I don't like the idea having long-running branches and especially TODO comments in code before a review. I would argue that it is better to discuss large architectural changes as a team, and split the task into smaller, more manageable tickets and branches which can be reviewed and/or QA'd individually. I think this is key to maintaining clean and reliable.

I think the approach outlined is a bit upside down, and significant API changes should be agreed on first. Not necessarily documented or thought about in detail, but the implementer should walk away from the preliminary discussion with a good idea of the agreed structure in order to be able to flesh it out further.

This means the architectural changes on a high level are understood and agreed on by the some or all of team first and the implementation is what actually needs to be reviewed afterwards.

Thinking about a solution, pushing a draft to code review with your thoughts explaining the changes typed out in full, responding to comments on the review two days later, and repeating the process all over again- it just seems like if you need this sort of feedback it would be way more efficient to schedule an informal exchange in front of a whiteboard with one or more team members.

Re: the first point. I agree that large changes should always be broken down into individually testable pieces, however I don't agree that they can always be broken down into individually shippable pieces. This is definitely true of large refactor projects, where a simple change (like upgrading a third party library) will unavoidably trigger thousands of changes throughout the codebase. I find it easier as a reviewer to look through a bunch of TODO comments that outline what needs to be done, and then each subsequent review addresses a manageable amount of the TODOs. I really see no other way since we can't ship two copies of, say, jQuery to customers.

I totally agree with you that the architectural changes should be agreed upon in advance, with direct conversations happening and whiteboards if necessary. I tried to make it clear in the post that architecture design should happen before any coding begins, but the gap I noticed was that even with full agreement on a schema or even down to the function level, there are still important details to work out before writing logic and UI on top of it, and that requires special attention from reviewers.

Just to play devil's advocate though, if you avoid face-to-face design sessions and do everything through code reviews, you will be forced to explain your thinking in text, and the whole conversation will remain as a record for future developers getting familiar with the code who may be confused about why things were done a particular way. The code review paper trail is immensely helpful to understanding a large codebase - reading through sparse meeting notes is never as good.

Very nice article. To be honest, I always review code from a architectural perspective, rather than just reviewing code quality, as getting that stuff wrong will hurt a project more so.

I have been looking for jobs recently. No one seems interested in testing my architectural decisions.

Instead I get silly coding trivial pursuit questions, or built an application to do this in 2 hours. Neither of these approaches seems to test what I am good at after 11 years of software engineering - coming up with an architecture that makes the correct trade offs (rushing a new project for a 2 hour deadline is probably the worst way for me to write code).

Completely agree. When I hire engineers, I get them to write an app at home for a week and then send it in. I don't set an arbitrary deadline, just send it when you think it's ready. I'm then looking at:

1. Code quality (mainly with a view to maintainability)

2. Unit test coverage (I don't even asked them to do that, the good ones know to provide them).

3. How the app uses dependencies (APIs, databases, caches), and if they support dependency injection.

4. Can the app scale up and down (i.e. can I scale horizontally by just adding another instance?).

5. How are they using 3rd party libs?

6. How well documented is the solution?

7. How easy is it to build and deploy?

8. How well structured is their database (PKs, FKs, indexes, no excessives joins etc.)?

9. Is their a clean separation of the main modules in the app? E.g. MVC, having a REST API for the data and de-coupled clients etc.

I could go on. At a senior level, I just assume an engineer can write nice code in whatever language, but also has a keen understanding of software architecture with a view to how their code impacts on scaling and performance.

As a candidate, if all I get asked at interviews is low-level algorithms, I move on.

While I think this is an ideal approach for selection, how does it work out in practice? How will the candidate find the time to do all that while they are (most likely) employed somewhere else, and are trying other options too?

That's quite a thorough list, which I appreciate, but on the other hand its a lot to do for an interview.

That's why I give them at least a week to do the code submission at home. It is too much to cover in just a few hours in an interview.

I have blogged about an alternative [here](1). The summary is: ask them to contribute to any open-source project of their choice. Win-win for all, IMO.

  (1) : http://lavadip.com/3948_opensource_hiring_and_lowhangingfruit.html

Its also quite a lot t expect. After a couple of half days worth, with no result at the end, I am getting kind of pissed off with them. Do you pay them for this?

No we don't pay, and you're right it is a lot to expect. I've been on the other side myself plenty of times so I know how that feels: nothing more frustrating than having your time wasted. But on the flip side, I now have to waste a lot of time screening, interviewing, and doing code reviews with candidates that did not work out. Hiring sucks from both sides, I wish you luck.