$push and $set same sub-document in an Array - node.js

Question

I'm trying to keep an history of states in a subdocument array with mongoosejs 4.9.5 and mongo 3.2.7
Example of document structure:
company (Schema)
employees (Schema): [ ]
currentState: String
states (Schema): [ ]
state: String
starts: Date
ends: Date
When I change the employee state, I want to change the currentState, add the new state into the states array, and update the last state for define the 'ends' timestamp
// I get the last state position from a previous find request
var lastStateIndex = employee.stateHistory.length - 1;
var changeStateDate = new Date();
// Prepare the update
var query = { _id: companyId, "employees._id": employeeId };
var update = {
$set: {
"employees.$.state": newState,
`employees.$.stateHistory.${lastStateIndex}.ends`: changeStateDate
},
$push: {
"employees.$.stateHistory": {
state: newState,
starts: changeStateDate
}
}
}
Company.findOneAndUpdate(query, update, { multi:false, new:true}, ... )
Mongo is returning the following error
{"name":"MongoError","message":"Cannot update 'employees.0.stateHistory.0.ends' and 'employees.0.stateHistory' at the same time","ok":0,"errmsg":"Cannot update 'employees.0.stateHistory.0.ends' and 'employees.0.stateHistory' at the same time","code":16837}
Any suggestions how to avoid running two updates for that purpose?
Any work around for avoid storing the 'ends' date, but being able to calculate it after based on the 'starts' of the next item in the array?
Thank you,

Answer 1

I expected this to already be answered elsewhere, but no other reasonable response seems to exist. As commented, you cannot actually do this in a single update operation because the operations "conflict" on the same path. But .bulkWrite() allows "multiple updates" to be applied in a single request and response.
Company.bulkWrite([
{ "updateOne": {
"filter": { "_id": companyId, "employees._id": employeeId },
"update": {
"$set": {
"employees.$.state": newState,
[`employees.$.stateHistory.${lastStateIndex}.ends`]: changeStateDate
}
}},
{ "updateOne": {
"filter": { "_id": companyId, "employees._id": employeeId },
"update": {
"$push": {
"employees.$.stateHistory": {
"state": newState,
"starts": changeStateDate
}
}
}
}}
])
Now of course .bulkWrite() does not return the "modified document" like .findOneAndUpdate() does. So if you need to actually return the document, then you need to add to the Promise chain instead:
Company.bulkWrite([
{ "updateOne": {
"filter": { "_id": companyId, "employees._id": employeeId },
"update": {
"$set": {
"employees.$.state": newState,
[`employees.$.stateHistory.${lastStateIndex}.ends`]: changeStateDate
}
}},
{ "updateOne": {
"filter": { "_id": companyId, "employees._id": employeeId },
"update": {
"$push": {
"employees.$.stateHistory": {
"state": newState,
"starts": changeStateDate
}
}
}
}}
]).then( result => {
// maybe inspect the result
return Company.findById(companyId);
})
Of course noting that it is "possible" that another modification can be made to the document in between when the .bulkWrite() is applied and the .findById() is executed. But that is the cost of the operation you are doing.
It is generally best to consider if you actually need the returned document or not. In most instances you simply already have the information and any "updates" you should be aware of because you are "issuing them", and if you want "truly reactive" then you should be listening for other change events on the data through a socket instead.
Note you could simply "chain" the "multiple" .findOneAndUpdate() calls, but this is indeed "multiple" calls and responses from the server, as opposed to the one using .bulkWrite(). So there really isn't anything to gain by doing otherwise.