An unpopular opinion: Get rid of Pull Requests! - Eugene Romero - NDC Oslo 2024

Summary

Eugene Romero's lightning talk addresses the common frustrations with pull requests (PRs) in software development. Drawing on his extensive experience, he highlights issues like the tendency for PRs to become overly large, delays in approvals, and shifting responsibility between coders and reviewers. Romero advocates for a mindful approach, emphasizing that not all changes require a PR and that keeping PRs small enhances clarity and review efficiency. He encourages teams to regularly assess their processes to ensure they remain valuable, concluding with a reminder to embrace flexibility in their development rituals.

Takeaways

• 😀 Pull requests (PRs) can become too large, complicating the review process for both authors and reviewers.
• 😀 Many developers experience 'LGTM syndrome,' approving PRs without thorough understanding, leading to potential issues.
• 😀 Delays in PR approvals can hinder deployment, affecting the urgency of changes needed in production.
• 😀 Responsibility shifting occurs, where authors and reviewers blame each other for code issues, complicating accountability.
• 😀 Keeping pull requests small is essential for effective reviews and maintaining high code quality.
• 😀 Not every change requires a pull request; simple changes may not need formal reviews, saving time and effort.
• 😀 Teams should regularly evaluate the necessity and effectiveness of their pull request processes to enhance productivity.
• 😀 Encouraging a culture of flexibility allows teams to adapt their practices to better suit current needs.
• 😀 Avoid waiting too long on open branches; PRs should ideally represent one or two days of work.
• 😀 Questioning established practices can lead to improved workflows and better collaboration among team members.

Q & A

What is the main topic of Eugene Romero's talk?

The main topic is the challenges and inefficiencies associated with the pull request process in software development.

What humorous term does Eugene use to describe quick approvals without thorough reviews?

He refers to this as 'Stevie Wonder approvals,' implying that reviewers approve changes without fully understanding them.

What common issues does Eugene identify with pull requests?

He identifies several issues, including large pull requests being hard to understand, delays in approvals, and responsibility shifting between developers.

How does Eugene suggest improving the pull request process?

He suggests keeping pull requests small and manageable, ensuring that only necessary changes go through the full review process.

When does Eugene believe a pull request is necessary?

He believes a pull request is necessary for new functionality, potentially destructive changes, and changes that are difficult to roll back.

What are some situations where Eugene argues pull requests may not be needed?

He argues that pull requests may not be needed for simple or repetitive changes, when there are no other experts available for review, or for easily reversible changes.

What principle does Eugene emphasize for maintaining pull requests?

He emphasizes the principle of keeping pull requests small, suggesting they should ideally represent only one or two days of work.

What is the potential consequence of having large pull requests?

Large pull requests can lead to prolonged review times, decreased code quality, and may cause reviewers to simply approve them without adequate understanding.

What does Eugene encourage teams to do regarding their development practices?

He encourages teams to regularly reassess their practices to ensure they still provide value, advocating for a mindset of questioning established rituals.

How can developers ensure their pull requests remain manageable?

Developers can break down changes into smaller parts, use short-lived branches, and implement feature gates to facilitate easier reviews.