venue_name": "venue #2",
"shows": [
{
"show_time": "2014-06-17T07:46:02.415Z",
"capacity": 20
},
{
"show_time": "2014-06-24T07:46:02.415Z",
"capacity": 40
}
]
}
]
}
Event Document #2
{
"event_name": "Another nice event",
"venues": [
{
"venue_name": "venue #1",
"shows": [
{
"show_time": "2014-06-19T07:46:02.415Z",
"capacity": 20
},
{
"show_time": "2014-06-16T07:46:02.415Z",
"capacity": 40
}
]
}
]
}
I need to query this collection of event documents and fetch the events with the closest shows, with respective to a particular time.
So for e.g., if I had to find events happening on or after 16 Jun, I should get document #2 followed by document #1, with the venue sub-document order as [venue #2, venue #1].
On the other hand, if I wanted events happening on or after 18 Jun, I should get document #1, with [venue #1, venue #2], followed by document #2.
Essentially, I need to be able to sort by the start_time of the nested sub-document. And this sorting should work on multiple venue sub-documents.
According to mongo's documentation, this doesn't seem to be supported, so is there a way of using aggregation to achieve this?
Or is there a way to rejig the schema to support such queries?
Or is mongoDB the wrong use-case for such scenarios altogether?

Answer 1

Really good question. Hoping that your dates are real date but the lexical form should not really matter here. The following form should do it, as long as you take the dates into consideration:
db.event.aggregate([
// Match the "documents" that meet the condition
{ "$match": {
"venues.shows.show_time": { "$gte": new Date("2014-06-16") }
}},
// Unwind the arrays
{ "$unwind": "$venues" },
{ "$unwind": "$venues.shows" },
// Sort the entries just to float the nearest result
{ "$sort": { "venues.shows.show_time": 1 } },
// Find the "earliest" for the venue while grouping
{ "$group": {
"_id": {
"_id": "$_id",
"event_name": "$event_name",
"venue_name": "$venues.venue_name"
},
"shows": {
"$push": "$venues.shows"
},
"earliest": {
"$min": {
"$cond": [
{ "$gte": [
"$venues.shows.show_time",
new Date("2014-06-16")
]},
"$venues.shows.show_time",
null
]
}
}
}},
// Sort those because of the order you want
{ "$sort": { "earliest": 1 } },
// Group back and with the "earliest" document
{ "$group": {
"_id": "$_id._id",
"event_name": { "$first": "$_id.event_name" },
"venues": {
"$push": {
"venue_name": "$_id.venue_name",
"shows": "$shows"
}
},
"earliest": {
"$min": {
"$cond": [
{ "$gte": [
"$earliest",
new Date("2014-06-16")
]},
"$earliest",
null
]
}
}
}},
// Sort by the earliest document
{ "$sort": { "earliest": 1 } },
// Project the fields
{ "$project": {
"event_name": 1,
"venues": 1
}}
])
So most of this looks reasonable straightforward if you have some experience with the aggregation framework. If not then there is some general explaining, plus there are some "funky" things happening as we evaluate further.
The first steps in aggregation are to $match just like any normal query and then to $unwind the arrays you want to process. The "unwind" statement effectively "de-normalizes" the documents contained in the array to be standard documents by themselves.
The next $sort ends up as a "prettying up" function as the "earliest" event in each "set" will be at the top as a result.
As there are "two" levels of arrays, you do the grouping in two stages via the $group pipeline stage.
The first $group "groups" by "document", "event_name" and "venue". All of the shows are put back into their original array form, but at this time we extract the $min value for the "show_time".
The value taken is not just the ordinary "minimal" value. Here we use the $cond operator to make sure that the value returned must be "greater than or equal to" the date that you were requesting in the query initially. This makes sure that any "earlier" values are not taken into consideration when "sorting".
The next thing to do is to $sort on that "earliest" date, to keep the entries for the "venues" in order. The following stages then do the same as above, but "grouping" back to the original documents this time, then finally "sorting" in the order of which "show_time" would be the "earliest".
The result from the dates shown as input would be your desired result for the 16th:
{
"_id" : ObjectId("53a95263a1923f45a6c2d3dd"),
"event_name" : "Another nice event",
"venues" : [
{
"venue_name" : "venue #1",
"shows" : [
{
"show_time" : ISODate("2014-06-16T07:46:02.415Z"),
"capacity" : 40
},
{
"show_time" : ISODate("2014-06-19T07:46:02.415Z"),
"capacity" : 20
}
]
}
]
}
{
"_id" : ObjectId("53a952b5a1923f45a6c2d3de"),
"event_name" : "Some nice event",
"venues" : [
{
"venue_name" : "venue #2",
"shows" : [
{
"show_time" : ISODate("2014-06-17T07:46:02.415Z"),
"capacity" : 20
},
{
"show_time" : ISODate("2014-06-24T07:46:02.415Z"),
"capacity" : 40
}
]
},
{
"venue_name" : "venue #1",
"shows" : [
{
"show_time" : ISODate("2014-06-18T07:46:02.415Z"),
"capacity" : 20
},
{
"show_time" : ISODate("2014-06-20T07:46:02.415Z"),
"capacity" : 40
}
]
}
]
}
And by changing the input to the 18th you also get the desired result:
{
"_id" : ObjectId("53a952b5a1923f45a6c2d3de"),
"event_name" : "Some nice event",
"venues" : [
{
"venue_name" : "venue #1",
"shows" : [
{
"show_time" : ISODate("2014-06-18T07:46:02.415Z"),
"capacity" : 20
},
{
"show_time" : ISODate("2014-06-20T07:46:02.415Z"),
"capacity" : 40
}
]
},
{
"venue_name" : "venue #2",
"shows" : [
{
"show_time" : ISODate("2014-06-17T07:46:02.415Z"),
"capacity" : 20
},
{
"show_time" : ISODate("2014-06-24T07:46:02.415Z"),
"capacity" : 40
}
]
}
]
}
{
"_id" : ObjectId("53a95263a1923f45a6c2d3dd"),
"event_name" : "Another nice event",
"venues" : [
{
"venue_name" : "venue #1",
"shows" : [
{
"show_time" : ISODate("2014-06-16T07:46:02.415Z"),
"capacity" : 40
},
{
"show_time" : ISODate("2014-06-19T07:46:02.415Z"),
"capacity" : 20
}
]
}
]
}
Also if you want to go further with this, just add an additional $match stage, and that can filter out "events" that occur before the date that is requested in the query:
db.event.aggregate([
{ "$match": {
"venues.shows.show_time": { "$gte": new Date("2014-06-18") }
}},
{ "$unwind": "$venues" },
{ "$unwind": "$venues.shows" },
{ "$match": {
"venues.shows.show_time": { "$gte": new Date("2014-06-18") }
}},
{ "$sort": { "venues.shows.show_time": 1 } },
{ "$group": {
"_id": {
"_id": "$_id",
"event_name": "$event_name",
"venue_name": "$venues.venue_name"
},
"shows": {
"$push": "$venues.shows"
},
"earliest": {
"$min": {
"$cond": [
{ "$gte": [
"$venues.shows.show_time",
new Date("2014-06-18")
]},
"$venues.shows.show_time",
null
]
}
}
}},
{ "$sort": { "earliest": 1 } },
{ "$group": {
"_id": "$_id._id",
"event_name": { "$first": "$_id.event_name" },
"venues": {
"$push": {
"venue_name": "$_id.venue_name",
"shows": "$shows"
}
},
"earliest": {
"$min": {
"$cond": [
{ "$gte": [
"$earliest",
new Date("2014-06-18")
]},
"$earliest",
null
]
}
}
}},
{ "$sort": { "earliest": 1 } },
{ "$project": {
"event_name": 1,
"venues": 1
}}
])
With the result:
{
"_id" : ObjectId("53a952b5a1923f45a6c2d3de"),
"event_name" : "Some nice event",
"venues" : [
{
"venue_name" : "venue #1",
"shows" : [
{
"show_time" : ISODate("2014-06-18T07:46:02.415Z"),
"capacity" : 20
},
{
"show_time" : ISODate("2014-06-20T07:46:02.415Z"),
"capacity" : 40
}
]
},
{
"venue_name" : "venue #2",
"shows" : [
{
"show_time" : ISODate("2014-06-24T07:46:02.415Z"),
"capacity" : 40
}
]
}
]
}
{
"_id" : ObjectId("53a95263a1923f45a6c2d3dd"),
"event_name" : "Another nice event",
"venues" : [
{
"venue_name" : "venue #1",
"shows" : [
{
"show_time" : ISODate("2014-06-19T07:46:02.415Z"),
"capacity" : 20
}
]
}
]
}