gotestdox

A tool for formatting Go test results as readable documentation

Programming language: Go

License: MIT License

Tags: Go Tools    

README

go install github.com/bitfield/gotestdox/cmd/gotestdox@latest

[Writing gopher logo](img/gotestdox.png)

gotestdox

gotestdox is a command-line tool for turning Go test names into readable sentences. For example, suppose we have some tests named like this:

TestRelevantIsTrueForTestPassOrFailEvents
TestRelevantIsFalseForNonPassFailEvents

We can transform them into straightforward sentences that express the desired behaviour, by running gotestdox:

gotestdox

This will run the tests, and print:

 ✔ Relevant is true for test pass or fail events (0.00s)
 ✔ Relevant is false for non pass fail events (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 report exit status 1.

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). It doesn't report fuzz tests, since they don't tend to have useful names.

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 returns a correctly configured server (0.00s)
 ✔ NewServer errors on invalid config options (0.00s)

github.com/octocat/mymodule/util:
 ✔ LeftPad adds the correct number of leading spaces (0.00s)

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 library

See pkg.go.dev/github.com/bitfield/gotestdox for the full documentation on using gotestdox as a library package.

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:

• Test names should be sentences

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!

Some examples

Here is the complete gotestdox rendering of its own tests (sorted for readability), in case it gives you any useful ideas:

github.com/bitfield/gotestdox:
 ✔ EventString formats pass and fail events differently (0.00s)
 ✔ ExecGoTest sets OK to false when command errors (0.02s)
 ✔ ExecGoTest sets OK to false when tests fail (0.76s)
 ✔ ExecGoTest sets OK to true when tests pass (0.63s)
 ✔ Filter keeps track of current package (0.00s)
 ✔ Filter sets OK to false if any test fails (0.01s)
 ✔ Filter sets OK to false on parsing error (0.00s)
 ✔ Filter sets OK to true if there are no test failures (0.01s)
 ✔ Filter skips irrelevant events (0.01s)
 ✔ NewTestDoxer returns testdoxer with standard IO streams (0.00s)
 ✔ ParseJSON errors on invalid JSON (0.00s)
 ✔ ParseJSON returns valid data for valid JSON (0.00s)
 ✔ Prettifier logs to debug writer (0.00s)
 ✔ Prettify (0.01s)
 ✔ Prettify accepts a single-letter test name (0.00s)
 ✔ Prettify accepts a single-word test name (0.00s)
 ✔ Prettify does not break words when a digit follows an '=' sign (0.00s)
 ✔ Prettify does not erase the final digit in words that end with a digit (0.00s)
 ✔ Prettify does not hang when name ends with initialism (0.00s)
 ✔ Prettify does not treat an underscore in a subtest name as marking the end of a multiword function name (0.00s)
 ✔ Prettify doesn't incorrectly title-case single-letter words (0.00s)
 ✔ Prettify eliminates any words containing underscores after splitting (0.00s)
 ✔ Prettify handles a test with no name, but with subtests (0.00s)
 ✔ Prettify handles multiple underscores, with the first marking the end of a multiword function name (0.00s)
 ✔ Prettify inserts a word break before subtest names beginning with a lowercase letter (0.00s)
 ✔ Prettify is okay with test names not in the form of a sentence (0.00s)
 ✔ Prettify keeps a trailing digit as part of an initialism (0.00s)
 ✔ Prettify keeps numbers within a hyphenated word (0.00s)
 ✔ Prettify keeps together digits in numbers that are standalone words (0.00s)
 ✔ Prettify keeps together hyphenated words with initial capitals (0.00s)
 ✔ Prettify knows that just 'test' is a valid test name (0.00s)
 ✔ Prettify preserves capitalisation of initialism when it is the first word (0.00s)
 ✔ Prettify preserves capitalisation of initialisms such as 'PDF' (0.00s)
 ✔ Prettify preserves capitalisation of two-letter initialisms such as 'OK' (0.00s)
 ✔ Prettify preserves initialisms containing digits (0.00s)
 ✔ Prettify preserves initialisms containing digits with two or more leading alpha characters (0.00s)
 ✔ Prettify preserves longer all-caps words (0.00s)
 ✔ Prettify recognises a dash followed by a digit as a negative number (0.00s)
 ✔ Prettify renders subtest names without the slash, and with underscores replaced by spaces (0.00s)
 ✔ Prettify replaces camel-case transitions with spaces (0.00s)
 ✔ Prettify retains apostrophised words in their original form (0.00s)
 ✔ Prettify retains capitalisation of initialisms in a multiword function name (0.00s)
 ✔ Prettify retains hyphenated words in their original form (0.00s)
 ✔ Prettify retains quoted words as quoted (0.00s)
 ✔ Prettify treats a single underscore as marking the end of a multiword function name (0.00s)
 ✔ Prettify treats a single underscore before the first slash as marking the end of a multiword function name (0.00s)
 ✔ Prettify treats consecutive underscores as a single word break (0.00s)
 ✔ Prettify treats numbers as word separators (0.00s)
 ✔ Prettify treats underscores as word breaks (0.00s)
 ✔ Relevant is false for non test pass fail events (0.00s)
 ✔ Relevant is true for test pass or fail events (0.00s)

Links

• Bitfield Consulting
• Test names should be sentences
• The Power of Go: Tests

Gopher image by MariaLetta