Vous êtes confronté à des délais serrés pour les nouvelles fonctionnalités. Comment maintenir la qualité du code sous pression ?

Understand the task and requirements clearly, and define the scope of the task precisely. After that, it is advisable to have a checklist to review the status of the work.

To efficiently meet user or company needs, prioritize existing solutions before developing from scratch. If a library fulfills your requirements, utilize or adapt it to save time and resources. For unique demands, break the project into smaller timelines, holding weekly meetings with clients or product managers to ensure alignment and timely delivery. Aim to produce an MVP by the deadline, ensuring the core features are in place. When the project scope is clear, implement tests to establish a robust foundation, facilitating easier future modifications and ensuring long-term success.

Prioritize the essentials, communicate with stakeholders, and follow through. Make a to-do list which focuses on minimum viability. Explain to project stakeholders what's possible in terms of delivery and what will need to be deferred until a future release. Deliver immediate value consistently and communicate progress. Identify blockers ahead of time if possible and if they are already blocking your path forward call everyone together to find a way forward. Once everyone is unblocked check-in regularly to get status updates. Keep everyone on track and be the driver of ways forward and unblocking of the team.

It is advisable to have a coding convention for the entire team to follow, and you can use third-party tools to check for code smells, etc. In summary, there should be a coding standard that specifies code formatting, log formatting, and other related practices.

As a team lead, it's crucial to establish clear ground rules and conventions for consistency in the codebase. Standardize code structure and naming conventions to enhance readability and collaboration. Implement tools like SonarLint in IntelliJ for Java to automate these checks. Regular code reviews are essential for maintaining quality, allowing team members to catch errors early and share knowledge. Emphasize code optimization and adherence to design principles to ensure clean, cohesive, and loosely coupled code that remains robust and adaptable over time, even under tight deadlines.

Automation testing is a crucial step that cannot be overlooked for quality control when time is limited. Additionally, it is advisable to include some unit tests for happy cases to ensure that functionality continues to work well in these scenarios when the source code changes.

Make the LLM write your tests from the client side using tools like puppeteer to drive browsers or similar clients that can tests how new code behaves. Test-driven-development or Behavior-driven-development in general is often the most efficient development workflow when using AI copilots or LLMs to generate code. If you're not using AI and LLMs to generate code you will be very slow and perhaps not meet expectations time-wise since almost everyone else is doing it. The old thinking is: Writing tests takes too long and we just need to get this thing out the door right now. The new thinking is: LLMs are gonna write code for you and 10x the generation of code, but you need to test the behavior to make sure that you get consistent behavior.

This one thing is something that i have done so many times despite that no product manager is agreed on to give you time to do so , so you have to find times for yourself cause you and your team are the ones that are gonna add or change features to the code base and not the CEO or the non technical guys which only see the delivery of the features and nothing more so despite the common believe of if it works don't touch it, actually i shall say refactor small portion of your code if you are confident enough that you are not gonna break something in the software or having downtime in the time of deployment .

In my view, this step can be applied from the beginning if you are implementing a new feature or project. However, if it is a subsequent phase, it might not be necessary to apply it right away, as all team members might not be familiar with it yet, which could lead to longer coding times and potentially affect quality.

Refactor as you go by not allowing developers to create friction through rigidity. Refactoring shouldn't be a project with a lot of sponsorship around "maintainability." In most cases, this so-called "maintainability" is just a waste of resources because it's a premature optimization around yesterday's development workflows. Code-maintainability should be assumed. You can prompt an LLM to write idiomatic code, for example. Code is inherently more disposable simply because it's 10x easier to generate with LLMs which understand natural language and programming language better than people do.

Reduce tech. debt in a simpler way. You a tool an metrics to constantly measure it. You might not be able to do much with the existent one, but you can do two things at least: stop adding new one (the tools helps with that), and analyze the existing debt, and only work in anything that can be related with the new features. This strategy will help you have a better code, while at the same time, will ensure maintainability and enhance quality.