don't overwrite options.foreignKey in associations

[wilzbach]

I use sequelize quite often and just discovered are rare case where it doesn't exactly do what imho it is supposed to do. Let's assume we have three tables, where A & B have a string ID and C has an integer ID. Now when we try to connect B and C with A, we run into the problem that if a foreignKey is given as input, it will be modified by the associations. Hence this to this weird behavior:

var Sequelize = require('sequelize');
var sequelize = new Sequelize('database', 'username', 'password', {
    dialect: 'sqlite',
    storage: 'foo.sqlite'
});

var A = sequelize.define('A', {
    id: {
        type: Sequelize.CHAR(20),
        primaryKey: true
    }
});
var B = sequelize.define('B', {
    id: {
        type: Sequelize.CHAR(20),
        primaryKey: true
    }
});
var C = sequelize.define('C', {});
var reqValidForeignKey = { foreignKey: { allowNull: false }};

//B.hasMany(A, reqValidForeignKey); // other associations also modify the options.foreignKey
A.belongsTo(B, reqValidForeignKey);
A.belongsTo(C, reqValidForeignKey);
var QueryGenerator = A.modelManager.sequelize.queryInterface.QueryGenerator;
console.log(QueryGenerator.attributesToSQL(A.rawAttributes, {
    context: 'createTable'
}));
//{ id: 'CHAR(20) PRIMARY KEY',
//  createdAt: 'DATETIME NOT NULL',
//  updatedAt: 'DATETIME NOT NULL',
//  BId: 'CHAR(20) NOT NULL REFERENCES `Bs` (`id`) ON DELETE CASCADE ON UPDATE CASCADE',
//  CId: 'CHAR(20) NOT NULL REFERENCES `Cs` (`id`) ON DELETE NO ACTION ON UPDATE CASCADE' }

Note: CId should be an integer.

This is due to the fact that we use the same foreignKey object which is passed into the function and modify. In my use case I discovered this because both tables are required to have foreign keys.

This PR adds a very simple fix for this problem and also a test case to check for this ;-)

[janmeier]

There's no need to call .sync here - which means the test can be converted to a unit test

[janmeier]

LGTM, just a single comment

If you change that, add a changelog entry and squash this should be good to merge

[wilzbach]

If you change that, add a changelog entry and squash this should be good to merge

Thanks for your help :)

Should I be worried about travis failing for MySQL?

[janmeier]

Not as long as it's only a single job failing - we have an intermittent failure which we've not been able to track down yet

[janmeier]

Woops, sorry if my comment was not clear enough, could you move the file to test/unit

[wilzbach]

Woops, sorry if my comment was not clear enough, could you move the file to test/unit

No worries - my fault ;-)

[janmeier]

Thanks! :)