Vote for the Question of the Week! (Jan 2017 #4) [duplicate]

Question

It's time to choose a Question of the Week! The Question of the Week demonstrates what a good on-topic question looks like. Last week, we picked:

• Is there a programming paradigm that promotes making dependencies extremely obvious to other programmers? by Christs_Chin
• How to specify a precondition (LSP) in an interface in C#? by Achilles

Suggest the next Question of the Week by answering here with a link to the question, and a short explanation why it's an exemplary on topic question. You may suggest any question that was posted between 2017-01-22 and 2017-01-28 and is not closed. You can use this search query to quickly find good candidates.

Vote on suggestions to select the next Question of the Week. Ask yourself:

1. Is this question firmly and unambiguously within the scope of Software Engineering?

2. Is this an interesting and well-written question?

3. Would I like to see more questions like that?

The next QotW will be the suggestion with the highest score by next sunday. If you have doubts about a suggestion, you can discuss it in the comments.

Vote for this meta question if you like the Question of the Week contest. This makes the contest more visible in the community bulletin on the right. It also indicates whether this contest should be continued. After this week, I'll open up a discussion on whether we want these contests to continue, and what could be improved.

Answer 1

How to correct a mistake in the test, after writing implementation

An interesting question about how to “recover” from when the ideal TDD cycle goes wrong.

Usually, TDD is only presented in terms of “best-case examples”. Maybe, there will be a mention of what not to test, or when to skip the green-refactor bit when it is really obvious how the fully-refactored feature-complete implementation should look like.

However, most books and tutorials are very light on what to do when the TDD cycle goes awry. What if a test doesn't test what you think it does? Or, as is the case in this question: the test is simply wrong?

Not only is this question a good question, it is also a question whose answers expand the sum total knowledge about software engineering in an area where not much (if any) knowledge was previously recorded.

Answer 2

Large class with single responsibility

Aka. “I have a god object, how can I cope with the complexity?” Design reviews are considered on topic, so this kind of question is absolutely acceptable. Luckily for us, OP didn't show us the whole 2500 line class but only a representative sample, and has focused the question on a single topic: how such a class can be made to obey the single-responsibility principle.

Answer 3

When should dependencies be updated?

Configuration management is one of the central topics in the SDLC, and this question is about a burning issue for many development teams. Moreover, this is nothing you can easily read in a text book, a good answer needs some real world experience. Though the topic seems to be broad and to some degree opinionated at a first glance, the question can be answered in a very sensible manner in a few paragraphs.

Answer 4

What are the actors in a backend server use case?

A short, simple question about creating good use cases. And that makes it perfectly on topic.

Answer 5

Professional way to produce a large problem without filling up huge arrays: C++, free memory from part of an array

Though worded in terms of "C++", this is clearly an conceptual question about a real problem the OP faces, mostly independent from the programming language.The topmost answer shows there is a non-trivial, but canonical solution for it. This is the kind of question this site is ideal for.