Last few months I’ve been using Go to write quite a lot of tools. In this post I intent to show not why I chose Go over others, but how I architect those tools, what libraries I use and what kind of automation I have in place.
I’m sure some folks will preffer others, but, right now I’m using:
- golang/dep for dependency management;
- caarlos0/spin for showing spinners when it makes sense;
- urfave/cli as cli “framework”;
- stretchr/testify for better testing;
- alecthomas/gometalinter for code linting;
- goreleaser/goreleaser for release automation.
At this point some will ask me why I built my own spinner library. The answer is: I didn’t like the two or three libraries I tested, so I wrote my own in the way I wanted.
A basic tree would look like:
. ├── Makefile ├── cmd │ └── example │ └── main.go ├── example.go ├── example_test.go ├── goreleaser.yml ├── Gopkg.lock └── Gopkg.toml
Makefile: contains common tasks for the project, like formating, testing, linting, etc;
cmd/example/main.go: is the cli entrypoint;
example_test.go: is the “library” of the application and its respective files. Could be more than one file, of course;
goreleaser.yml: the GoReleaser configuration;
Gopkg.toml: dependencies locks and manifest.
Of course, creating this all the time I want to write some tool would be a lot of work, so I kind of automated it.
Starting a new project
To help me (and hopefully others) start projects faster, I created caarlos0/example, which contains the directory struture we just talked about, as well the deps, Makefile, License, readme, code of conduct and all that.
To use it, we can simply:
$ cd $GOPATH/src/github.com/myuser $ git clone firstname.lastname@example.org:caarlos0/example.git myapp $ cd myapp $ ./script/setup myuser MyApp # notice the case on the second arg
It is actually a working app (that does nothing), to run it:
$ make setup $ go run ./cmd/example/main.go -h
Now, we create a GitHub repository for our new app and push it:
$ git push origin master
If we check the README file, we’ll see that there are already a few badges on the top, but some of them are not working. Let’s fix them!
Badges are important! They show cool stats about our project and may show off other tools we use.
caarlos0/example includes badges for:
- Latest release
- Travis build status
- Codecov Coverage status
- GoReportCard status
- SayThanks.io, so people can thank us for our projects
- GoReleaser badge
We need to enable our new repository on Travis settings to make the travis badge work.
Coverage reports will require us to just have an account at Codecov website. Yep, just as simple.
Now, we need commit and push something to fire a Travis build and check if both Travis and Codecov badges work:
$ git commit --allow-empty -m 'fire travis build' $ git push origin master
Wait a few minutes, and the badges should be displayed on the README!
SayThanks.io is as easy as creating an account on the site and adjusting the username on the badge.
With that done, we can focus on write the features we want!
First of all, we can read the
It is the “newcomer guide” and usually helps having more contributions,
so, it’s good to make sure it’s always right and up to date.
tl;dr: there are some make tasks:
$ make setup # install tools and deps $ make fmt # format code $ make test # runs tests $ make lint # runs gometalinter
I usually write my features in the root folder. After that, I call
them from my
And that’s it. Feature is done.
Of course, more complex tools might require more folders and stuff like that, but most simple tools I have written fit in this category.
Once we are done coding the features, we’ll want to distribute our app, right?
For some time I had a quick-and-dirty shell script to automate this for me, but I ended up needing more power, and therefore started the GoReleaser project. It is approaching v1.0.0 very soon, and is already being used by ~50 public projects, despite being very new.
caarlos0/example is already set up to:
- build both Linux and macOS binaries for amd64;
- push a Homebrew recipe to a tap.
To enable it, we just need to create a new
repo box checked.
Then, we need to add it as an environment variable named
Travis project settings.
We also need a tap repository. Mine is caarlos0/homebrew-tap, which
is kind of a standard. I also created a
Formula folder inside it.
Finally, we need to change the
gorelaser.yml file, pointing to our
new homebrew-tap repository.
Now, commit and push everything:
$ git add goreleaser.yml $ git commit -m 'setup goreleaser' $ git push origin master
To fire a release, GoReleaser expects us to create semantic versioned
tags, for example,
v1.2.3. We can do that using either the command line
or GitHub web interface:
$ git tag v1.0.0 $ git push origin v1.0.0
Travis will build the tag, and, on the
post-success hook, will check if
its build a tag, and then, download and execute GoReleaser, which will do the
heavylifting for us.
You can check how a release will look like in the example releases.
This is how I work right now. I’m not saying it is perfect nor the only way of doing this kind of stuff. I only kind of automated some stuff and wanted to share it so other people may save some time as well.
I’m curious, though, what tools do you use? How do you structure your projects? Let’s discuss on the comments bellow!
comments powered by Disqus