Awesome
gotestdox
is a command-line tool for formatting Go test results as readable documentation, as recommended in my book The Power of Go: Tests.
Here's how to install it:
go install github.com/bitfield/gotestdox/cmd/gotestdox@latest
In any Go project, run:
gotestdox ./...
What does it do?
gotestdox
runs your tests and reports the results, but it formats their names in a special way. It converts test names WrittenInCamelCase into ordinary sentences.
For example, suppose we have some tests named like this:
TestValidIsTrueForValidInputs
TestValidIsFalseForInvalidInputs
We can transform them into readably-spaced sentences that express the desired behaviour, by running gotestdox
:
gotestdox
This will run the tests, and print:
✔ Valid is true for valid inputs (0.00s)
✔ Valid is false for invalid inputs (0.00s)
Why?
I read a blog post by Dan North, which says:
My first “Aha!” moment occurred as I was being shown a deceptively simple utility called
agiledox
, written by my colleague, Chris Stevenson. It takes a JUnit test class and prints out the method names as plain sentences.The word “test” is stripped from both the class name and the method names, and the camel-case method name is converted into regular text. That’s all it does, but its effect is amazing.
Developers discovered it could do at least some of their documentation for them, so they started to write test methods that were real sentences.
—Dan North, Introducing BDD
How?
The original testdox
tool (part of agiledox
) was very simple, as Dan describes: it just turned a camel-case JUnit test name like testFailsForDuplicateCustomers
into a space-separated sentence like fails for duplicate customers
.
And that's what I find neat about it: it's so simple that it hardly seems like it could be of any value, but it is. I've already used the idea to improve a lot of my test names.
There are implementations of testdox
for various languages other than Java: for example, PHP, Python, and .NET. I haven't found one for Go, so here it is.
gotestdox
reads the JSON output generated by the go test -json
command. This is easier than trying to parse Go source code, for example, and also gives us pass/fail information for the tests. It ignores all events except pass/fail events for individual tests (including subtests).
Getting fancy
Some more advanced ways to use gotestdox
:
Exit status
If there are any test failures, gotestdox
will print the output messages from the offending test and report status 1 on exit.
Colour
gotestdox
indicates a passing test with a ✔
(check mark emoji), and a failing test with an x
. These are displayed as green and red respectively, using the color
library, which automagically detects if it's talking to a colour-capable terminal.
If not (for example, when you redirect output to a file), or if the NO_COLOR
environment variable is set to any value, colour output will be disabled.
Test flags and arguments
gotestdox
, with no arguments, will run the command go test -json
and process its output.
Any arguments you supply will be passed on to go test
. For example:
gotestdox -run ParseJSON
will run the command:
go test -json -run ParseJSON
You can supply a list of packages to test, or any other arguments or flags understood by go test
. However, gotestdox
only prints events about tests (ignoring benchmarks and examples).
Since fuzz test cases are autogenerated and don't tend to have useful names, these are not included in gotestdox
output unless they are failing.
Multiple packages
To test all the packages in the current tree, run:
gotestdox ./...
Each package's test results will be prefixed by the fully-qualified name of the package. For example:
github.com/octocat/mymodule/api:
✔ NewServer errors on invalid config options (0.00s)
✔ NewServer returns a correctly configured server (0.00s)
github.com/octocat/mymodule/util:
x LeftPad adds the correct number of leading spaces (0.00s)
util_test.go:133: want " dummy", got " dummy"
Multi-word function names
There's an ambiguity about test names involving functions whose names contain more than one word. For example, suppose we're testing a function HandleInput
, and we write a test like this:
TestHandleInputClosesInputAfterReading
Unless we do something, this will be rendered as:
✔ Handle input closes input after reading
To let us give gotestdox
a hint about this, there's one extra transformation rule: the first underscore marks the end of the function name. So we can name our test like this:
TestHandleInput_ClosesInputAfterReading
and this becomes:
✔ HandleInput closes input after reading
I think this is an acceptable compromise: the gotestdox
output is much more readable, while the extra underscore in the test name doesn't seriously interfere with its readability.
The intent is not to perfectly render all sensible test names as sentences, in any case, but to do something useful with them, primarily to encourage developers to write test names that are informative descriptions of the unit's behaviour, and thus (as a side effect) read well when formatted by gotestdox
.
In other words, gotestdox
is not the thing. It's the thing that gets us to the thing, the end goal being meaningful test names (I like the term literate test names).
Filtering standard input
If you want to run go test -json
yourself, for example as part of a shell pipeline, and pipe its output into gotestdox
, you can do that too:
go test -json | gotestdox
In this case, any flags or arguments to gotestdox
will be ignored, and it won't run the tests; instead, it will act purely as a text filter. However, just as when it runs the tests itself, it will report exit status 1 if there are any test failures.
As a package
See pkg.go.dev/github.com/bitfield/gotestdox for the full documentation on using gotestdox
as a package in your own programs.
So what?
Why should you care, then? What's interesting about gotestdox
, or any testdox
-like tool, I find, is the way its output makes you think about your tests, how you name them, and what they do.
As Dan says in his blog post, turning test names into sentences is a very simple idea, but it has a powerful effect. Test names should be sentences.
Test names should be sentences
I don't know about you, but I've wasted a lot of time and energy over the years trying to choose good names for tests. I didn't really have a way to evaluate whether the name I chose was good or not. Now I do!
In fact, I wrote a whole blog post about it:
It might be interesting to show your gotestdox
output to users, customers, or business folks, and see if it makes sense to them. If so, you're on the right lines. And it's quite likely to generate some interesting conversations (“Is that really what it does? But that's not what we asked for!”)
It seems that I'm not the only one who finds this idea useful. I hear that gotestdox
is already being used in some fairly major Go projects and companies, helping their developers to get more value out of their existing tests, and encouraging them to think in interesting new ways about what tests are really for. How nice!
Links
<small>Gopher image by MariaLetta</small>