Did My Jasmine Expect Method Get Called?

When unit testing with Jasmine, expect() calls are not mandatory. That is, calling expect() at least once is not enforced by Jasmine. I recently ran into a problem which caused me to ask myself “did that expect method get called?”. I couldn’t count on Jasmine for this – in fact, my tests pass whether I include the expect() call or comment it out! So I went digging..

I determined that I could simply create spies for my expect() calls. This is an easy way to leverage Jasmine to inspect your tests. Simply create your spy:

const expectSpy: jasmine.Spy = spyOn(window, 'expect').and.callThrough();

I am using TypeScript for my unit tests. Since the expect() method is global and I am running my tests in a browser, I use the window object directly. There are ways to obtain the global object without this sort of hard-coding but, that is besides the point.

Moving on, the expect() calls must work properly so and.callThrough() is called. This is important. Without including and.callThrough(), your tests will fail because, rather than Jasmine’s expect() execution, you will be limited to a jasmine.Spy.

Here is a more complete example of a test with an expect spy – slightly modified from a sample Angular 2 application I have been working on:

it('should trigger on selection change', async(() => {
  const expectSpy: jasmine.Spy = spyOn(window, 'expect').and.callThrough();

  const triggerSpy = spyOn(component, 'triggerThemeChanged');
  const select = fixture.debugElement.query(By.css('select.theme-selector'));
  dispatchEvent(select.nativeElement, 'change');

  fixture.whenStable().then(() => {
    expect(triggerSpy).toHaveBeenCalledTimes(1);
  }).then(() => {
    expect(expectSpy).toHaveBeenCalledTimes(2);
  });
}));

There are a few things about this test that are not the point of this article – what the heck is async() and the apparent improper use of dispatchEvent()? The important bits are the use of Promises as implied by the use of then() callbacks, the creation of the expect spy, and the inspection of the expect spy.

The test creates the expect spy and then uses expect() as usual within the test until it finally inspects the expect spy. Remember, the inspection of the expect spy counts as an expect() call! This is why expect(expectSpy).toHaveBeenCalledTimes(2) is called with 2 rather than 1.

I stopped at the call count. This test could be extended to further leverage the Jasmine API by looking at expectSpy.calls with other handy methods to make sure the expect() calls were made properly. I’ll leave that for an exercise for the reader. Just make sure your testing, at a minimum, covers the scope of your problem.

If you have had similar issues or have explored this in more depth I would be very interested in hearing about your journey! Comments are welcomed and appreciated.

Meteor Hang-up: Extracting Package….

In the world of Node.js and NPM, things can change at an increasingly rapid pace. This causes pain when starting or upgrading projects that require NPM packages. While there are sites like Greenkeeper, I see them as symptoms of a flawed system. Yes, I will say that without offering alternative solutions because at the moment I am aware of exactly zero. Suggestions welcome!

It is a wonderful world of possibility.

Complaining about NPM is not the point of this article. I’ll stop wasting time:

Recently I came across a few excellent tutorials about using Meteor, Ionic 2, Angular and React. They eventually brought me to Telescope Nova. My first thought was: this looks promising.

After forking and cloning and other Gitisms, I was ready to start the application:

npm install
npm run start

Of course I have a Microsoft development background so when I saw a bunch of red because of ‘.sh’ I wondered why these two letters were such a problem. I ended up having to update my start script to exclude this bit of code. The script I excluded simply renames a sample_settings.json file to settings.json. I figured that was a safe thing to shortcut in this case by renaming it myself.

My next step was to try it again!

> Nova@1.0.0 start C:\Demo\Telescope
> meteor --settings settings.json

[[[[[ C:\Demo\Telescope ]]]]]

=> Started proxy.
=> Started MongoDB.
=> Extracting std:account-ui@1.2.17

To be honest, I let it try for a few hours and it just could not get that pesky package extracted. Certainly something had gone wrong before that. After digging into the depths of Nova, Meteor, and NPM, I finally explicitly searched within Stack Overflow for: Extracting std:accounts-ui.

The search came up with only 2 results which are both linked at the bottom of this article. Most importantly: following the suggestions solved my problem.

I fixed the issue by relocating the 7z executable (7z.exe) from: C:\Users\[UserName]\AppData\Local\.meteor\packages\meteor-tool\1.4.2_3\mt-os.windows.x86_32\dev_bundle\bin to a place outside of any source/build code/tools location. I relocated it because I didn’t want to mess up my machine any more than it already may have been. Turns out, the missing 7z.exe was all it took to get my Meteor package installed properly!

It figures that the solution was to create a sort of FileNotFound scenario.

In an effort to spread the word, the following links lead me to this solution:

http://stackoverflow.com/questions/41155583/meteor-1-4-2-3-adding-package-extracts-forever-windows

http://stackoverflow.com/questions/41195227/meteor-package-extracting-forever

https://github.com/studiointeract/accounts-ui/issues/67

https://github.com/meteor/meteor/issues/7688

I hope this helps. It is a rather simple solution in the end. I am very interested in learning of your past issues with our current favorite packaging system and its various dependencies. Feel free to comment if you have hard-fought wisdom to share!

UPDATE: just a quick update here, turns out this approach can be helpful when updating packages or if you get stuck here ‘Extracting meteor-tool@1.4.2_5’ (after the recent patch). Note: extracting meteor tools can take awhile (upward of 30+ minutes) so expect to wait awhile to know if it fails.

Bring Your Own Shell (Update 1.9 for VS Code)

I received the recent update for VS Code with pleasant surprises for the Integrated Terminal! Most importantly, it now defaults to Powershell. This can be changed using VS Code settings. This ability is not necessarily a new option within VS Code but, if you rather your own shell, then here we go!

Accessing Settings

To access VS Code settings, use the menu under File > Preferences and click Settings. A settings.json file opens with a list of Default Settings. These can be copied to your workspace or user settings.

With the Settings open, finding settings is easy – just type the setting name (or scroll) as shown in Figure 1.

settingsjson
Figure 1: Looking for ‘shell’ commands in settings.json

Bring Your Own Shell

According to VS Code Documentation, a different shell executable can be used given that it adheres to this very important requirement:

..the shell executable must be a console application so that stdin/stdout/stderr can be redirected..

This offers a lot of flexibility to support your given scenario and preferences. Update the shell using the format: “terminal.integrated.shell.platform”: “full path to shell executable”. For example, when specifying the setting for Windows this would be similar to:

"terminal.integrated.shell.windows":
    "C:\\WINDOWS\Sysnative\\WindowsPowerShell\\v1.0\\powershell.exe"

Shell Arguments

Arguments can be passed to this executable using the following format: “terminal.integrated.shellArgs.platform”: [“arg1”, …]. Specifying the setting for Windows would be similar to:

"terminal.integrated.shellArgs.windows": ["Get-Help", "*"]

-or-

"terminal.integrated.shellArgs.windows": ["Get-Help *"]

With those sample arguments, the terminal window will close after the help text displays for Powershell. While that is less than useful, the settings are there for your benefit. Choose the shell you want within VS Code and configure it accordingly. Of course, make sure you first have it installed.

Platforms

I’m sure you all noticed the “platform” placeholder within the setting formats. Currently the platforms for terminal settings specified within VS Code are:

  • linux
  • osx
  • windows

VS Code is awesome, what do you think? I am eager to hear of your experiences using Visual Studio Code!

Thankful for Integrated Terminals in Visual Studio Code

Since before I started my software development career, I have been using Visual Studio. It is a beast-mode IDE with tons of heavy-weight features and a bountiful supply of extensions. Don’t get me wrong, Visual Studio is near and dear. It has been the key to successful delivery of many kinds of projects. Since Visual Studio was so useful, I honestly wondered why Visual Studio Code was even created. After using it though, my thoughts changed from “why?” to “this is amazing!”

To be able to view everything on a single screen is hard to do – some would say unnecessary or even Noise. One of the great thing about VS Code is that it accomplishes the two feats of allowing a developer to quickly discover what they need during development without needing to wade through the noise. This is done by a simple tab-like interface including Explorer, Search, Git, Debug, and Extensions. The best part about this is the Integrated Terminal. It is very simple to get started, easily accessible, and easily ignored.

Integrated Terminals

vscintegratedterminal

My primary use for VS Code is front-end development. VS Code offers great tooling for creating high-quality user interfaces using the latest frameworks such as Angular 2 and React. Full-stack development can be done at once with a single language (JavaScript). With Integrated Terminals, Node.js operations can be performed quickly and easily.

With Extensions offering further integration, I find VS Code especially useful when developing with Node. Its Integrated Terminals allow an easy way of managing Node operations. When developing a MEAN stack application, each tier can get its own terminal for operations specific to that tier.

Dedicated terminals per tier is also beneficial because certain operations are long-running such as running MongoDB, Node and Angular. I tend to create a terminal per long-running operation. VSC makes it very easy to get these started – and to ignore them when they are not appropriate for the current task. For example, I can ignore the DB and Services when styling an Angular 2 component while still maintaining the benefit of them running so I can quickly test my work.

When writing front-end code, I would definitely consider Visual Studio Code.

Commands

  • Ctrl+`
    • Toggle Integrated Terminal Panel
  • Ctrl+Shift+`
    • Create New Integrated Terminal Instance
  • Ctrl+Shift+P
    • Begin searching the Command Palette

User Interface

vscintegratedterminalpane

Interact with Integrated Terminals like you would a Command Prompt. Multiple instances can be created while being accessible by simply making a selection from the drop-down. From the small toolbar, instances can be removed, more instances can be created, and the down-arrow closes the entire panel.

and to close,

Write End-to-End Tests for Your Angular 2 Applications With Protractor

Testing your applications is a critical step in ensuring software quality. While many agree, it can be hard to justice the budget for it. This issue can occur when development and testing are imagined as two separable activities. In fact, there are common practices in place designed around their coupling (see TDD). End-to-End (E2E) testing is another common practice designed in this way.

E2E testing can be thought of as an additional direction or angle in which to test an Angular 2 application’s logic. It tests, from a user’s perspective, if the application does what is expected. If these tests fail, your users will fail and the failure will be right in their face. This user-centric approach is critical to ensuring an intended user experience. Writing these tests will produce a detailed user experience specification enforced by simply making sure the tests pass.

End to End testing does not replace QA resources. All software is written by people and people are not perfect – no matter what they say.

Interacting with Your Application

To begin creating our tests we need a way to interact with the application. This is made easier with Protractor. With Protractor, Selenium can be leveraged within JavaScript including integration with Angular applications. A few critical components are exposed to you through Protractor.

browser – browser-scoped operations such as navigating to a particular URL.
element – provides a way to retrieve the UI components of your application from within your tests.
by – What UI component do you want and how should it be found?
promise – support asynchronous operations.
ElementFinder – perform operations on retrieved UI components.
ElementArrayFinder – perform operations on an array of retrieved UI components.

Importing from Protractor

Importing what you need from Protractor is as easy as importing anything else:

import { browser, element, by, promise, ElementFinder } from 'protractor';

Navigating to a URL

Get to the correct page using browser.get():

navigateTo() {
    return browser.get('/list');
}

Retrieving and Manipulating UI Components:

Use the Protractor API to retrieve elements and perform operations in your application:

clickEditButton(): promise.Promise {
    return element(by.css('.edit-button')).click();
}

Creating Page Objects

Now that a few fundamentals are out of the way, lets take a look at organizing our tests. First, we need to create Page Objects. These objects encapsulate interactions with UI components. Implementing Page Objects could be thought of as user experience mapping including structure and operations.

With single page applications, a “Page Object” could get quite unwieldy! This is why we need to consider the structure of the application. Considering the Angular Components within the application is a great way to begin dividing your Page Objects. Take a simple Todo application shown in Figure 1:

todoscreenshot
Figure 1: Sample todo application

Even with this simple application there is a lot to include in just a single Page Object. This screen can be divided up into four distinct areas: header, menu, list, and details. Let’s look at how these can be created using separate Page Objects:

Header

todoheader

// header.po.ts
import { promise, ElementFinder } from 'protractor';

export class Header {
    getContainer(): ElementFinder {
        return element(by.css('.header'));
    }
    getHeader(): ElementFinder {
        return this.getContainer().element(by.css('.header-text'));
    }
    getHeaderText(): promise.Promise {
        return this.getHeader().getText();
    }
}

Menu

todomenu

// menu.po.ts
import { element, by, promise, 
    ElementFinder, ElementArrayFinder } from 'protractor';

export class Menu {
    getContainer(): ElementFinder {
        return element(by.css('.menu'));
    }
    getMenuItems(): ElementArrayFinder {
        return element.all(by.css('.menu-item'));
    }
    getActiveMenuItem(): ElementFinder {
        return this.getContainer()
            .element(by.css('.menu-item.active-menu-item'));
    }
    getActiveMenuItemText(): promise.Promise {
        return this.getActiveMenuItem().getText();
    }
    getCompletedMenuItem(): ElementFinder {
        return this.getContainer()
            .element(by.css('.menu-item.completed-menu-item'));
    }
    getCompletedMenuItemText(): promise.Promise {
        return this.getCompletedMenuItem().getText();
    }
    getAddMenuItem(): ElementFinder {
        return this.getContainer().element(by.css('.menu-item.add-menu-item'));
    }
    getAddMenuItemText(): promise.Promise {
        return this.getAddMenuItem().getText();
    }

    clickActiveMenuItem(): promise.Promise {
        return this.getActiveMenuItem().click();
    }
    clickCompletedMenuItem(): promise.Promise {
        return this.getCompletedMenuItem().click();
    }
    clickAddMenuItem(): promise.Promise {
        return this.getAddMenuItem().click();
    }
}

List

todolist

// list.po.ts
import { element, by, promise, 
    ElementFinder, ElementArrayFinder } from 'protractor';

export class List {
    getContainer(): ElementFinder {
        return element(by.css('.list'));
    }
    getItems(): ElementArrayFinder {
        return element.all(by.css('app-todo'));
    }
    getItem(index: Number): ElementFinder {
        return this.getItems().get(index);
    }
    getEditButton(index: Number): ElementFinder {
        return this.getItem(index).element(by.css('.edit'));
    }
    getDeleteButton(index: Number): ElementFinder {
        return this.getItem(index).element(by.css('.delete'));
    }

    clickEditButton(index: Number): promise.Promise {
        return this.getEditButton(index).click();
    }
    clickDeleteButton(index: Number): promise.Promise {
        return this.getDeleteButton(index).click();
    }
}

Details

tododetails

// details.po.ts
import { element, by, promise, ElementFinder } from 'protractor';

export class Details {
    getContainer(): ElementFinder {
        return element(by.css('.details'));
    }
    getDetailHeader(): ElementFinder {
        return this.getContainer().element(by.css('.detail-header'));
    }
    getDetailHeaderText(): promise.Promise {
        return this.getDetailHeader().getText();
    }
    getDescription(): ElementFinder {
        return this.getContainer().element(by.css('.detail-description'));
    }
    getDescriptionText(): promise.Promise {
        return this.getDescription().getText();
    }
}

Alright, that was a lot of code. Let’s take a step back for a moment. We are creating Page Objects representing the distinct areas of the UI that we have determined. Each Page Object has offered the ability to retrieve (e.g. getContainer) and operate (e.g. clickDeleteButton) on UI elements from their respective areas. The last item we need now is a root Page Object to complete our Page Object hierarchy.

The Page Object Hierarchy

todopageobjecthierarchy

Yes, I can’t seem to help it – there is one additional Page Object that has been included along with the root Page Object. The Item Page Object will encapsulate the structure and logic of each todo item within the List.

An instance of each leaf Page Object (which includes header, menu, list, and details) is stored on the root Page Object (TodoApp). This provides the ability to write complex operations while exposing a simple API to your tests:

// snippet todo-app.po.ts

export class TodoApp {
    constructor() {
        this.list = new List();
    }
    private list: List;

    removeAllItems(): promise.Promise<void[]> {
        let promises: Array<promise.Promise> =
            new Array<promise.Promise>();
        this.list.getItems().each(item => {
            promises.push(item.clickDeleteButton());
        });
        return promise.all(promises);
    }
}

In removeAllItems() each todo item is found and its Delete button is clicked. We should no longer have any todo items. This is a testable scenario. Testing this would mean answering the question – what happens when there are no items? We can use our Page Objects to create tests around this scenario!

Testing with Page Objects

Creating tests using Page Objects can clean up your tests and allows more explicit definition of your test scenarios. This way also helps keep the user interface definition out of the tests making it easier to maintain – keep in sync with your application!

Jasmine Syntax

If you have written tests using Jasmine before, you know how to create tests for Protractor. It is a great tool for writing unit tests! It is just as good for writing tests with Protractor. Simply follow their tutorials to learn more about Jasmine.

Writing Tests

Now lets see some tests. We will test the ability to remove all items by clicking their Delete button:

// todo-app.e2e-spec.ts
import { TodoApp } from './todo-app.po.ts'

describe('Todo Item', () => {
   let page: TodoApp;
   beforeEach(() => {
      page = new TodoApp();
   });

   it('delete button should remove todo item', (done) => {
      page.navigateTo();
      const text: String = 'test text';
      page.addItem(text, text).then(() => {
          page.removeAllItems().then(() => {
             expect(page.getItems().count()).toBe(0);
             done();
          });
      });
   });
});

This test sample is following a few common practices. First, the root Page Object is imported:

import { TodoApp } from './todo-app.po.ts';

Next, the top-level test suite is defined. Within this test suite a variable is defined to hold our root Page Object. Then comes the initialization of each test in the suite:

beforeEach(() => {
    page = new TodoApp();
});

This is a fairly rudimentary example but, the initialization code will run once before each test in the suite. Make sure to place any necessary test initialization code within this block such as initializing Page Objects and setting up the UI.

Next comes the fun. The test is defined using typical Jasmine syntax with support for asynchronous test code:

it('delete button should remove todo item', (done) => {
   page.navigateTo();
   const text: String = 'test text';
   page.addItem(text, text).then(() => {
       page.removeAllItems().then(() => {
          expect(page.getItems().count()).toBe(0);
          done();
       });
   });
});

The first step in our test is to navigate to the TodoApp and create a variable to hold our new item text and description.

Considering that the test is for ensuring items get deleted when their Delete button is clicked, we need to be sure there is an item to remove! Using a Page Object method for adding an item, a new todo item is added to the screen using the new item text and description defined earlier. Since the method returns a promise, we use then() to continue execution.

We have added an item. Now its time to remove it. We know that the Page Object contains a method that clicks a todo item’s Delete button. We also know that the Page Object provides the ability to remove all of the todo items in the list using this method. This means we can simply call removeAllItems() and check the list to make sure it is empty.

If the test passes we can say the Delete button works for each todo item.

Mean TODO

There is a sample application demonstrating Angular 2 in the MEAN Stack that includes a set of E2E tests. Simply follow the guide on GitHub to get started!

Thank you for reading this far! If you have any comments, questions or concerns please send a message.

Angular 2 in the MEAN Stack – A Project Template

I recently became aware of the beauty of developing with the MEAN stack. It started with my desire to brush up on Angular 2 and using the Angular CLI. I ended up with a TODO application that runs on Node using MongoDB, Express.js, and Angular 2 (from a foundation found here: https://scotch.io/tutorials/mean-app-with-angular-2-and-the-angular-cli).

I am also a fan of LESS. It makes writing modular CSS a breeze. I am also not yet sold on the Styled Component approach Angular is now pushing. I do believe Component-based UI architectures make writing modular CSS quite simple. I just don’t think styles should be located across the application when, with an optimally modular UI, the CSS used can be quite minimal. Beyond that, an application should have a look and feel that gives the user at least the illusion of cohesiveness. Easily done when styles are in one place.

Using Feature Modules with Angular 2 makes a lot of sense. Each feature can be created in isolation while still being integrated with the rest of the application. The module structure of the TODO application is shown in Figure 1. This structure allows the application to be extended and makes features easy to find for updates and bug fixes. The approach also provides a more SOLID application.

todofeaturemodule
Figure 1: Module structure

With the focus on Angular 2, the server-side code is quite minimal and I don’t have much to say about it. Figure 2 depicts the main ideas.

todoapi
Figure 2: Server-side

I almost forgot to mention that this app is on GitHub: https://github.com/calebmagenic/ng2-mean. Let me know your thoughts!

I look forward to continuing my work with the MEAN stack and sharing what I find that works – and what doesn’t work so well. Look for more in the future.