I haven’t been good at blogging for a while. Most of my writing has been toward a new book (details forthcoming), but I did manage to write up how I see designing service objects for Rails applications over at Stitch Fix’s Eng Blog.

Anatomy of a Rails Service Object:

We’ve given up on “fat models, skinny controllers” as a design style for our Rails apps—in fact we abandoned it before we started. Instead, we factor our code into special-purpose classes, commonly called service objects.

The post goes over six rules of thumb that I’ve found useful, including class design, method parameter, and return values.

Post by yours truly on Sitepoint’s Ruby section, “What You Don’t Get With ActiveJob”:

ActiveJob provides simple solutions to two problems faced by Rails developers when writing background jobs: queueing jobs (via the ActiveJob API) and serialization of ActiveRecord objects (via GlobalID).

It doesn’t, however, help with the third problem: writing resilient jobs that can survive in a production environment.

Giles Bowkett posted a great piece on the deep flaws of Agile processes like Scrum. It’s spot-on with my experience. There’s so many choice quotes, it’s hard to pull just one, but I’ll try. Toward the end of the article:

I’ve never seen Scrum-like frameworks for transmuting the work of designers, marketers, or accountants into cartoonish oversimplifications like story points. People are happy to treat these workers as adults and trust them to do their jobs.

I don’t know why this same trust does not prevail in the culture of managing programmers.

After a job converting JIRA Tickets into Story Points into diffs, I eventually realized that good developers deliver results, not promises, points, or even features. I even wrote an entire book about it.

Thankfully, the team at Stitch Fix has no sprints, story points, velocity, or planning poker. We just solve the most important problem at hand, and move onto the next most important one.

@garybernhardt linked to a short academic paper on code readability, called “What Makes Code Hard to Understand?”. It’s a quick read that details an experiment where researchers showed severals versions of the same program to a bunch of programmers and asked them to guess the output. Each version had the same cyclomatic complexity and lines of code, but differed only in formatting, variable naming, and use of types.

Choice quote:

The physical aspects of notation, often considered superficial, can have a profound impact on performance [of programmers to understand what code will do]

In particular, in a program that relied on order of operations, two versions were given, one in which one space was used around operators (zigzag) and another where the operators were all vertically aligned (linedup):

Programmers were more likely to respect the order of mathematical operations in the linedup version of whitespace, showing how horizontal space can emphasize the common structure between related calculations.

Also, they presented three versions of code to calculate the area of a rectangle, using free variables, tuples, and a Rectangle class:

Programmers took longer to respond to the tuples version of rectangle despite it having fewer lines than the class version. It is not uncommon in Python to use tuples for (x, y) coordinates, but the syntactic “noise” that is present in the tuples version for variable names (e.g., r1_xy_1) and calculations (e.g., width = xy_2[0] - xy_1[0]) likely gave programmers pause when verifying the code’s operation.

This was not something that was initially obvious to me as I learned programming, but I have come to realize the importance of typography in writing code. From my post on the subject over a year ago:

But, it’s not just the content - the code itself - that affects readability. How it’s presented matters and if we’re going to talk about presentation, we have to talk about typography.

Over on Stitch Fix’s engineering blog, I wrote a new post on how we deal with tricky job failures. Namely, make your jobs idempotent:

Trapping Resque::TermException is really just kicking the can down the road. You might prevent some of your jobs from failing, but you’ll still get failed jobs. Granted, they will fail at a lower rate, but it’s still a rate correlated to the scale of your business.

The underlying problem is that jobs aren’t idempotent.

Check out the entire thing here.

Monitoring multiple Resque instances, especially from a mobile phone browser, is a huge pain. That pain is now over. From the Stitch Fix blog

Today, we’re introducing Resque Brain, an open source application that monitors multiple resque instances. It has a mobile-friendly UI, and provides better tools for managing the failed queue. It also includes some basic monitoring and statistics tasks.

Check it out.