The Pros and Cons of Four Kinds of Code Reviews

[article]

review. On the author's end, he has to figure out how to gather the files together. For example, if this is a review of changes being proposed to check into version control, the user has to identify all the files added, deleted, and modified, copy them somewhere, then download the previous versions of those files (so reviewers can see what was changed), and organize the files so the reviewers know which files should be compared with which others. On the reviewing end, reviewers have to extract those files from the email and generate differences between each.

The version control system can assist the process by sending the emails out automatically. The automation is helpful, but for many code review processes you want
to require reviews before check-in, not after.

Like over-the-shoulder reviews, email pass-arounds are fairly easy to implement. They also work just as well across the hall or across an ocean. A unique advantage of email-based review is the ease in which other people can be brought into conversations, whether for expert advice or complete deferral. And unlike over-the-shoulder, emails don't break developers out of "the zone" as they are working; reviews can be done whenever the reviewer has a chance.

The biggest drawback to email-based reviews is that they can quickly become an unreadable mass of comments, replies, and code snippets, especially when others are invited to talk and with several discussions in different parts of the code. It's also hard to manage multiple reviews at the same time. Imagine a developer in Hyderabad opening Outlook to discover 25 emails from different people discussing aspects of three different code changes he's made over the last few days. It will take a while just to dig though that before any real work can begin.

Another problem is that there's no indication that the review is "done." Emails can fly around for any length of time.

So:

    • Pro: Fairly easy to implement
    • Pro: Works with remote developers
    • Pro: SCM system can initiate reviews automatically
    • Pro: Easy to involve other people
    • Pro: Doesn't interrupt reviewers
    • Con: Usually no verification that defects are really fixed
    • Con: How do you know when the review is "complete?"
    • Con: Impossible to know if reviewers are just deleting those emails
    • Con: No metrics or process measurement/improvement

Pair-programming (review)
Most people associate pair-programming with XP and agile development in general. Among other things, it's a development process that incorporates continuous code review. Pair-programming is two developers writing code at a single workstation with only one developer typing at a time and continuous free-form discussion and review.

Studies of pair-programming have shown it to be very effective at both finding bugs and promoting knowledge transfer. And some developers really enjoy doing it. (Or did you forget that making your developers happy is important?) There's a controversial issue about whether pair-programming reviews are better, worse, or complementary to more standard reviews. The reviewing developer is deeply involved in the code, giving great thought to the issues and consequences arising from different implementations. On the one hand, this gives the reviewer lots of inspection time and a deep insight into the problem at hand, so perhaps this means the review is more effective.

On the other hand, this closeness is exactly what you don't want in a reviewer; just as no author can see all typos in his own writing, a reviewer too close to the code cannot step back and critique it from a fresh and unbiased position. Some people suggest using both techniques — pair-programming for the deep review and a follow-up standard review for fresh eyes. Although this takes a lot of developer time to implement, it would seem that this technique would find

About the author

Jason Cohen's picture Jason Cohen

Jason Cohen is the author of Best Kept Secrets of Peer Code Review, the definitive work on the new process of Lightweight peer review with 7,000 copies in circulation. Jason conducted and published the largest case study of peer code review at Cisco Systems®—2,500 reviews over 10 months with a real, delivered product. Jason is also the founder of Smart Bear Inc., makers of Code Collaborator that assists developers and managers with lightweight peer code review processes. Jason has spoken at many conferences, trade shows, and ACM/IEEE society meetings across the country.

AgileConnection is one of the growing communities of the TechWell network.

Featuring fresh, insightful stories, TechWell.com is the place to go for what is happening in software development and delivery.  Join the conversation now!

Upcoming Events

Oct 12
Oct 15
Nov 09
Nov 09