π
goldplate
$ goldplate -j2 tests/
Found 32 specs
Running 49 executions in 2 jobs
Finished in 0.84s
Ran 32 specs, 49 executions, 146 asserts, all A-OK!
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
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.
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 filehello-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
...
echo.goldplate: stdout: does not match
echo.goldplate: fixed ./hello-world.txt
...
Ran 1 specs, 1 executions, 2 asserts, 1 failed.
After hello-world.txt
has been created with proper contents, subsequent goldplate
invocations will pass:
$ goldplate echo.goldplate
...
Ran 1 specs, 1 executions, 2 asserts, all A-OK!
You can view the full example here:
Feeding input on stdin
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.
Globbing input files
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
Installation through 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.
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 asGOLDPLATE_INPUT_FILE
but without the extension.
To recap, if we have a specification prettify-js.goldplate
that uses:
"input_files": "module-*.js"
And module-1.js
matches the glob, we'll get an execution with the following environment variables set:
Variable | Value |
---|---|
GOLDPLATE_FILE |
prettify-js.goldplate |
GOLDPLATE_NAME |
prettify-js |
GOLDPLATE_INPUT_FILE |
module-1.js |
GOLDPLATE_INPUT_NAME |
module-1 |
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.