Strongloop UUID - loopbackjs

heres a snippet of the model I've created. I would like the id property to be ALWAYS automatically generated via the uuid default function. However, when even added to hidden the id property could be decide by the user if they pass that into the post create method.
Other than using the method hooks OR operation hooks, to always overwrite the value is there another Loopback approach or flag (I have tried the foreId property as well) to ensure that the id is always a uuid even when the user provides a value? If not then for my case, "defaultFn": "uuid" completely pointless, if I'm always overwriting it.
{
"name": "Account",
"base": "PersistedModel",
"strict": true,
"idInjection": false,
"options": {
"validateUpsert": true
},
"properties": {
"id": {
"id": true,
"required": true,
"type": "string",
"defaultFn": "uuid"
},
},
"validations": [],
"relations": {},
"acls": [],
"methods": {}
}

For some reason, the id can be redefined by the user through the REST api and is not protected when using idInjection: false and defining a custom id property.
I'll open a ticket on github if it's not already done, in the meantime you can fix it easily using a before save operation hook
In Account.js, using the uuid package
const uuidV4 = require('uuid/v4');
module.exports = function(Account) {
Account.observe('before save', function(ctx, cb){
if (ctx.instance) {
ctx.instance.id = uuidV4();
} else {
ctx.data.id = uuidV4();
}
cb();
});
};

Related

Loopback custom connector implementation

I am trying to implement a custom loopback connector and it's not clear to me how this all works.
Here are my models:
{
"customer": {
"dataSource": "qb",
"public": false
},
"company": {
"dataSource": "qb",
"public": true
},
"payment": {
"dataSource": "qb",
"public": false
},
"invoice": {
"dataSource": "qb",
"public": false
}
}
The most important part to the model (and to save space) is
{
"relations": {
"company": {
"type": "belongsTo",
"model": "company",
"foreignKey": "id",
"primaryKey": "id"
}
}
}
And, in company.json
{
"name": "company",
"plural": "companies",
"base": "Model",
"idInjection": true,
"options": {
"validateUpsert": true
},
"properties": {
"id": {
"type": "string",
"required": true
}
},
"validations": [],
"relations": {
"customers": {
"type": "hasMany",
"model": "customer",
"foreignKey": "customerId"
},
"payments": {
"type": "hasMany",
"model": "payment",
"foreignKey": "customerId"
},
"invoices": {
"type": "hasMany",
"model": "customer",
"foreignKey": "customerId"
}
},
"acls": [],
"methods": {}
}
which, as expected, produces URLs like:
/companies/${id}/customers/${fk}
So, I try the swagger UI and submit: GET /companies/4620816365214377730/customers/456
The problem I have is now 2 fold:
It calls the all function on my connector every time - right away, that doesn't make sense. I've given it 2 specific ID's why would it possible want all of anything?
I managed the above and produced the results asked, but then loopback reports a 404:
{
"error": {
"statusCode": 404,
"name": "Error",
"message": "could not find a model with id 4620816365214377730",
"code": "MODEL_NOT_FOUND",
"stack": "Error: could not find a model with id 4620816365214377730"
}
}
So, I definitely don't get it - the first param in callback is the err, and the second is the result. I have literally hardcoded it to be right (I think)
How do I implement simple CRUD? Why does it not call my findById function? I have breakpoints everywhere
const {Connector: connector} = require('loopback-connector')
const util = require("util");
exports.initialize = function initializeDataSource(dataSource, callback) {
dataSource.connector = new QbConnector(dataSource.settings);
dataSource.connector.dataSource = dataSource;
};
exports.QbConnector = QbConnector
function QbConnector(settings, datasource) {
connector.call(this, 'quickbooks', settings)
this.datasource = datasource
this.client = require(`./qb`)(require('./axios'))
}
util.inherits(QbConnector, connector);
// connector.defineAliases(QbConnector.prototype, 'find', 'findById');
QbConnector.prototype.create = function(data, callback) {
console.log()
}
QbConnector.prototype.replaceOrCreate = function(model, data, options, cb) {
console.log()
}
QbConnector.prototype.findOne = function (filter,cb) {
console.log()
}
QbConnector.prototype.all = function(model, filter, callback) {
this.client[model]?.get(filter.where.id)
?.then(data => callback(null,{id: filter.where.id}))
?.catch(e => callback(JSON.stringify(e.response.data,null,4)))
}
QbConnector.prototype.count = function (whereClause,callback) {
console.log()
}
QbConnector.prototype.save = function(model, data, options, cb) {
console.log()
}
QbConnector.prototype.findById = function (id, filter, options) {
console.log()
}
When I step into the callback it's definition is a guaranteed error (the message I am seeing)
(function anonymous(number, plural, select, pluralFuncs, fmt
) {
return function(d) { return "could not find a model with id " + d["0"]; }
})

Loopback inserts object id values as sequence and not random - how can i change it?

Loopback inserts object id values for my MongoDB as sequence and not random.
for security reason, how can i change it to be random by default for all my models?
Will it affect the DB performance?
You can change the way the ID is generated in the description file of your model:
https://loopback.io/doc/en/lb3/Model-definition-JSON-file.html#properties
model.json:
{
"name": "model",
"base": "PersistedModel",
"strict": true,
"idInjection": false, // disable default id!
"properties": {
"id": {
"id": true,
"required": true,
"type": "string",
"defaultFn": "uuid" // "guid" / "uuid" / "uuidv4" / "now"
},
},
"validations": [],
"relations": {},
"acls": [],
"methods": {}
}

Loopback setting a property that somehow doesn't persist to the database sql?

We're porting our api from C# to Loopback ^v3.19.0 and have run into a blocker.
Many of our models have shared properties, so we've created a base model "Base" which they inherit from.
{
"name": "Base",
"base": "PersistedModel",
"idInjection": true,
"options": {
"validateUpsert": true
},
"mixins": {
"Timestamp": {}
},
"properties": {
"created-by": {
"type": "number",
"postgresql": {
"columnName": "created_by"
}
},
"created-date": {
"type": "date",
"postgresql": {
"columnName": "created_on_utc"
}
},
"updated-by": {
"type": "number",
"postgresql": {
"columnName": "updated_by"
}
},
"updated-date": {
"type": "date",
"postgresql": {
"columnName": "updated_on_utc"
}
},
"soft-deleted": {
"type": "boolean",
"postgresql": {
"columnName": "is_deleted"
}
},
"deleted-by": {
"type": "number",
"postgresql": {
"columnName": "deleted_by"
}
},
"deleted-date": {
"type": "date",
"postgresql": {
"columnName": "deleted_on_utc"
}
},
"tenant-id": {
"type": "number",
"postgresql": {
"columnName": "tenant_id"
}
}
},
...
}
Inside the Timestamp mixin (our own), those properties get set accordingly
module.exports = function(Model, options) {
Model.observe('before save', function event(ctx, next) {
const token = ctx.options && ctx.options.accessToken;
const userId = token && token.userId;
const now = new Date().toISOString();
if (ctx.instance) {
ctx.instance['created-by'] = userId;
ctx.instance['created-date'] = now;
ctx.instance['updated-by'] = userId;
ctx.instance['updated-date'] = now;
} else {
if (ctx.data['soft-deleted'] &&
ctx.data['soft-deleted'] === true) {
ctx.data['deleted-by'] = userId;
ctx.data['deleted-date'] = now;
ctx.data['is-active'] = false;
}
ctx.data['updated-by'] = userId;
ctx.data['updated-date'] = now;
}
next();
});
};
This works great when creating a new model. It was working great for updates (PATCH /modelname/:id), but unexpectedly broke and we can't figure out why. (This is consistent across all the models that inherit from this Base model.)
The mixin correctly sees the model and adds the updated properties like so
LoopbackJS | ************* 'before save' ctx.data **************
LoopbackJS | { 'is-active': false,
LoopbackJS | 'updated-by': 1,
LoopbackJS | 'updated-date': '2018-08-16T17:57:23.660Z' }
LoopbackJS | ************* END 'before save' ctx.data **************
But when loopback executes the update SQL, it somehow omits/removes the value for updated-by? (2nd param should be 1, not null)
LoopbackJS | 2018-08-16T17:57:23.666Z loopback:connector:postgresql SQL: UPDATE "public"."asset_types" SET "is_active"=$1,"updated_by"=$2,"updated_on_utc"=$3::TIMESTAMP WITH TIME ZONE,"tenant_id"=$4 WHERE "id"=$5
LoopbackJS | Parameters: [false,null,"2018-08-16T17:57:23.660Z",1,5]
updated_by in Postgres is nullable, so that shouldn't generate an error... but Loopback is sending a stringified function?
LoopbackJS | 2018-08-16T18:04:12.522Z loopback:connector:postgresql error: invalid input syntax for integer: "function () { [native code] }"
LoopbackJS | at Connection.parseE (/home/src/back-end/node_modules/pg/lib/connection.js:553:11)
LoopbackJS | at Connection.parseMessage (/home/src/back-end/node_modules/pg/lib/connection.js:378:19)
LoopbackJS | at TLSSocket.<anonymous> (/home/src/back-end/node_modules/pg/lib/connection.js:119:22)
LoopbackJS | at emitOne (events.js:115:13)
LoopbackJS | at TLSSocket.emit (events.js:210:7)
LoopbackJS | at addChunk (_stream_readable.js:264:12)
LoopbackJS | at readableAddChunk (_stream_readable.js:251:11)
LoopbackJS | at TLSSocket.Readable.push (_stream_readable.js:209:10)
LoopbackJS | at TLSWrap.onread (net.js:587:20)
If we don't touch the updated_by column, the SQL is correct and updates.
Incidentally, if we soft-delete and the deleted_by column is in play, the same thing happens there.
Feels like I'm spinning in circles here and probably overlooking something basic. Any suggestions?
EDIT
So it appears that it's not limited to a mixin... when we remove it completely and manually set the k:v pair in the payload (ie 'created-by': 1) we still get the same error back from Postgres.
The root cause of this was due to incorrect relationships.
I created this as a gist, but pasting it here too in case it helps someone else.
It's a PostgreSQL best-practice to use lowercase names, using snakecase if you need to. ie, my_column_name.
Also, since I'm using a JSON API client, I've installed the excellent loopback-component-jsonapi to handle the de/serialization stuff... but that just added additional complexities.
JSON API calls for dasherized property names. When you start with something like my-property-name, Loopback or the PostgreSQL driver (doesn't really matter) collapses the dasherized property down to mypropertyname by default.
This is bad... especially when you have an existing schema you're working with.
It's worse when you're working with relationships, because Loopback also appends the id suffix by default, so now you have issues unless you happen to have a mypropertynameid column.
An example
Let's say we have a Customer model. I needed endpoints that are lowercase (and dasherized, where applicable), so just change the plural to match here.
{
"name": "Customer",
"plural": "customers",
"base": "PersistedModel",
...
}
Inside of options.postgresql, you can set a tableName. Loopback will use the name value by default, but remember PostgreSQL doesn't like CamelCase. You need to override this unless you use lowercase model names.
(It's a religious preference, but I like my tables to be plurals. Fight me.)
{
...
"options": {
"validateUpsert": true,
"postgresql": {
"tableName": "customers"
}
}
...
}
Back to the properties, use the postgresql.columnName property to map to the correct column name in the db. If it's not a dasherized property name (ie status) then you can ignore the postgresql.columnName bit.
{
...
"properties": {
"is-active": {
"type": "boolean",
"default": false,
"postgresql": {
"columnName": "is_active"
}
}
}
}
Relationships can be a headache.
Let's say our Customer has people who work there. To do a basic one-many relationship between the models...
{
...
"relations": {
"people": {
"type": "hasMany",
"model": "Person",
"foreignKey": "customer_id"
}
},
...
}
people is the name of the relationship element of the JSON API payload.
A "gotcha" here for me was the foreignKey property.
The Loopback docs say it's optional - and it is - but if you leave it out then it adds the id suffix to the name (people) and then looks for that column in your customers table. That wasn't highlighted very well, but it was clear enough.
This part wasn't clear => I originally thought the foreignKey value pointed to the property of the Person model, so I had the dasherized customer-id property here. That's incorrect. It's literally asking you for the database column name, which feels like a bit of an antipattern... In the properties you had to define a columnName if you wanted to refer to the db columns under the ORM.
Also, note that the foreignKey property is reused in relationships but it means different things to different type contexts. In a hasMany, it's asking "Which column there maps to the primary key here?"
Final Customer model:
{
"name": "Customer",
"plural": "customers",
"base": "PersistedModel",
"options": {
"validateUpsert": true,
"postgresql": {
"tableName": "customers"
}
},
"properties": {
"name": {
"type": "string"
},
"is-active": {
"type": "boolean",
"default": false,
"postgresql": {
"columnName": "is_active"
}
}
},
"validations": [],
"relations": {
"people": {
"type": "hasMany",
"model": "Person",
"foreignKey": "customer_id"
}
},
"acls": [],
"methods": {}
}
The Person model on the other end of the relationship.
The foreignKey for a belongsTo relationship is asking the opposite question... "Which property here maps to the primary key there?"
Also, if you have properties you don't want exposed (especially if you've inherited a model and don't want/need all those properties for whatever reason) then you can hide them with the hidden element. See below.
{
"name": "Person",
"plural": "people",
"base": "User",
"idInjection": false,
"options": {
"validateUpsert": true,
"postgresql": {
"tableName": "people"
}
},
"hidden": [
"emailVerified",
"realm",
"username",
],
"properties": {
"first-name": {
"type": "string",
"postgresql": {
"columnName": "first_name"
}
},
"last-name": {
"type": "string",
"postgresql": {
"columnName": "last_name"
}
},
"email": {
"type": "string"
},
...
},
"validations": [],
"relations": {
"customer": {
"type": "belongsTo",
"model": "Customer",
"foreignKey": "customer_id"
}
},
"acls": [],
"methods": {}
}

scope in inherited model

I have a contact db table and model. Employee model inherits from contact.
If i do GET employees/ it returns all the contacts.
How should I set up my employee.json if I want to return only the contacts with partnerId = 1?
{
"name": "employee",
"base": "contact",
"strict": false,
"idInjection": false,
"options": {
"validateUpsert": true,
"postgresql": {
"schema": "public",
"table": "contact"
}
},
"scope": {
"where": {
"partnerId": 1
}
},
//...
}
Debug says calling GET employees/ makes the following query:
SELECT "name", "position", "email", "password", "id" FROM "public"."contact" ORDER BY "id"
It does not seem that scope is added.
models/partner.json
{
"name": "partner",
// ...
"properties": {
"name": {
"type": "string",
"required": true
},
// ...
},
"validations": [],
"relations": {
"contacts": {
"type": "hasMany",
"model": "contact"
}
//...
},
"acls": [],
"methods": {}
}
Try using the where filter, either in the REST API
/employees?filter[where][partnerId]=1
or in your Employee.js
Employee.find({ where: {partnerId:1} });
https://docs.strongloop.com/display/APIC/Where+filter

loopback hasMany relation remoteMethod hook

I am using loopback for develop my own website.
But recently I had a problem of hasMany remoteMethod.
Here is the problem:
I have two models :
person.json:
{
"name": "Person",
"base": "PersistedModel",
"strict": true,
"idInjection": true,
"properties": {
/*...
....
*/
},
"validations": [],
"relations": {
"friends": {
"type": "hasMany",
"model": "Friend",
"foreignKey": "personId"
}
},
"acls": [],
"methods": []
}
friend.json
friend.json:
{
"name": "friend",
"base": "PersistedModel",
"strict": true,
"idInjection": true,
"properties": {
/*...
....
*/
},
"validations": [],
"relations": {
},
"acls": [],
"methods": []
}
I want to use beforeRemote when I call POST /api/Persons/{id}/friends.
So I code in person.js
module.exports = function(Person) {
Person.beforeRemote('__create__friends', function(ctx, instance, next) {
/*
code here
*/
});
};
But it does not work!
At the beginning I think it's the matter of '__create__friends',but when I
code in person.js like :
module.exports = function(Person) {
Person.disableRemoteMethod('__create__friends');
};
I could disable the '__create__friends' successfully.
So what's the problem?
Can anyone help me?
Because methods for related models are attached to Person prototype, you should register hook like this:
Person.beforeRemote('prototype.__create__friends', function() {
next()
})