Cinq erreurs de débutants dans la conception de tests automatisés. Vous lisez probablement ceci parce que vous êtes intéressé par la conception de tests automatisés ou simplement parce que vous travaillez avec cette technologie. Mais avant de passer en revue ces cinq erreurs de débutant, permettez-moi d’abord de vous recommander de toujours vérifier le fournisseur de technologie. Les « tests » en général ont beaucoup évolué au cours des deux dernières années. À partir d’un modèle, vous pouvez bien sûr générer une suite de tests complète, mais vous pouvez également générer directement des scripts de test exécutables de manière entièrement automatique. Je ne suis pas là pour en parler, mais c’est vraiment une recommandation : Vérifiez la technologie de votre fournisseur 🙂
Commençons par nos cinq erreurs de débutants.
1. PARTIR DE RIEN
Je sais qu’une nouvelle technologie est toujours amusante et que certains d’entre vous veulent se lancer dans cette nouvelle façon de travailler. Après avoir été frustré par la rédaction de tonnes de documents Word pour les tests manuels, vous souhaitez consacrer du temps à la modélisation à partir de zéro.
Mais posez-vous toujours la question :
Que puis-je réutiliser pour être plus rapidement sur la bonne voie ?
Parce que oui, nous pouvons réutiliser les actifs des tests et des exigences !
Par exemple : diagrammes Visio, enregistrements, cas de test manuels existants, plans de test, fichiers WSDL, diagrammes BPMN, etc.
So don’t waste your time reworking something you’ve already done. Instead, import it!
2. TOO DETAILED MODELING OF INTERFACES
Modeling in general is fascinating. In a simple language, you can express all the interfaces that your application provides for testing. But think carefully: Do you really want to model every possible detail? So my second piece of advice is:
Think AGILE
Target only the most important interfaces of your application: what brings the most value? Tell yourself that if you have time, you will improve your diagram and go into details later. Don’t target everything. Remember that 20% of your work will bring 80% of the value. And 80% is still much better than the old way of working!
3. EXCESSIVE USE OF COMBINATORIAL DATA
This point is something I see a lot. You have a lot of possible data that you want to test. Do you really want to test your 500 logins and passwords as part of functional testing? Are you really interested in running tests that test every combination of your 4 form fields that can each have one of 10 possible values? In other words:
Do you really want to run 10,000 possible tests?
Yes, it may sound funny when you read this, but many new users do this. So always try to take a value that makes sense and don’t test every value you can test. You should always ask yourself questions about this. What should I include to test the functionality of the SUT? – Probably not every combination of data.
4. MODEL THE ONLY HAPPY PATHS
Most users, when they start modeling, only model happy paths. That’s a good thing, but why not model alternative and invalid behaviors?
By modeling alternatives and invalid behaviors, this is where the tool generates test cases that you can’t think of! So, ask yourself [what would happen if ….]? The tool will generate the corresponding test case for you.
5. OVERUSE OF CUSTOM MODELING EXTENSIONS
Let me explain what a custom template extension is: it is a way to extend your templating language, for example with a custom action, when you are stuck in a situation where no predefined action provided by your templating language « seems » appropriate.
Here the important word is « feels » in this sentence. When you decide to create a custom modeling extension, you need to be 100% sure that there is no other way to achieve it. Introducing such an extension as part of a library that can be reused not only for a specific purpose, but also in different diagrams or even in different projects, solves this problem much more elegantly, as it brings readability and consistency and allows a direct transition between generated tests and automatic execution.
I hope these 5 points will free your mind and bring you added value during your modeling. In conclusion, modeling changes a lot of things in the way we work. It is always good to question and discuss the way you work and how to improve modeling within your team. All this will put you on the path to becoming a JEDI TEST DESIGN MASTER!
May the Modeling Force be with you and I hope you never make the five beginner mistakes again!
Thanks for reading! Please feel free to share your comments with us!
