Welcome to my space where I will share with you technical development guides and software project management experiences.
I have been a software engineer specialising in quality assurance since 2006.
Join me over the next few months as I unpack my analysis on our industry and bring to light some of the burning issues that cause projects to fail and ultimately have put some businesses at risk.
Count property should match rows.length, Count must be greater than 1
should filters resources
GET
Call endpoint with filter parameters (limit, sort, start, filter)
200
Count property, rows.length, id of first and last resource
should return a specific resource
GET
Call endpoint with a resource ID
200
Check each property
should return a 404 if resource not found
GET
Call endpoint with a fake resource ID
404
should create a resource
POST
Send full valid data
201
Check each property
should fail returning all mandatory properties
POST
Send a single non mandatory property
400
Count number of errors
should fail if …
POST
“Send data against business logic (null value, blank value, unicity, shorter than expected, bad relation …)”
400
Check reason/code of error
should update the resource
PATCH
Send full valid data (set a property id which should be ignored)
200
Check each property
should fail if …
PATCH
“Send data against business logic (null value, blank value, unicity, shorter than expected, bad relation …)”
200
Check reason/code of error
should return a 404 if resource not found
PATCH
Call endpoint with a fake resource ID and send full valid data
404
should delete the resource
DELETE
Call endpoint with a resource ID
204
If hard delete, check if the resource doesn’t exist anymore in DB. If soft delete, check the resource has a deletedAt value not null
should delete the resource
DELETE
Call endpoint with a fake resource ID
204
Detailed Test Scope
Where {id} is a UUID, and all GET endpoints allow optional query parameters filter, sort, skip and limit for filtering, sorting, and pagination.
#
Test Scenario Category
Test Action Category
Test Action Description
1
Basic positive tests (happy paths)
Execute API call with valid required parameters
Validate status code:
1. All requests should return 2XX HTTP status code
2. Returned status code is according to spec: – 200 OK for GET requests – 201 for POST or PUT requests creating a new resource – 200, 202, or 204 for a DELETE operation and so on
Validate payload:
1. Response is a well-formed JSON object
2. Response structure is according to data model (schema validation: field names and field types are as expected, including nested objects; field values are as expected; non-nullable fields are not null, etc.)
Validate state:
1. For GET requests, verify there is NO STATE CHANGE in the system (idempotence)
2. For POST, DELETE, PATCH, PUT operations – Ensure action has been performed correctly in the system by: – Performing appropriate GET request and inspecting response – Refreshing the UI in the web application and verifying new state (only applicable to manual testing)
Validate headers:
Verify that HTTP headers are as expected, including content-type, connection, cache-control, expires, access-control-allow-origin, keep-alive, HSTS and other standard header fields – according to spec.
Verify that information is NOT leaked via headers (e.g. X-Powered-By header is not sent to user).
Performance sanity:
Response is received in a timely manner (within reasonable expected time) – as defined in the test plan.
2
Positive + optional parameters
Execute API call with valid required parameters AND valid optional parameters
Run same tests as in #1, this time including the endpoint’s optional parameters (e.g., filter, sort, limit, skip, etc.)
Validate status code:
As in #1
Validate payload:
Verify response structure and content as in #1.
In addition, check the following parameters: – filter: ensure the response is filtered on the specified value. – sort: specify field on which to sort, test ascending and descending options. Ensure the response is sorted according to selected field and sort direction. – skip: ensure the specified number of results from the start of the dataset is skipped – limit: ensure dataset size is bounded by specified limit. – limit + skip: Test pagination
Check combinations of all optional fields (fields + sort + limit + skip) and verify expected response.
Validate state:
As in #1
Validate headers:
As in #1
Performance sanity:
As in #1
3
Negative testing – valid input
Execute API calls with valid input that attempts illegal operations. i.e.:
– Attempting to create a resource with a name that already exists (e.g., user configuration with the same name)
– Attempting to delete a resource that doesn’t exist (e.g., user configuration with no such ID)
– Attempting to update a resource with illegal valid data (e.g., rename a configuration to an existing name)
– Attempting illegal operation (e.g., delete a user configuration without permission.)
And so forth.
Validate status code:
1. Verify that an erroneous HTTP status code is sent (NOT 2XX)
2. Verify that the HTTP status code is in accordance with error case as defined in spec
Validate payload:
1. Verify that error response is received
2. Verify that error format is according to spec. e.g., error is a valid JSON object or a plain string (as defined in spec)
3. Verify that there is a clear, descriptive error message/description field
4. Verify error description is correct for this error case and in accordance with spec
Validate headers:
As in #1
Performance sanity:
Ensure error is received in a timely manner (within reasonable expected time)
4
Negative testing – invalid input
Execute API calls with invalid input, e.g.:
– Missing or invalid authorization token – Missing required parameters – Invalid value for endpoint parameters, e.g.: – Invalid UUID in path or query parameters – Payload with invalid model (violates schema) – Payload with incomplete model (missing fields or required nested entities) – Invalid values in nested entity fields – Invalid values in HTTP headers – Unsupported methods for endpoints
And so on.
Validate status code:
As in #1
Validate payload:
As in #1
Validate headers:
As in #1
Performance sanity:
As in #1
5
Destructive testing
Intentionally attempt to fail the API to check its robustness: Malformed content in request
Wrong content-type in payload
Content with wrong structure
Overflow parameter values. E.g.: – Attempt to create a user configuration with a title longer than 200 characters
– Attempt to GET a user with invalid UUID which is 1000 characters long
– Overflow payload – huge JSON in request body
Boundary value testing
Empty payloads
Empty sub-objects in payload
Illegal characters in parameters or payload
Using incorrect HTTP headers (e.g. Content-Type)
Small concurrency tests – concurrent API calls that write to the same resources (DELETE + PATCH, etc.)
Other exploratory testing
Validate status code:
As in #3. API should fail gracefully.
Validate payload:
Validate headers:
As in #3. API should fail gracefully. As in #3. API should fail gracefully.
So, there you are… having a UI and an entire suite of manual test cases. These tests are tedious, take forever and lets be honest…can get repetitive and boring!
My friend, you are in the frontline for automating your testing!
Tools
Lets not beat around the bush – Selenium has been around for ages. The product has had millions of users, has become a W3 standard and is launching selenium 4 pretty soon.
This is the tool for us.
There are others yes, most of which will wrap some selenium WebDriver capability into a pretty package and sell you that at a pretty penny.
Lets not get the wool pull over our eyes.
We can do the same, implement at the same level and in fact, have far greater control on our test product ….cheaper, faster, better.
Ok, before we get butterflies in our tummies over this tool, there are some pitfalls…urhgggg, of course.
Not to worry, we have help – in our demo below, I will show you how I implemented Selenide – an open source project to fill the gaps that were obvious in Selenium 3.
* @author Suveer Prithipal
public class GoogleTest {
@Test
public void googlePageTest(){
/**
No need to create a webdriver instance!
Selenide provides with easy to use API's that provide rich functionality.
On a normal day with selenium, we would need to create a webdriver instance, and bind it to a browser.
We would also need to define page elements to use them.
Selenide removes the need to do this by wrapping up that into a singe API.
Below, we use "open" to create the webdriver instance, and bind it to a class.
Passing it a class, provides the shape for the instance, giving it methods and defined functionality.
*/
GooglePage googlePage = open("http://www.google.com",GooglePage.class);
/**
Now that we have an instance of webdriver up and we are on our test app, Google.
We can then search something.
Searching for something means it will return a result.
Therefore, we need a class to take the shape of these results.
*/
SearchResultsPage searchResultsPage = googlePage.searchGoogle("selenide");
/**
Tests.
Now that we have results, we can perform tests.
Below, we use the searchResultsPage and query the class for expected results.
*/
searchResultsPage.checkResultsSize(8);
searchResultsPage.getResults().get(0).shouldHave(text("Selenide: concise UI tests in Java"));
/**
Use page object models and design patterns
This example is to demo the ease of use with Selenide.
Its important to separate out your implementation for better maintenance, easy of reading and debugging.
*/
}
}
In Code – Get going in 3 easy steps
This project is written in Java, uses the page object model and is triggered by a BDD Cucumber feature file. Learn these terms well.
Lets get into it:
Step 1: Define our Page Objects
The things you want to interact with on a page like:
buttons
lists
labels
text inputs
Step 2: Step Definition
Step definitions are the glue that will bring a workflow to your elements and bind them your feature file.
Here we chain our actions together and feed them input using our feature file.
Eg, the loginCuke() method:
we open up the site in a web page using Selenide. A 1 liner which will handle a lot of the grunt of selenium behind the scenes.
we then proceed to run the login method that will:
take our input username and password from the feature file and process the login method.
Step 3: Create your feature file
Feature files are BDD scribed tests.
We write these in plain english and translate that to code…as we did above
Thats it! run your feature file and wait for the results.
You’ll see the browser opening and doing things. Screenshots are a default on failure with Selenide so you’ll have that too!
Reporting:
Reporting is a must in any project. Please have a look at my other content for more details or pursue your own.
Conclusion:
UI testing has been around for a while and is getting easier, cheaper and simpler to implement.
With the older versions selenium we had to code a lot to get a page to open, while today we can achieve this in 1 line with a very fluent api.
Write once, test 1000 times on all browsers
our tools allow us to test on all browsers and any version…. simultaneously.
A bless in the UI testing world as it allows us to drastically reduce our testing time and therefore our cost to service testing.
Writing our own frameworks gives us greater advantage in scale and capability.
Nothing is more fore-filling than overcoming a challenge by learning, trying and failing!
Having automated tests are great! … but not sharing the results or centralizing them is not so great. 😦
Your automated tests have undeniable business value, don’t be shy about it. Excellence does not happen overnight. So even if your tests are in early development, share the results with your team.
So what can you do?
Setup a nightly run using some CI tool. I like Jenkins 😉
Promote the results by using a dashboard, email or a tool like slack.
Address failed tests daily. Fix them or add them to a backlog. Fix these soonest.
Prep a backlog of test automation scenarios that the entire team supports and can contribute towards.
In conclusion:
Visibility and Reliability are really important. This makes them trust worthy, needed… valuable.
Pursue it viciously. Take small incremental steps towards implementation.
Docker is a great resource to use as we have the ability to spin up and environment for testing and then very quickly throw it away again.
OMG why would you do that?
Well that the power and presence of docker and containers. We can spin up, tear down and re-use this resource multiple times and we should do so, without attachment,
Docker support is amazing. there are heaps of predefined containers that just require us to pull and use instead of us having to write and maintain those scripts, environments and data.
Its like magic at your fingertips
Selenium Grid
Grid is an extension out from Selenium which give us the ability to run our tests remotely.
Lets look at some of the key benefits:
Run tests remotely
Run tests in different browsers with the change of a config
Run tests in parallel – our biggest win.
Reduce test time and therefore the feedback cycle.
Implementation
Start up the Selenium Grid
We’ll make use of the predefined selenium-docker containers.
This setup makes use of a dockerfile which will spin up 3 environments:
You are viewing the grid, and the child worker nodes it has available.
Thats it! your grid is up and running –
how damn easy was that!
But we’re not done. We need to change our test application code to hit this new environment.
Implement the Hook into the grid
To open the browser we need to establish some browser capabilities to start up our browser.
These are seen as ChromeOptions(),
We need to apply these setttings so that we can start up a browser on a linux terminal with no display adapter. We set the headless option in particular for this.
We then need to point our RemoteWebDriver to the new url,
We do so by setting the urlToRemoteWD variable, which point to the url of the grid.
Run your test
When we run our test, –
Our test is published to the master
The master will inspect which node is free and available and will push the test to run there.
The results are fed back to the master and then fed back to our test application for reporting.
You are ALL DONE and ready to run a battery of tests against AWS or locally, headed or headless.
While this worked for many years, we found limitations in processing performance and handling data in near real time but nothing I guess prepared our world for the
creation of the big data concept.
Storing data is always costly and usually the bulk of your software budget. To make thinks faster in this model, devs
had to either scale up the number of Db instances they had of use a multiple Db instances. This is a sore sore SORE point for member of our tech community, big pockets or small.
This entry point also made our applications the centre of attention and building our applications more complex.
Multi-threading and bad message handling was super key to ensure this was successful.
How do we manage the load?!
However it came at a cost.
it was challenging for devs
it made dev and testing complex
it bloated our codebases
it wasn’t always done right
issues were usually only spotted once these threads ran live and encountered the real world for the first time. We’re were now in the thick of things…
This gave rise to the idea of data streaming, which rise to Kafka!
Kafka and Data streams for the win
Yes, my eyes are still googlied <- legit
I love high performance applications. I mean, we all can build a hello world project,
but to build a highly performant, fault tolerant scalable and real time application that can process a million messages quickly
takes something hell of a special. (quick shout out to Reece and the SDP Platforms team )
A Kafka broker is just a running instance of kafka.
Its the hub, the mother ship the ring leader, the conductor, the centre piece of the show the thriller in Manila…ok, too far but you get my vibe.
TOPIC
Brokers are were we create topics for consuming.
A topic is really just a layer of business logic events. for eg, if we were in the business of baking bread,
a topic of interest would be:
“how long has the dough rested”
“is the oven hot enough to begin”
“how long has the bun been in the oven”
PRODUCER
A producer is one that will create or add to a topic. When an event is raised, the producer springs into action.
So in our case above, the oven would produce an event to the topics:
“is the oven hot enough to begin”
“how long has the bun been in the oven”
But what about the other topic?
The oven has no need to care about how long the bread has rested. It just cares if it itself is ready and if it has completed the job.
This then means the baker would be the only producer to the other topic
“how long has the dough rested”
CONSUMER
A consumer of a topic is one that is simply subscribes to a topic and is updated when something in that topic changes.
In our bread winner example above, our baker or bakers would the consumers of all these topics…well, we hope for bread sakes!
The Kafka value
So far I’ve explained some differences and some terms, but not the value.
As seen in the Kafka image above, we are now able to do a couple of things BEFORE our application and databases are hit.
BOOM! VALUE!
We now can:
filed multiple threads
compute at mad speeds
and store only when necessary
doing all this at near real time, almost seamlessly at lower operational costs.
MORE FOR LESS -> BOOM! EVEN MORE VALUE!
The real 11 herbs and spices are in the manner in which the data is sharded, partitioned, stored, processed and read. But that a technical walk though for another day, but dang
Its like the Oprah of big data processing eh?..You feeling my love for this now?
Confluent
With all nice things in this world, people cant wait to just onboard. thankfully so!
Working with streams and stream applications, especially from a QA point of view, can be challenging, merely on the fact that
there is a very, very small set of tools for us to use to validate and gain some peek preview of data under the hood.
The team of confluent have done a job of this, check them out here:
I guess being in QA for so long has really trained my brain to push past what even i as a developer would deem normal behaviour.
i.e. just testing the application as a consumer is not enough.
The REST client is awesome and lets me integrate my integration testing platform along side my dev team and application code.
As we produce topics and messages, I get to peer into these, even before consuming it.
I can validate
the producer
the message
the topic
the partitions
the success
and the failures
speed at which we are processing and latency between them. STUNNING!
Conclusion
I hope this has enlightened you a little on Kafka, streams and some tools you can use to simply your life.
We actually live in a world a streaming data…
NETFLIX
YOUTUBE
SPOTIFY
Your favourite ONLINE RETAIL store… you know those suggestions on “OTHER ITEMS YOU MIGHT LIKE”….. well…stream data…It’s all around you and I.
The introduction of tooling over Kafka is epic.
As a QA, this was imperative. I need to follow the data from start to end, leaving no bread unbaked.
It was not a simple undertaking for me.
I researched and I learnt a lot, tried loads of things and failed, leaned on people with experience to help me through and eventually found success.
All of these come neatly wrapped in docker containers. These can be run as a whole or as components. It makes working with queues a lot simpler by giving us hooks into our streams.
Dont let a subject like big data processing scare you. Get in there, get your hands dirty and fail! Fail and fail till you bake your bread.
Over the past then years that I’ve been in the industry, I have been fortunate enough to bear witness to the beautiful way software delivery methods have evolved.
From long lasting design of monoliths to quick and in some cases, daily deploys of micro-services, we have come a long long way but our journey is not done.
Today, we’re moving faster than ever into cloud, serverless, containers and SaaS.
The Revolution
The key to this revolution was communication. There is indeed, strength in numbers.
Knowing that we were not struggling alone. That our issues were not unique and that others solved them was comforting, motivating and enlightened us.
The Bitter Truth
We all became students and masters at the same time. We read, research, learn insatiably. GIVE ME ALL THE KNOWLEDGES!
Learning how to do things better,
Learning how others found success in similar situations! YAS!
We’re amped! We’re pumped, we try it out and its going to be amazing…and it fails!
Its hard…its not working…its taking longer than expected.
wait.. what?!
What just happened
Variables. Life just happened. It sucks, but its a lesson that was always yours…waiting for just you.
Project Risk
As a lead, take careful consideration of your projects strength and weaknesses.
These Risks will surely impact your delivery.
Focus on quality early
With the introduction of micro-services, we have seen testing get pushed left.
Like into the bush and we forget about it? Erhm…NO.
By pushing testing left, we attempt to push the top tiers of the pyramid down.
So an example here would be, refining our system tests to pose as integration tests instead.
The goal of this concept has always been the same, faster feedback.
TIP: Dont get bogged down by the concept but rather focus on the goal, shift can happen in both directions.
Have a repeatable game plan
Build a vision for delivery with your team.
Understand each move your people and software makes and draw a picture.
Create flow diagrams and maintain this as you and the team grows…making it your own and not an internet copy and paste.
Most of the time, everything you read will be wrong for you. OMG…
I just mean that when we read something, most of the time we dont have the full picture. The new start in your team or the exit of an old soul will dramatically change how you deliver. Shift and adapt with it.
TIP: Dont forget to address your delivery process often!
Below is an example and a good head start for you to build and share your flow. Make it your own!
Conclusion
As you forge forward in your ambitions to delivery high quality software sooner, build a plan that is lean but high value to your team.
)
(The Random Tester) -> 