Hot on the heels of Functional February came Mechanical March in Exercism’s #12in23 challenge. There was a veritable feast of languages from which to choose and I opted for Go (apparently it’s not Golang despite all of the easier-to-find on the web references to the contrary. Go is one of the dominant languages in my domain at work, where tools like Docker and Kubernetes are in plentiful supply. I last played with Go in December 2012 when I was considering working for one of Go’s biggest users and remember quite liking it after years of C, C++ and Python. I remember really liking the approach of returning errors as a second return value, so that it is handled by the caller. The language was only about three years old then, so I was keen to see how it had matured.

The only valid measurement of code quality: WTFs/minute, or so it’s said. I had so many WTF moments playing with Go this month that I am just chosing to share a few of them with you.

Before I get into Go, I’m going to call out how excellent asdf is when working wth multiple programming languages. Install the relevant language plugin from asdf-plugins (Go’s is called golang just because), asdf install golang latest && asdf local golang latest and off you go.

Oh and how good is Dash for docs? I’ve been using it with Alfred for years but it also integrates really nicely with VSCode, which is my editor of choice these days.

Comments

I am considerably less anti-comment (and put much less stock into what the referenced author has to say) than when I wrote Comments Are Lies! in 2012 but /* increment i by one */ style comments still make my teeth itch. Exercism’s Weather Forecast exercise was painful to work through. Documenting this module following the Go Doc Comments specification felt like an exercise in futility. Comments must start with specific words and end with a “period”. Does this really make code or generated content easier to understand?

Parsing and formatting dates and times

Most languages I have worked with use strftime(3)-style formatting specifications for handling date and time strings. The other timestamp etched in my programmer’s brain is 00:00:00 UTC on Thursday, 1 January 1970, the beginning of the Unix epoch. These are well-understood and recognised concepts.

Imagine my surprise when the format specifiers I had been using for over a quarter of a century didn’t work when undertaking the Booking up for Beauty exercise. A bit of digging around informed me…

// The reference time used in these layouts is the specific time stamp:
//
//  01/02 03:04:05PM '06 -0700
//
// (January 2, 15:04:05, 2006, in time zone seven hours west of GMT).

https://github.com/golang/go/blob/5e191f8/src/time/format.go

The what now!?

I’m sure there’d be some who’d argue that this is as simple as 1, 2, 3, 4, 5, 6, 7 (which just happens to be a Monday) but developers working with the Gregorian calendar in other languages have been seasoned to use the International Standard (i.e. 2006-01-02T15:04:05Z-07:00) and 6, 1, 2, 15, 4, 5, 6, -7 is arguably not as simple nor as useful as a mnemonic. This made the Booking up for Beauty exercise more confusing in Go than I would have expected. The Description() implementation may be harder to understand to folks coming from other languages because the placeholders are indistinguishable from regular text.

I found JavaScript’s zero-based months easier to reason with.

How to test whether a key is present within a map

This is a doozy. Given a map…

units := map[string]int{
  "quarter_of_a_dozen": 3,
  "half_of_a_dozen":    6,
  "dozen":              12,
  "small_gross":        120,
  "gross":              144,
  "great_gross":        1728,
 }

…how do you determine whether units contains the string key “dozen”? Accessing a missing element using an index expression (units["score"]) returns a zero value for the stored type.

First I checked out Map types in the The Go Programming Language Specification. Nothing there. Next I went to Index expressions and failed to spot the pertinent text (hence “doc: defragment index expressions spec for map types”). From there I ended up on a blog post, Go maps in action, which has the answer:

A two-value assignment tests for the existence of a key:

i, ok := m["route"]

I mentioned earlier how I liked the Go approach of returning an error as the second return value. It transpires that there’s another opposing “comma ok” idiom for maps. Because of course there is. To make matters more confusing i := units["score"] will set i to 0 in the example above but i, ok := units["score"] is also valid with a second return value. After a month with the language I’m still unsure how you specify exactly how many return values to expect in these cases, since the following doesn’t compile:

func GetUnit(units map[string]int, item string) (int, bool) {
 return units[item] // not enough return values
}

…but this does…

func GetUnit(units map[string]int, item string) (int, bool) {
 i, ok := units[item]
 return i, ok
}

OK Go!

Named Captures in Regular expressions

One of the many things I like about Ruby is how easy it is to use regular expressions to get things done. I guess it gets this from Perl and AWK.

For example to get a couple of words:

/^(?<first>\w+)\s+(?<second>\w+)$/.match('Hello World') # => #<MatchData "Hello World" first:"Hello" second:"World">

In Go, I have three problems (as my long-suffering friend Marcus points out). After specifying the regular expression, I need to create my own map to get the indexes for the named matches.

package main

import (
 "fmt"
 "regexp"
)

func main() {
 matcher := regexp.MustCompile(`^(?P<first>\w+)\s+(?P<second>\w+)`)
 matches := matcher.FindStringSubmatch("Hello World")
 namedMatches := map[string]string{}
 for i, name := range matcher.SubexpNames() {
  namedMatches[name] = matches[i]
 }
 fmt.Println(namedMatches)
}
go run main.go
# map[:Hello World first:Hello second:World]

So much boilerplate code to get stuff done. I had similarly verbose experiences with simple programming concepts like returning a positive integer (there’s no abs() in the standard library and finding an element in an array a slice – roll your own for-loop). And to get the last item from a slice, you need first to determine the length of said slice. I really expected Go to be considerably better than its C-like predecessors but it’s not and I’m not even going to dig into pointers in this post. It’s already too ranty!

While I am glad I have enough familiarity with its syntax to be able to review Go code at work, it is not a language I’ll be rushing to for solving hobby programming problems.

Okay, April, what do you have in store for me?