javascriptnode.jshapi.jshapi.js-lab

Inconsistent Hapi.js Lab test results


I am trying to test a Hapi.js plugin with registration function:

exports.register = function(server, options, next) {

    server.route({
        method: 'POST',
        path: '/register',
        config: {
            payload: {
                allow: 'application/json'
            },
            validate: {
                /* deleted for brevity */
            }
        },

        handler: function(request, reply) {

            if (!server.app.mongoose) {

                server.log('error', 'Failed to find an active MongoDB connection.');

                return reply(Boom.badImplementation());
            }

            var response = reply().hold();

            var mongoose = server.app.mongoose;

            var User = mongoose.model('User');

            var password = request.payload.password;

            return new Promise(function(resolve, reject) {

                bcrypt.genSalt(10, function(err, salt) {

                    if (err) {

                        server.log('error', 'Failed to generate bcrypt salt: ' + err);

                        return reject();
                    }

                    bcrypt.hash(password, salt, function(err, hash) {

                        /* deleted for brevity */

                        user.save(function(err, savedUser) {

                            if (err) {

                                server.log('error', 'Failed to save user to the database: ' + err);

                                return reject(Boom.conflict());
                            }

                            server.log('debug', 'Registered new user with e-mail validation code: ' + validationCode);

                            resolve({});
                        });
                    });
                });

            }).then(function(data) {

                response.statusCode = 201;
                response.source = data;
                response.send();

                return response;

            }, function(err) {

                if (!err)
                    err = Boom.badImplementation();

                response.statusCode = err.output.statusCode;
                response.source = err.output.payload;
                response.send();

                return response;
            });
        }
    });

    next();
};

My test file is here:

const Lab = require('lab');
const expect = require('code').expect;

const server = require('../');
const lab = exports.lab = Lab.script();

const mongoose = require('../plugins/mongo.js').mongoose;

lab.experiment('Registration', function() {

    lab.before(function(done) {

        mongoose.connection.collections['users'].drop(function(err, resp) {

            if (err) {

                console.error(err);

            } else {

                console.log(resp);
            }
        });

        var User = mongoose.model('User');

        /* deleted for brevity */

        user.save(function(err, savedUser) {

            if (err) {

                server.log('error', 'Failed to save user to the database: ' + err);

                done(err);
            }

            done();
        });
    });

    lab.test('/register endpoint with empty payload', function(done) {

        server.inject({
            method: 'POST',
            url: '/register',
            payload: {}
        }, function(response) {

            expect(response.statusCode).to.be.equal(400);
            expect(response.result.message).to.match(/^child "\w+" fails because \["\w+" is required\]$/);

            done();
        });
    });

    lab.test('/register endpoint with invalid email', function(done) {

        server.inject({
            method: 'POST',
            url: '/register',
            payload: {
               ...
            }
        }, function(response) {

            expect(response.statusCode).to.be.equal(400);
            expect(response.result.message).to.be.equal('child "email" fails because ["email" must be a valid email]');

            done();
        });
    });

    lab.test('/register endpoint with short password', function(done) {

        server.inject({
            method: 'POST',
            url: '/register',
            payload: {
                ...
            }
        }, function(response) {

            expect(response.statusCode).to.equal(400);
            expect(response.result.message).to.startWith('child "password" fails because ["password" length must be at least');

            done();
        });
    });

    lab.test('/register endpoint with invalid password', function(done) {

        server.inject({
            method: 'POST',
            url: '/register',
            payload: {
                ...
            }
        }, function(response) {

            expect(response.statusCode).to.equal(400);
            expect(response.result.message).to.startWith('child "password" fails because');

            done();
        });
    });

    lab.test('/register endpoint with existing username', function(done) {

        server.inject({
            method: 'POST',
            url: '/register',
            payload: {
                ...
            }
        }, function(response) {

            expect(response.statusCode).to.equal(409);
            done();
        });
    });

    lab.test('/register endpoint with valid payload', function(done) {

        server.inject({
            method: 'POST',
            url: '/register',
            payload: {
                ...
            }
        }, function(response) {

            expect(response.statusCode).to.equal(201);

            done();
        });
    });
});

Everything was fine until I added the '/register endpoint with existing username'.

Now subsequent invocations of lab command exits with success and then a failure for the tests 5 and 6 (it goes on and on like this, one success and then one failure). Basically, it looks like resulting status codes are swapped after a successful test launch.

Looks like a synchronization issue but I could not find where the cause is. Any ideas?


Solution

  • I managed to solve this problem by seperating server instances per tests and registering only the plugin(s) to be tested as recommended by the doc.

    I am answering my own question for anyone who may encounter the same problem in the future.