Comment déboguer facilement du code complexe ?

The debugger provides great services, but it is very important to understand that its effectiveness relies on two very important points: first, a great rigor on the part of the debugger's user (tests show that there is a problem, the debugger helps to find and explain why), and second, well-written and especially, very well-documented code. It's like with tests, if we don't know what the code is supposed to do, we don't have a point of comparison to identify that it is not doing, at that specific point, what we expect from it.

To debug complex code effectively, start by understanding the problem and breaking it into smaller parts. Use logging, print statements, and debugger tools to identify issues. Employ unit tests to validate individual components. Utilize breakpoints and step through code to trace execution flow. Collaborate with team members for fresh perspectives. Leverage version control to isolate changes. Document your findings, and iteratively refine the code. Embrace a systematic and patient approach, focusing on one issue at a time for efficient debugging.

Rubber ducking. This is a technique where you would discuss the code line by line with an inanimate object (like a rubber duck). By doing so you will often discover what the error is as your explanation will not match what the code says.

Using a debugger is crucial for effective debugging. It enables you to inspect code in real-time, set breakpoints, analyze variables, and understand code flow. Popular debuggers include gdb, pdb, and Visual Studio Debugger for various languages. Integrated development environments (IDEs) like PyCharm and Visual Studio Code often provide user-friendly debugging interfaces, enhancing the debugging process.

Using a debugger in an http request context can be extra challenging. Reproducing the request in an end-to-end test first can provide a more robust debugger experience. It allows for an easy reset of sometimes complex initial conditions. It also prevents accidental debugger disconnects that can happen in live reloading local web servers. Finally, as a side effect, you have lowered the bar to make this a solid addition to your test suite.

For code that you write from scratch, THINK. Think carefully about the overall algorithm, the specification for each function, and all the corner cases. Write an invariant for each loop. Debugging is much easier if one knows exactly what each function should do. Unfortunately, most software engineers must debug code that has been created. modifed, and extended by others over time, which makes testing and debugging extremely difficult. When making modifications, first create a series of tests that establish a baseline. Then, after adding your changes, rerun the tests to ensure that the baseline outputs remain the same. Debuggers can help track down small details, but I prefer printf (or kprintf for operating system debugging).

When debugging complex code, having tests is almost crucial, as they enable you to mock output from one module to the other for example, which helps to narrow down your hypothesis about what could trigger the bug. In higher level tests, you can add your own testing logic and try to test for different possibilities when the code is complex in runtime as well, e.g. when you have lots of threads running in parallel and the bug is stochastic. In such cases tests can be run for a certain number of times with a certain set of inputs to quickly reproduce the bug and generate clues for you to fix them.

Express errors in code with automated tests that fail. Once you have a failing test, change the code so that the test passes. The collection of automated tests, if run periodically and whenever any part of the code changes, will confirm the code is operating without errors.

Writing tests is essential for ensuring code reliability. Tests, like unit tests, integration tests, and system tests, validate different aspects of your code, preventing and identifying errors. Testing frameworks such as unittest, JUnit, and NUnit facilitate the creation and execution of tests, promoting a robust development process.

Integrating testing into your CI/CD pipeline can greatly enhance the quality and reliability of your software. Automated tests can be run every time changes are made to the codebase, ensuring that new code integrates smoothly and doesn't break existing functionality.

There are several approaches that can be effective for debugging complex code issues during the development process: Adding sysout: This is a simple and straightforward approach where you can add sysout statements at various places in the code to see the values of variables and the flow of execution. Using a debugger: A debugger allows you to step through the code line by line and set breakpoints to pause the execution of the code at a specific point. Logging: Logging is a way to record information about the execution of a program. This can be useful for identifying specific points of failure. Code review: Sometimes, the best way to debug code is to have someone else review it. Often, a combination of these methods will be most effective.

Importance of logging and print statements can be understood hugely by people who are stuck in an on-call issue in a highly scalable application say Hotstar streaming WorldCup finals. Just remember a logging statement can help you with pulling your hair out on a P0 issue :) Stay safe and make logging a part of engineering practice in your team.

While debugger is helpful to debug in your local environment, logging is very crucial to catch errors in production. Print statements write messages to standard console and are used to debug locally. On the other hand, loggers generally log the message in a file and provides different log levels and formatting. You can log various things: 1. Stack trace: Logging the stack trace helps in understanding what was the error, why it occurred and how it can be corrected. 2. Method arguments and return types: This helps to ensure if the intended data is passed to the function and whether the function is returning desired result. 3. Just a message to ensure that a particular function was called or a piece of code got executed.

Logging is a good tool and if used wisely can help debug the code faster. Differentiated logging i.e. ability to see different levels /details of the debug logs is also a good thing here. e.g. to see memory dumps or say just the output of various print statements, we can use different levels that can be defined at runtime i.e. while running the module.

I would add a caveat about not shipping debug logs to production unless its a highly experimental feature and there is no other good way to get the same information. Another pointer I would add is to configure a chosen logging framework/library to format the log lines in uniform way and to show information which beginners sometimes omit - Time stamp - Level - Logger name. A whole section could be written on how to manage logger names. - Version of code being run, such as git sha. - file name and line number from where the message was emitted.

1 thing I follow is ‘unit-of-work’. Keep a single responsibility of a unit of code. It could be a method or a simple code block. It is more about architecture and design of code. When accompanied by debugger tools, it gives a clearer understanding and purpose of what this unit is supposed to do. Debugging becomes quite easy. In a complex code structure, we can always backtrack the dependants and eliminate the usual issues on single unit of code.

Code organization is a both a necessary discipline for bug prevention and a useful tool in finding and squashing existing bugs. In conjunction with succinct logging and thorough tests, having your code base organized into atomic modules—where each file only contains the code related to one clearly defined purpose—helps to isolate where errors are occurring. It also reduces the amount of code to read through to find the source of the error. Sometimes, I have eliminated bugs simply by splitting one module into two or more separate modules, each cohesively containing code pertaining to a specific task or objective. They say bugs find their way into cluttered spaces, so it wouldn’t hurt to keep your code clean.

Apply the principles of Clean Code to make your code more maintainable and easier to debug. Techniques like breaking down functions into smaller, more manageable pieces and writing self-documenting code can simplify the debugging process

Tenho um mantra que acho que combina com isso, algo que tenho comigo desde a época da faculdade: Um software deve sempre "buscar" baixo acoplamento e alta coesão. Basicamente quer dizer que devemos isolar as "coisas"/estruturas bem para que a manutenção seja mais simples e que você tenha que realizar alterações no mínimo de lugares possíveis.

It's very helpful to make your functions pure whenever it's possible. This will make them much easier to test and isolate when debugging. Sometimes when I work with unfamiliar code, I'm trying to refractor this code to understand it and to make it cleaner

Follow Rubber duck method - Try explaining your code to rubber duck - Go over all the presumptions and explain why you did each statement as simple as for loop - Question your logic and assumptions - Voila! soon you will find the issue and problem solved :)

Unfortunately, not everyone would appreciate the value of seeking help from the community. If you're stuck on a problem without good documentation, chances are others are too or will be. So, reaching out on places like GitHub Issues or Stack Overflow, and sharing the details of your issue and what you've tried, not only gets you help but also benefits other developers facing the same problem in the future.

Def. seek feedback will help and improve our skills. Receiving feedback has tremendous benefits such as identifying errors, code exploration, data integration, management, collaboration, communication, and more.

Another pair of eyes on the problem can definitely be helpful. Sometimes just explaining the problem can help. The main point though is to gain as much understanding as possible before a solution is implemented. Focus on what the problem is (why is it a bug?) and what's potentially causing it - be sure you can explain the problem to someone with limited knowledge. Be sure to stick to the facts and not your interpretation of them - this is the biggest challenge for most people. Try your best to know how to reproduce the issue and identify modules associated with the problematic behavior.

In my journey as a developer, seeking help and feedback has been invaluable in debugging complex code. On one project, I was stuck on a particularly challenging issue. After sharing my dilemma in an online forum, I received suggestions that led to a breakthrough. This experience taught me the power of collective knowledge and diverse perspectives. Platforms like Stack Overflow, GitHub discussions, and even informal chats with peers have often provided fresh insights. It’s important to be clear and specific when seeking help, detailing the problem and what you’ve tried. Respectful engagement not only aids in finding solutions but also builds strong professional relationships and a supportive community.

Don't write complex code. One way of doing this is to write everything two or three times until you get the hang of it (various case studies in the past including an early one from Bell Labs). If its somebody else's complex code, consider throwing it away. It's unlikely it does what it was supposed to do assuming anybody knows. I don't think I have ever seen a complex buggy piece of code successfully recovered by patching. It usually requires major surgery.

Para ajudar a depurar código complexo de forma eficiente algumas dicas podem ser seguidas: - Quebre o problema em partes menores e teste cada uma ("dividir para conquistar"). - Faça diagramas e anotações para entender a lógica (a parte visual é, muitas vezes, mais fácil de entender que só ficar olhando para linhas de código). - Compare com códigos similares que funcionam (aprender com outros exemplos é um bom caminho para encontrar erros). Pense de forma algorítmica e siga o fluxo linha a linha (tenha uma visão local). Foque nos indícios e erros em vez de procurar por tudo (seja estratégico e sistemático). Tenha paciência e persistência (a solução virá com o tempo).

I've always found having a curious mindset when debugging helps to build a more complete picture of what the code is doing. Having a complete picture of the code will not only lead to successfully debugging an active issue, but you may find some other areas to improve or harden the code. Fix the problem ASAP, then go back and explore the code more. Being curious when debugging means, when you see a function being called, you dive into that function to see how it works, and you keep diving in when you see something that you don't understand. This mindset will not only help you when debugging, but it will also help you exercise the muscles you need to use when you're starting a new job and trying to understand your new code base.

There are undoubtedly many tools that can help avoid debugging: software architecture, documentation, code review, 'lean programming' approach, 'test-driven development', and most importantly, a very good understanding of algorithms. When programming, you are driving the processor; it's you who must adapt to what it is capable of doing, not the other way around. This, therefore, requires a good knowledge of its operation (especially for memory exchanges, calculation precision, etc.) because sometimes the code is correct, but its execution does not allow achieving the goal due to the limited functioning of the machine.

To me, the very first and most important task in debugging is to reproduce the error. Some bugs are obvious because they happen in a systematic manner. This kind of bug is of little interest and can be resolved very easily. The most difficult bugs are the ones that produce errors in a seemingly random or inconsistent manner. In this case, debugging becomes an investigation in which you have to try variations on the parameters and exogeneous stimuli of your algorithm until you find the deterministic trigger(s). Once you have identified the deterministic cause, the problem is almost solved. I find debugging fun because it is a serious game that starts like a crime investigation, with the collection of evidences.