Using Fawn
The issue is one of usage with the Fawn library and comes from some misconceptions about the naming of mongoose models and how these interact with the library itself. As such the best way to demonstrate is with a minimal example of working code:
const { Schema } = mongoose = require('mongoose');
const Fawn = require('fawn');
const uri = 'mongodb://localhost:27017/fawndemo';
const opts = { useNewUrlParser: true };
// sensible defaults
mongoose.Promise = global.Promise;
mongoose.set('debug', true);
mongoose.set('useFindAndModify', false);
mongoose.set('useCreateIndex', true);
// schema defs
const oneSchema = new Schema({
name: String
});
const twoSchema = new Schema({
counter: Number
});
// don't even need vars since we access model by name
mongoose.model('One', oneSchema);
mongoose.model('Two', twoSchema);
// log helper
const log = data => console.log(JSON.stringify(data, undefined, 2));
(async function() {
try {
const conn = await mongoose.connect(uri, opts);
// init fawm
Fawn.init(mongoose);
// Clean models
await Promise.all(
Object.entries(conn.models).map(([k,m]) => m.deleteMany())
)
// run test
let task = Fawn.Task();
let results = await task
.save('One', { name: 'Bill' })
.save('Two', { counter: 0 })
.update('Two', { }, { "$inc": { "counter": 1 } })
.run({ useMongoose: true });
log(results);
// List objects in models
for ( [k,m] of Object.entries(conn.models) ) {
let result = await m.find();
log(result);
}
} catch(e) {
console.error(e)
} finally {
mongoose.disconnect()
}
})()
Note how the mongoose models are registered here:
mongoose.model('One', oneSchema);
mongoose.model('Two', twoSchema);
That first argument is the registered name which mongoose uses for the model in it's internal logic. From the perspective of mongoose itself, once you have registered the model name with the schema as above, you can actually call an instance of the model as follows:
const One = mongoose.model('One');
Typically people export
the result of the initial registration and then just use the returned value which is a reference to mongoose's own internal storage of the model details and attached schema. But the line of code is equivalent to that same thing as long as the registration code has already been run.
A typical exports
considering this can therefore be used as:
require('./models/one');
require('./models/two');
let results = await mongoose.model('One').find();
So you might not see that often in other code examples, but that is really to show what is actually happening from the perspective of the Fawn library with later code.
With that knowledge you can consider the following code in the listing:
let task = Fawn.Task();
let results = await task
.save('One', { name: 'Bill' })
.save('Two', { counter: 0 })
.update('Two', { }, { "$inc": { "counter": 1 } })
.run({ useMongoose: true });
Here the methods of update()
and save()
familiar to mongoose and MongoDB users actually have a different first argument specific to their implementation on the Fawn.Task()
result. That first argument is the "registered model name" for mongoose, which is what we just explained with the previous example.
What the Fawn library is actually doing is calling similar code to:
mongoose.model('One').save({ name: 'Bill' })
Well actually it's doing something a lot more complicated than that as is evidenced in the output of the example listing. It's actually doing a lot of other things related to two phase commits and writing temporary entries in another collection and eventually moving those over to the target collections. But when it does actually go to the collections for the registered models, then that is basically how it is doing it.
So the core issue in the code in the question is that you are not using the names that were actually registered to the mongoose models, and a few other things are missing from the documentation steps.
You're also not awaiting asynchronous functions correctly, and the try..catch
within the question code is not doing anything with calls in this context. The listing here however demonstrates how to do that correctly using async/await
.
You can alternately just use the native Promise.then(...).catch(...)
aproach if your NodeJS version does not have async/await
support, but there really is little other change than doing that and of course removing the try..catch
since promises in that form will ignore it. Which is why you catch()
instead.
NOTE - With some brief testing there appear to be a number of things which are supported mongoose/mongodb features which are not actually implemented and supported on this library's methods. Notably "upserts" was a prime example of a useful and common thing which the "two phase commit" system implemented here does not appear to support at all.
This partly seems an oversight in the code of the library where certain "options" to the methods are actually being ignored or stripped completely. This is a concern for getting the most out of MongoDB features.
Transactions
The whole usage of this library though at least seems suspicious to me that you picked it up because you "thought" this was "Transactions". Put plainly the two phase commit is NOT a transaction. Furthermore the implementation of any attempt at such control and rollback etc seem very loose at best.
If you have a modern MongoDB 4.0 server or above, and where you actually configured it to be named as a "replica set" ( which you can also do for a single member, where a common misconception is you need more than one ) then there is support for real transactions, and they are very easy to implement:
const { Schema } = mongoose = require('mongoose');
const uri = 'mongodb://localhost:27017/trandemo';
const opts = { useNewUrlParser: true };
// sensible defaults
mongoose.Promise = global.Promise;
mongoose.set('debug', true);
mongoose.set('useFindAndModify', false);
mongoose.set('useCreateIndex', true);
// schema defs
const orderSchema = new Schema({
name: String
});
const orderItemsSchema = new Schema({
order: { type: Schema.Types.ObjectId, ref: 'Order' },
itemName: String,
price: Number
});
const Order = mongoose.model('Order', orderSchema);
const OrderItems = mongoose.model('OrderItems', orderItemsSchema);
// log helper
const log = data => console.log(JSON.stringify(data, undefined, 2));
// main
(async function() {
try {
const conn = await mongoose.connect(uri, opts);
// clean models
await Promise.all(
Object.entries(conn.models).map(([k,m]) => m.deleteMany())
)
let session = await conn.startSession();
session.startTransaction();
// Collections must exist in transactions
await Promise.all(
Object.entries(conn.models).map(([k,m]) => m.createCollection())
);
let [order] = await Order.create([{ name: 'Bill' }], { session });
let items = await OrderItems.insertMany(
[
{ order: order._id, itemName: 'Cheese', price: 1 },
{ order: order._id, itemName: 'Bread', price: 2 },
{ order: order._id, itemName: 'Milk', price: 3 }
],
{ session }
);
// update an item
let result1 = await OrderItems.updateOne(
{ order: order._id, itemName: 'Milk' },
{ $inc: { price: 1 } },
{ session }
);
log(result1);
// commit
await session.commitTransaction();
// start another
session.startTransaction();
// Update and abort
let result2 = await OrderItems.findOneAndUpdate(
{ order: order._id, itemName: 'Milk' },
{ $inc: { price: 1 } },
{ 'new': true, session }
);
log(result2);
await session.abortTransaction();
/*
* $lookup join - expect Milk to be price: 4
*
*/
let joined = await Order.aggregate([
{ '$match': { _id: order._id } },
{ '$lookup': {
'from': OrderItems.collection.name,
'foreignField': 'order',
'localField': '_id',
'as': 'orderitems'
}}
]);
log(joined);
} catch(e) {
console.error(e)
} finally {
mongoose.disconnect()
}
})()
That is really just a simple listing with the class Order
and related OrderItems
. There really is nothing special in the code and you should see that it's basically the same as most listing examples you will see with a few small changes.
Notably we initialize a session
and also session.startTransaction()
as an indicator that a transaction should be in progress. Note that session
would generally have a wider scope where you would typically re-use that object for more than just a few operations.
Now you have session
and the transaction is started, this is simply added to the "options" of the various statements being executed:
let [order] = await Order.create([{ name: 'Bill' }], { session });
let items = await OrderItems.insertMany(
[
{ order: order._id, itemName: 'Cheese', price: 1 },
{ order: order._id, itemName: 'Bread', price: 2 },
{ order: order._id, itemName: 'Milk', price: 3 }
],
{ session }
);
Admittedly this is a brief example that does not fully cover all write error possibilities and how to handle that within separate try..catch
blocks. But as a very basic example should any error occur before the session.commitTransaction()
is called, then none of the operations since the transaction was started will actually be persisted within the session.
Also there is "causal consistency" in that once a normal write acknowledgement has been confirmed, then within the scope of the session the data appears written to the respective collections right up until the transaction commit or rollback.
In the event of a rollback ( as demonstrated in the final operation ):
// Update and abort
let result2 = await OrderItems.findOneAndUpdate(
{ order: order._id, itemName: 'Milk' },
{ $inc: { price: 1 } },
{ 'new': true, session }
);
log(result2);
await session.abortTransaction();
These writes though reported to be made as seen in the operation result, are indeed "rolled back" and further operations see the state of the data before these changes were made.
The full example code demonstrates this by adding the items with another update action in one transaction, then beginning another to alter data and read it then abort the transaction. The final data state shows of course only what was actually committed.
NOTE Operations like find()
and findOne()
or anything that retrieves data must include the session
whilst a transaction is active in order to see the current state, just in the same way that write operations are doing as shown in the listing.
Without including the session
, these changes in state are not visible in the "global" scope until the transaction is resolved.
Listing Outputs
Code listings given produce the following output when run, for reference.
fawndemo
Mongoose: ones.deleteMany({}, {})
Mongoose: twos.deleteMany({}, {})
Mongoose: ojlinttaskcollections.deleteMany({}, {})
Mongoose: ojlinttaskcollections.insertOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a"), steps: [ { dataStore: [], _id: ObjectId("5bf765f7e5c71c5fae77030d"), index: 0, type: 'save', state: 0, name: 'One', data: { name: 'Bill' } }, { dataStore: [], _id: ObjectId("5bf765f7e5c71c5fae77030c"), index: 1, type: 'save', state: 0, name: 'Two', data: { counter: 0 } }, { dataStore: [], _id: ObjectId("5bf765f7e5c71c5fae77030b"), index: 2, type: 'update', state: 0, name: 'Two', data: { '*_**ojlint**escape$*__tx__00***___string$inc': { counter: 1 } } } ], __v: 0 })
Mongoose: ojlinttaskcollections.updateOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") }, { '$set': { 'steps.0.state': 1 } })
Mongoose: ones.insertOne({ _id: ObjectId("5bf765f7e5c71c5fae77030e"), name: 'Bill', __v: 0 })
Mongoose: ojlinttaskcollections.updateOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") }, { '$set': { 'steps.0.state': 2 } })
Mongoose: ojlinttaskcollections.updateOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") }, { '$set': { 'steps.1.state': 1 } })
Mongoose: twos.insertOne({ _id: ObjectId("5bf765f7e5c71c5fae77030f"), counter: 0, __v: 0 })
Mongoose: ojlinttaskcollections.updateOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") }, { '$set': { 'steps.1.state': 2 } })
Mongoose: twos.find({})
Mongoose: ojlinttaskcollections.updateOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") }, { '$set': { 'steps.2.state': 1 } })
Mongoose: twos.update({}, { '$inc': { counter: 1 } }, {})
(node:24494) DeprecationWarning: collection.update is deprecated. Use updateOne, updateMany, or bulkWrite instead.
Mongoose: ojlinttaskcollections.updateOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") }, { '$set': { 'steps.2.state': 2 } })
Mongoose: ojlinttaskcollections.deleteOne({ _id: ObjectId("5bf765f7e5c71c5fae77030a") })
[
{
"_id": "5bf765f7e5c71c5fae77030e",
"name": "Bill",
"__v": 0
},
{
"_id": "5bf765f7e5c71c5fae77030f",
"counter": 0,
"__v": 0
},
{
"n": 1,
"nModified": 1,
"opTime": {
"ts": "6626877488230301707",
"t": 139
},
"electionId": "7fffffff000000000000008b",
"ok": 1,
"operationTime": "6626877488230301707",
"$clusterTime": {
"clusterTime": "6626877488230301707",
"signature": {
"hash": "AAAAAAAAAAAAAAAAAAAAAAAAAAA=",
"keyId": 0
}
}
}
]
Mongoose: ones.find({}, { projection: {} })
[
{
"_id": "5bf765f7e5c71c5fae77030e",
"name": "Bill",
"__v": 0
}
]
Mongoose: twos.find({}, { projection: {} })
[
{
"_id": "5bf765f7e5c71c5fae77030f",
"counter": 1,
"__v": 0
}
]
Mongoose: ojlinttaskcollections.find({}, { projection: {} })
[]
transdemo
Mongoose: orders.deleteMany({}, {})
Mongoose: orderitems.deleteMany({}, {})
Mongoose: orders.insertOne({ _id: ObjectId("5bf7661c3f60105fe48d076e"), name: 'Bill', __v: 0 }, { session: ClientSession("e146c6074bb046faa7b70ed787e1a334") })
Mongoose: orderitems.insertMany([ { _id: 5bf7661c3f60105fe48d076f, order: 5bf7661c3f60105fe48d076e, itemName: 'Cheese', price: 1, __v: 0 }, { _id: 5bf7661c3f60105fe48d0770, order: 5bf7661c3f60105fe48d076e, itemName: 'Bread', price: 2, __v: 0 }, { _id: 5bf7661c3f60105fe48d0771, order: 5bf7661c3f60105fe48d076e, itemName: 'Milk', price: 3, __v: 0 } ], { session: ClientSession("e146c6074bb046faa7b70ed787e1a334") })
Mongoose: orderitems.updateOne({ order: ObjectId("5bf7661c3f60105fe48d076e"), itemName: 'Milk' }, { '$inc': { price: 1 } }, { session: ClientSession("e146c6074bb046faa7b70ed787e1a334") })
{
"n": 1,
"nModified": 1,
"opTime": {
"ts": "6626877647144091652",
"t": 139
},
"electionId": "7fffffff000000000000008b",
"ok": 1,
"operationTime": "6626877647144091652",
"$clusterTime": {
"clusterTime": "6626877647144091652",
"signature": {
"hash": "AAAAAAAAAAAAAAAAAAAAAAAAAAA=",
"keyId": 0
}
}
}
Mongoose: orderitems.findOneAndUpdate({ order: ObjectId("5bf7661c3f60105fe48d076e"), itemName: 'Milk' }, { '$inc': { price: 1 } }, { session: ClientSession("e146c6074bb046faa7b70ed787e1a334"), upsert: false, remove: false, projection: {}, returnOriginal: false })
{
"_id": "5bf7661c3f60105fe48d0771",
"order": "5bf7661c3f60105fe48d076e",
"itemName": "Milk",
"price": 5,
"__v": 0
}
Mongoose: orders.aggregate([ { '$match': { _id: 5bf7661c3f60105fe48d076e } }, { '$lookup': { from: 'orderitems', foreignField: 'order', localField: '_id', as: 'orderitems' } } ], {})
[
{
"_id": "5bf7661c3f60105fe48d076e",
"name": "Bill",
"__v": 0,
"orderitems": [
{
"_id": "5bf7661c3f60105fe48d076f",
"order": "5bf7661c3f60105fe48d076e",
"itemName": "Cheese",
"price": 1,
"__v": 0
},
{
"_id": "5bf7661c3f60105fe48d0770",
"order": "5bf7661c3f60105fe48d076e",
"itemName": "Bread",
"price": 2,
"__v": 0
},
{
"_id": "5bf7661c3f60105fe48d0771",
"order": "5bf7661c3f60105fe48d076e",
"itemName": "Milk",
"price": 4,
"__v": 0
}
]
}
]