Adding refs to another mongodb collection by matching multiple criteria - mongodb

Question

I'm totally new to MongoDB and can't sort this out from the docs. I have two collections, TREES and PLOTS. Every tree is in a plot; every plot has, say, 4-150 trees. So far the collections look something like this:
// TREES
[{"_id": objId(), "Tree": "1", "Project": "Alpha", "Plot": "A", "Year": 1979, "Size": 20},
{"_id": objId(), "Tree": "1", "Project": "Alpha", "Plot": "A", "Year": 1986, "Size": 21},
...
{"_id": objId(), "Tree": "54", "Project": "Omega", "Plot": "Z", "Year": 2016, "Size": 17}]
// PLOTS
[{"_id": objId(), "Plot": "A", "Project": "Alpha", "Year": 1979},
{"_id": objId(), "Plot": "A", "Project": "Alpha", "Year": 1986},
...
{"_id": objId(), "Plot": "Z", "Project": "Omega", "Year": 2016}]
I want to add a reference field to all the Trees with the objId of the appropriate Plot document, matching on Project, Plot and Year. I'd also like to add a refs array to all the Plots to contain the objIds with all of each one's Trees [Edit: Although maybe that's really not necessary?]. The real schemas both have 30-40 fields so embedding would be mad. The application development will most likely be done in pymongo, if there's anything relevant there.
To Clarify:
My problem is in matching trees to plots on the three criteria -- it seems like $lookup is no use here and I've tried $unionWith but can't figure it out. The docs and tutorials are full of toy problems where you add inter-collection references by matching on one field... and I can't figure out how to generalize that. Best result has been from doing
db.TREES.aggregate([
{
'$lookup': {
'from': 'PLOTS',
'localField': 'Plot',
'foreignField': 'Plot',
'as': 'TooManyPlots'
}
}
])
That gives me an array of all Plots with the right name -- but from all Years and all Projects, and I can't figure out how to weed that out and arrive at an updated Tree document with just the one correct Plot.
I haven't yet developed the Mongo-vision to see the proper flow for this.
Could be I'm also having some XY trouble -- plus it could be that MongoDB isn't the best fit for our project anyway. It seems worth a try though.

Answer 1

Okay -- no great mystery to it except for understanding how to understand it. In hopes of giving someone else a hand here's how I got it to work, closely matching the doc reference #D.SM noted in the comments, and my thanks to them.
db.TREES.aggregate(
[{$lookup: {
from: 'Plots',
let: { tPlot: '$Plot', tProj: '$Project', tYear: '$Year'},
pipeline: [
{ $match:
{ $expr:
{ $and:
[{$eq: ['$Plot', '$$tPlot']},
{$eq: ['$Project', '$$tProj']},
{$eq: ['$Year', '$$tYear']}]
}
}
},
{ $project: { _id:1} }
],
as: 'Plot_id_A'
} // That does what I wanted; from here it's just tidying things up
}, {$unwind: {
path: '$Plot_id_A',
}}, {$set: {
Plot_id: '$Plot_id_A._id'
}}, {$project: {
Plot_id_A : 0
}}]
Note that in the let: line the right-hand names are all coming from the TREES collection we're working in, while in the $and: the same-named left-hand names are all coming from the PLOTS collection we're joining, and so for the loss of clarity there I acknowlege D.SM's beef with my example field names. Anyway it works.
If anyone has an improvement to suggest though... lemme know!