goldplate 🏅
$ goldplate -j2 tests/
1..26
ok tests/prettify-json.goldplate: exit_code
ok tests/prettify-json.goldplate: stdout
ok tests/env.goldplate: exit_code
ok tests/env.goldplate: stdout
...
goldplate is a cute and simple opaque golden test runner for CLI
applications. You place your test cases in a directory, annotate them with
.goldplate files, and that's it. It is completely language agnostic. And
perhaps its best feature is that it can automatically✨🪄 fix your tests outputs!
Give goldplate a try if:
- You are testing a CLI application
- The application produces deterministic output (e.g. compilers, prettifiers,
convertors)
- The application is quick to start (a process is created for every test)
- Your output is text-based and not huge in size
At Fugue, we've been using internal versions of this tool
since 2016, so it should be pretty stable.
goldplate produces output compatible with the Test Anything Protocol.
Table of Contents
Tutorial
Creating a first test
Imagine we are up to testing the behaviour of echo command. In this very
simple example, we run echo "Hello, world!" and expect it to print Hello, world! to the stdout stream as a result.
Create a new file echo.goldplate and add the following content:
{
"command": "echo",
"arguments": ["Hello, world!"],
"asserts": [
{"exit_code": 0},
{"stdout": "hello-world.txt"}
]
}
Let's go through this bit by bit.
The test invocation is specified by the command and arguments fields. We
are invoking the echo process with a single argument, "Hello, world!".
The expected results of our test live in the asserts field. This simple test
has two asserts:
- We verify that the exit code is 0 (success).
- We check the
stdout (output) of the command against the contents of
the file hello-world.txt.
We haven't created hello-world.txt yet, but that's not a problem. We can
invoke goldplate --fix to create it:
$ goldplate echo.goldplate --pretty-diff --fix
1..2
ok echo.goldplate: exit_code
not ok echo.goldplate: stdout
diff:
0a1
> Hello, world!
fixed ./hello-world.txt
After hello-world.txt has been created with proper contents, subsequent
goldplate invocations will pass:
$ goldplate echo.goldplate
1..2
ok echo.goldplate: exit_code
ok echo.goldplate: stdout
You can view the full example here:
View example:
You can pass one or multiple lines of input to the command by using the stdin
field.
Setting environment variables
View example:
The environment field can be used to set environment variables for the
program.
goldplate also sets a number of environment
variables. You can use these directly within the
configuration JSON. In this example, we use:
{"stdout": "${GOLDPLATE_NAME}.stdout"}
Rather than:
{"stdout": "env.stdout"}
We found this to be good practice, it makes mass-renaming of tests much easier.
Setting work directory
View example:
The working_directory field can be used to set the work directory in which the
command will be executed. It can either be an absolute path or a path relative
to the goldplate file. If a working_directory is specified then the other
fields like input_files and stdout need to be relative to the
working_directory as well.
If a work directory is not specified the command will be executed in the
same directory as the goldplate file.
View example:
.goldplate files are fairly small but if you have a whole directory of files
that you just want to run the same command on, they can get very repetitive.
This is why goldplate provides a simple way to pull in multiple input files.
If the input_files field is set to a glob, all asserts will be ran for every
matching input file. goldplate will set the following variables:
${GOLDPLATE_INPUT_FILE}: the path to the input file
${GOLDPLATE_INPUT_NAME}: the input file without extension
Post processing: find and replace
View example:
Sometimes you may want to do a find-and-replace on the actual output, for
example to filter out timestamps or other information that you do not expect to
match up against the expected output.
Post processing: prettify JSON
View example:
Many modern CLI tools output JSON. You can use the prettify_json post
processor to make sure the JSON is converted to a normalized form with sorted
keys and consistent indentation.
Created files and directories
View example:
goldplate is geared towards checking the stdout and stderr outputs of a
program, but you can also check that files were created with specific contents.
If you do this, goldplate will remove these files and directories afterwards
to leave a clean repository behind.
Installation
Binaries for Mac OS and Linux are available on the
releases page.
Installation from source is done using standard Haskell tooling -- Cabal
and stack both work well.
Using stack
- Install stack for your platform.
- Clone this repository and
cd into it.
- Run
stack install.
- Make sure
$HOME/.local/bin is in your $PATH.
Using Cabal
- Install Cabal for your platform.
- Clone this repository and
cd into it.
- Run
cabal install.
- Make sure
$HOME/.cabal/bin is in your $PATH.
Reference
Syntax
Environment variables can be spliced into the configuration using ${VAR}
syntax within strings. To escape this syntax, use $${VAR} to get a literal
${VAR}, $$${VAR} to get a literal $${VAR}, and so on.
Environment variables
The test is always executed in the directory that holds the .goldplate file.
goldplate will always set the following environment variables:
GOLDPLATE_FILE: The filename of the .goldplate file.
GOLDPLATE_NAME: The filename of the .goldplate file without the
extension.
GOLDPLATE_BASENAME: The basename (filename without directory) of
the .goldplate file.
When dealing with multiple input files, the following
additional variables are set:
GOLDPLATE_INPUT_FILE: The input file name (relative to the current
directory).
GOLDPLATE_INPUT_NAME: The same as GOLDPLATE_INPUT_FILE but without
the extension.
GOLDPLATE_INPUT_BASENAME: The basename (filename without directory) of
the input file.
Here is an example that outputs all of these environment variables:
Similar projects
A similar project is smoke. I think goldplate has two major advantages
over smoke:
- It can fix the output files automatically using
--fix! This is very
useful if you make a change to your tool that affects a lot of test
files. You still need to manually review the diff, but at least you don't
need to manually update the specs.
- You can avoid most repetitive configs by using
simple globbing.