Protractor: Tips for better End-to-End Tests

Standard

protractor-logo

So last year I shared my initial experiences of getting started with Protractor. Here are some things i’ve learnt since then.

1. Use testSuites

As your Protractor tests grow you are undoubtably going to make use of testSuites. This allows a specific group of tests to be run by passing in the testSuite name into the run parameters.

2. Use only one protractor configuration file

Protractor uses a configuration file which contains various settings for your tests such as baseUrl, tests to run, web driver port etc. When running tests in different places it is tempting to have multiple configuration files. For example a local config and one for your CI server. Avoid this temptation and only have one config file.

3. Manage your Protractor package and dev dependencies

Keeping Protractor up-to-date as well as other packages, such Screenshot Reporter, which you depend upon is really important. If you don’t manage your packages they will quickly become out of date. Luckily using a package.json to store all your packages makes this super easy. You can declare if you want to auto update or use a specific package version when you run npm install.

4. Use a Task Runner such as Gulp

Your protractor tests will just be a small part of your JavaScript eco system. Using a Task runner can help to keep all your tasks in a single place. It also means it’s very simple for anyone in the team to run the Protractor Tests.

5. Learn how to Debug Protractor Tests

My experience is that writing E2E WebDriver tests in Java or C# is a lot easier because the IDE, such as Visual Studio, gives you lots of guidance. As Protractor Tests are JavaScript you get less help so need to understand how to debug. The two main options are browser.pause() and browser.debugger(). Aside from the Protractor debug commands I’d also recommend learning how to use browser development tools such as Chrome Dev tools to help debug JavaScript.

6. Use Protractor for non-angularJs sites only when necessary

Protractor can be used to test non-AngularJS sites but i’d only recommend this when testing a user-flow which jumps out of your AngularJS site, for example page redirects. I would suggest not using this capability to mean you can test non-Angular sites using Protractor.

7. Use other node packages to do stuff

The joy of using Node is that Node Package Manager allows you to do loads of stuff. An example is http-request which you can use to call API’s in your test flow. This is particularly useful if you want to setup some data via a direct API call.

8. Use Page Object Model

Just do it.

9. Multi Thread your Tests

In Protractor it’s so easy to run your tests in parallel and across multiple browsers. Even with a handful of tests you are going to have to look at running tests in parallel for time saving benefits. Running tests in parallel also forces you to have independent tests which is a good way to keep tests maintainable.

10. Handling Cookies and Sessions

Protractor by design does not open/close the browser between specs. This is useful in one way because it speeds up your tests. Closing and Opening browsers between tests has a significant footprint. There are cases though when you need to potentially close a browser to at least clear cookie or session history. In this case you have a couple of options. Firstly WebDriver can interact with cookies or alternatively you can add restartBrowserBetweenTests: true to your protractor config file.

11. Learn AngularJS

Having an understanding of the underlying technology is a huge help in writing good tests. As you’ll probably be using Protractor to tests AngularJS sites then you should definitely learn Angular. Build your own site in Angular and write some tests for it. There are tonnes of great sessions on Pluralsight to help. Learn also how to unit test AngularJS code as this will give you a better appreciation of where to place your test efforts.

12. Use ES6 Generators

At AngularConnect Uri Shaked did a great talk on how to use ES6 generators to control the flow of your Protractor Tests. This is particularly useful if you need certain tasks to occur in a  particular sequence.

13. Understand what Protractor does

Protractor is a wrapper around WebDriver specifically designed for AngularJS apps. This means most of the benefit will come when using AngularJS apps. The two primary benefits i’ve found are:

  • Page synchronisation – As protractor hooks into the angular digest cycle it makes one of the typical problems with E2E tests, When has the page loaded?, problems a lot easier to manage.
  • Angular Specific Locators – by.Binding, by.Model, by.Repater gives you more options for finding stuff on a page.

With this in mind i wouldn’t ever consider using Protractor for a site which isn’t built in AngularJS.

Thats more than enough tips for now.

Enjoy Protractor.

Advertisements

Using Gulp for Testing

Standard

Last year I shared some posts on testing AngularJS web apps:

Getting started with Protractor to automate E2E tests for AngularJS apps

How to steep a unit test environment for AngularJs Application

I thought i’d start 2016 by sharing some other tips in that area.

What is Gulp?

gulp.jpg

Gulp is a JavaScript Task Runner. Now it took me a while to actually get my head around what Gulp actually does. Gulp enables you to encode in JavaScript repetitive tasks that you perform regularly. This could be starting a web server, JavaScript minification or Testing which will be the focus of this blog.

One of the core benefits of Gulp is that it encodes these tasks in a JavaScript file which is typically named gulp.js. This means all of these encoded tasks can be shared across the team and kept in synch using version control.

Another benefit is that gulp utilises the wide variety of packages available via NPM(Node Package Manager) which means you can automate tasks incredibly easily.

In this blog I won’t go through how to setup Gulp from scratch or how to install Node Packages as the official documents do this very well. Instead my focus will be how you can use Gulp to support testing.

If you want to get started from scratch in addition to the official docs i’d highly recommend John Papas Gulp course on Pluralsight – JavaScript Build Automation with GulpJs

Note: if you don’t currently have a Pluralsight account you can get 3 Months FREE by creating a free Visual Studio Dev Essentials account

How to use Gulp for Testing

Gulp can be used for many tasks but in this blog post I’ll focus on tasks that can be used to support testing.

  1. Run Unit Tests

This task runs your JavaScript unit tests using Karma. It utilises your karma configuration file which allows you to configure run parameters, such as single run and also what outputs we want to generate, such as Istanbul Code Coverage reports and jUnit XML. You can read more about configuring your unit tests in my previous post

Using Karma you don’t need to install any additional gulp packages as you can call Karma directly through Gulp as explained here

The gulp task looks as follows:

Snip20160103_2

2. Run Protractor E2E Tests

As detailed in a previous blog post, Protractor, is an E2E testing framework mostly used for AngularJS web apps. This gulp task allows you to execute your protractor tests. You’ll need to make some additional tweaks to the task to allow a test suite to be specified. We use protractor test suites to allow control of what tests are run. To do this we used an NPM package called yargs. This blog helps.

What is cool with this task is that in a single task it will get the latest web driver jar, start a selenium server, run the tests and generate the output.

To run your protractor tests using gulp you’ll need to install a couple of node packages:

Gulp Protractor

Yargs (Used for specifying testSuites)

The gulp task looks as follows:

Snip20160103_3.png

 

3. Linting

Linting is an activity to discover common mistakes in code and is incredibly useful for JavaScript as its not a compiled language. It’s a great example of static testing that can yield great benefits. Here is an example of the output you get:

reporter_problems.png

Using this gulp task you can Lint your code to identify any simple mistakes or errors keeping the code base clean.

To do the linting you need to install the node package JsHint-Stylish

The gulp task looks as follows:

Snip20160103_5

 

Putting it all together

 

To run these tasks you simply start up a nodeJs command prompt and enter gulp followed by your task name. For example:

gulp protractor –testSuiteName

This will run your specified e2e protractor test suite.

As NPM gives you such a wide variety of options there are many more tasks you can add to support your testing. These are probably the three most common. Another example could be starting your mock services.

Local vs CI

You obviously want to avoid duplication so you should use the same Gulp file both locally and on your CI tool such as Jenkins. To achieve this you just need to learn about file paths and ensure your gulp.js file is located sensibly, such as within the project directory.

When running any of the three testing tasks outlined above there is one main difference locally vs CI. Locally you should set Karma and JsLint to watch your files. This provides instant feedback when developing your code. On your CI tool you will want these tasks to execute once, exit and produce the relevant outputs. Executing your gulp tasks through a CI tool such as Jenkins is super easy. In the case of running your Protractor tests the CI job essentially has two commands:

npm install –> Installs your node packages at build time

gulp e2e –testSuiteName –> Executes Protractor tests

You then just need to configure your CI tool to pick up any outputs generated.

Challenges

So this all sounds pretty straightforward but for me it did take a little while to get everything working smoothly. The main challenges I had:

1. What the hell is Gulp!?!?

Sounds simple, but it actually took me a few hours to actually work out the concepts behind Gulp. Hopefully this post has helped clear that up for you.

2. Handling Node Packages Correctly

Gulp in the most part uses node packages to “Do Stuff” so you’ll need to have a good grasp of how that works. Take the time to understand how to use a package.json, devDependencies, npm install and the differences between installing globally v locally. Interestingly, you’ll also need to install gulp both locally and globally.

3. Folder Structres and File Paths

You’ll need to make sure you have your node packages and gulp.js file in a sensible location.The standard is to install dev node packages within the project and your gulp file in the project root. Your gulp file will reference underlying configure files such as karma.conf.js and also declare output directories so you’ll need to understand how to move around the directory structure using JavaScript.

Alternatives

If you keen to crack on and use Gulp, at this point i’d highlight that there are alternatives out there which I haven’t used, the most popular is Grunt.

Finally, a heads up that Gulp 4 is currently in beta and will be launched soon

Pedestrian Crossings, Continuous Integration and Culture

Standard

6792_AA13-Create-website-graphics-from-2013-press-advertising_Walking-Men_DE1-1024x376

You have just arrived into a new country, checked into your hotel and ready to head out exploring the area. Soon after leaving the hotel you will face your first challenge; the Pedestrian Crossing.

I’ve been lucky to visit a few different countries and not so lucky to experience some of the world’s wildest Pedestrian Crossings (Could make a great TV Show!). I live in the UK where Pedestrian Crossings are, at least to me, well signposted and generally observed both by pedestrians and drivers(with exceptions).Across the rest of Europe you’ll find experiences, in particular Italy. My strategy when i visit; Step out onto the pedestrian crossing, hope the driver notices you and stops, if he doesn’t get ready to run! Seems like i’m not alone! Other parts of the world such as India, Thailand, Vietnam introduce other examples of even riskier behaviour to cross the street.

Japan is one of the most fascinating countries i’ve visited. Incredible cities, beautiful countryside and very obedient pedestrians. Late one night, at around 3am, i was returning to my hotel. The roads were empty and not a vehicle in sight. Regardless to the apparent lack of danger, if the “Wait” sign was shown, pedestrians did not cross.

17464672-TOKYO-DECEMBER-15-Shibuya-Crossing-December-15-2012-in-Tokyo-JP-The-crossing-is-one-of-the-world-s-m-Stock-Photo.jpg

Pedestrian Crossings and Culture

As you travel the world you realise culture has a huge impact on pedestrian behaviour. An international research team did a study into the impact of Culture; Different risk thresholds in pedestrian road crossing behaviour: A comparison of French and Japanese approaches.

“The group set up observations posts at various crossings in the French city of Strasbourg and the Japanese city of Inuyama. The results were both clear and striking. At the legal crossing in Strasbourg, France, about 67 percent of pedestrians crossed against the red light. Only 7 percent of walkers in Inuyama did the same. At the unmarked crossing, the researchers measured how much time between cars a pedestrian needed before deciding to walk. In Strasbourg that time clocked in at 9 seconds, on average; in Inuyama, walkers felt free from risk at 16-second gaps.”

These findings match my anecdotal experience across Europe and in Japan.

Continuous Integration

After reflecting on a near death experience at one crossing in Bangkok recently, Continuous Integration came to mind. Continuous Integration is a practice now widely adopted across the industry. Join a development team and they will probably be using tools such as Jenkins, Bamboo and Team City to monitor the build process. All these tools use indicators very similar to a Pedestrian Crossing.

Lets take Jenkins for example:

Snip20160101_1

Blue Ball = Everything appears Good.

Red Ball = Something went wrong.

These coloured indicators look simple but in the same way as pedestrian crossings, they influence behaviour in different ways.

Lets take the following two teams:

Team A

The build fails and a Red light is shown on the Jenkins dashboard. The team stops, identifies the source of the failure, perhaps by nudging the last person who checked in. The source of the problem is identified, rectified and build is run returning to normal.

Team B

The build fails and a Red light is shown on the Jenkins dashboard. The team ignores the failure, continues checking-in and the builds continue to fail. Later that day, a developer in the team identifies a compile issue locally, checks-in a fix. The build returns to normal.

How do you cross the street?

What actions we take based upon the build indicator is influenced  by a number of factors. I’ve seen teams confidence of build indicators eroded by false positives/negatives leading to ignorance of builds. However, the biggest factor in my experience is team culture.

In the example of Team A, they are clearly influenced by their CI tool and a culture to investigate problems quickly and implement solutions. However in the example of Team B, their culture is to ignore the build failure potentially with the view that it will be resolved at a later date.

So what is your teams behaviour towards your Continuous Integration tool?

Are you like the Japanese? Do you stop when the build goes Red?

Or…..

Are you like the French who in most cases ignore the warning signs?

In a perfect world we’d all agree that if the pedestrian light says stop, we would all agree thats the best action to take. However, i’m sure we have all crossed the road, ignoring the sign.

Maybe it’s the same for build indicators, if the build is Red, maybe its OK to ignore……….sometimes?

Ignorance is bliss

Ignoring builds might be ok in some cases but the danger comes when we grow a team culture that blindly ignores the warning signs. The occasional ignored failed build can often develop into a culture where build failures are always ignored. This culture might yield some good results but it’s likely to result in some near misses, like this gentleman:

Testers are second class citizens.

Standard

This week I attend the London Agile Discussion Group which had an interesting debate topic:

“Testers are second class citizens”

To give the evening some structure we were split into two groups, the proposition and the opposition. Each group had 15 minutes to prepare their statement which they would then present back to the audience.

The Proposition

Up first the proposition who agreed testers are second class citizens.

As they began I was actually surprised how passionately they put across their statement. I bit my tongue and resisted the urge to shout “You are wrong!” but at the same time i found it fascinating to hear this side of the debate. The debate spurred up a strange sense of emotions. It made me thankful that i don’t have these discussions in my own team otherwise i’d be a very angry man!

These are some of the arguments put forward by the Proposition.

Testers are second class citizens because……….

– Testers are paid less
– Testers are generally more junior members
– Testers are less skilled than other team members such as developers
– Testers are just a “distraction”
– Testers don’t add any value to the software
– You can make money without testers
– If you do TDD or BDD you don’t need testers
– Why use testers when your customers can do testing for you in Production
– Testing can be done by anyone
– It is commonly sent off shore
– Testers/Testing is an entry level for IT industry
– Testing is not seen as a career
– Testers are failed developers or individuals that want to eventually become developers
– We should aim for generalising specialists not specific roles. So based upon this the “tester” role itself doesn’t exist, as doesn’t “developer”, “BA”. We are all just team members.
– If you had to sack all your developers or all your testers. Who would you sack?

The Opposition

Second up was the opposition. I’d conveniently been placed in this group. There were 3 other testers in my group and as you can imagine we passionately believed in the value testers can add however being testers we did have questions around the debate topic:

– What do we mean by “Tester”?
– To whom are testers second class citizens? Developers? Business Owners? God?
– Who may think this?

To present our opposition we first looked at why people might think testers are second class citizens. We all had experiences where testing had been sent “off shore” maybe to cut costs because testing is seen as an unwanted cost, we also had experiences where developers had thought anyone could test and that as testers don’t add value to the product they are not needed.

These were the arguments we put forward as the opposition:

Testers are NOT second class citizens because……….

– We think its important to look beyond roles in Agile Teams. We are “Team Members”
– There should be no “class system” within agile teams
– Whole Team approach to quality is important
– We value that idea of “T-Shaped” team members and having team members with a deep knowledge and understanding is very, very important. We feel that its important for the following skills to be present(many of which are found in testers):
— Communication
— Problem Solving
— Analysis Skills
— Creativity
— Development/Programming skills
— Understanding the domain/business
— Put the customer first
— Perseverance/Determination

We also identified a relationship between ability and second class citizenship in teams. Often team members, sometimes testers, with less ability can be given less challenging tasks within a team, this can create a “class system” within the team. An example from the testing world would be “Lets get the testers to do the regression testing, we’ve got more important things to do.” This in turn becomes a self fulfilling prophecy, rather than bringing these “second class” citizens into “first class citizenship” we further push them away.

It was a really fascinating debate but upon reflection I was disappointed at how i struggled to clearly articulate my position. Over the past year I’ve worked in a team were everyone is treated with the same level of respect and valued equally(or at least that’s my perception) so i rarely have to justify my position in the team.

I’m planning to spend sometime thinking about how in future i can more clearly articulate why testers are not second class citizens.

I’d be interested in hearing other thoughts on the debate topic, either those for or against. Please leave a comment or tweet me!