Como você usa padrões e princípios de design para evitar dívidas técnicas no desenvolvimento de software?

For an application to be extensible and avoid technical debt ,design patterns should be incorporated at an initial stage. Also any third party integration should not be coupled with vendor specific implementation details. Few good practices - Adapter pattern - caching example. Implementing an abstract cache storage so that cache adapter of different flavors (redis, memCache, in memory) can be implemented and relevant one injected using a DI framework. - Strategy pattern allows us to extract the algorithm or logic in seperate classes which can be applied at run time based on condition. This improves code mantainability and testability. -Template pattern for processes having same steps but different logic . Code reusability is achieved.

"avoid technical debt" is a strong statement. Technical debt is necesary and is OK if it is managed. On design patterns, I don't recommend using them unless you're solving a problem (or have a context) for which the pattern is the appropriate fit. Instead use the Four Rules of Simple Design, especially the Simple Design Dynamo. It's OK. :)

Choosing appropriate design patterns is a key strategy to avoid technical debt in software development. Here are some tips: Understand the Problem: Before choosing a design pattern, make sure you fully understand the problem you’re trying to solve. Use Established Patterns: Leverage established design patterns that have been proven to solve certain types of problems. Don’t Overengineer: While design patterns can be powerful, it’s important not to overuse or misuse them. Keep Code Clean and Maintainable: Follow good coding practices such as writing clean, simple, and readable code. Refactor Regularly: Regular refactoring can help keep technical debt in check.

Some use cases where they can be applied - If there is a need to add functionality to a core module . In this case Open/Closed principle can be applied by inheriting from the core class or using the decorator pattern to enhance the functionality to reduce risk of regressions - If there are classes which implement interfaces with few methods that are not needed Throwing Not Implemented Exceptions for multiple methods are indicative of the fact that the interface has become unwieldly and become a candidate for Interface segregation . - If there are a few God classes SRP can be applied by defining responsibilities clearly . - If modules are not testable due to concretions use Dependency Inversion

Do you know that "Refactoring to Patterns" is a book that talks exactly about that? I really wish the LinkedIn AI was modest enough to mention external resources instead of trying to make people dependent on the platform… I really recommend to every developer to read more classical books and be less in TL;DR mode.

For code refactoring to be a success following must be kept in mind - Test Coverage Ensure the module being refactored has a high code coverage using unit and integration tests. Try to keep signature of public methods intact so that test code can execute as is. - Refactoring goals Do not refactor just due to preference for a particular coding style e.g. linq queries over foreach. There should be a tangible benefit like increased readability of code, simplified debugging and better extensibility. - Professional judgement Static Analysis Tools and design rules can be overwhelming at times. In some cases, adopting a particular rule can cause reoccurrence of another. Use your professional expertise in such cases.

As others have said, be sure the tests have your back. Consider mutation testing to test your tests. In terms of refactor, minimise duplication and improve naming.

Code reviewers should have a holistic look at the module being reviewed . Some examples. Are the cross cutting concerns such as Authorization addressed ? Are security parameters (OWASP, static analysis issues) addressed ? Is logging effective ? Are relevant log levels implemented with proper information ? Have unit tests being added with expected code coverage ? Is there sufficient information for users to be able to consume this feature ? Is there scope to move some functionality to a utility component for reuse ? Is instrumentation provided for monitoring and observability of new features ? Others items such as performance impact evaluation, accessibility and i18n, l10n. The review should add value and reduce technical debt .

Code review may include: Is the code consistent with the design pattern, system design, or architecture of the system? is it modular or componentized? is it readable enough to be comprehended easily? is it in line with the consistent database design or free of duplicity or performant enough?

Code reviews are better than nothing, pair programming is the best you can do. Code reviews are based on asynchronous communication and are downstream. A lot of time and exchange is lost here. Pair programming, if done correctly, improves the software design immediately and without delay because different perspectives are taken into account.

Documenting architecture via diagrams created by a professional diagramming tool will help various stakeholders in visualizing the architecture better. Popular Diagram types - Class diagrams details the structure of the classes and the relationship between other classes - Sequence diagrams are used to show interactions between various components - Dataflow diagram(DFD) show dataflow between various entities - Deployment diagrams which show physical architecture . Use of cloud specific symbols will provide deeper context.

Documentation comes from document, not diagram. Every software design is a compromise and optimizes certain features and goals at the expense of others. This is often not really understandable for other developers who work with the code at a later stage. Diagrams can only convey this with difficulty or not at all. I have had bad experiences with diagrams in this respect. Written comments, on the other hand, are very good.