There are no “Usage” comments at the top of the file, but there are some “Summary” comments. So we’ll start there.

“Summary” comments

# Summary: Show the current Ruby version and its origin
#
# Shows the currently selected Ruby version and how it was
# selected. To obtain only the version string, use `rbenv
# version-name'.

According to the comments, the expected output of the command is a version number plus an “origin”.

I try the command in my terminal, and get the following:

bash-3.2$ rbenv version

2.7.5 (set by /Users/myusername/.rbenv/version)

2.7.5 is my Ruby version number, and /Users/myusername/.rbenv/version is the origin.

Now let’s move on to the test file.

Tests

“Installing” a test Ruby version

Our first block of code is:

create_version() {
  mkdir -p "${RBENV_ROOT}/versions/$1"
}

This is a helper function to “install” a fake Ruby version inside RBENV’s versions/ directory. If RBENV sees a sub-directory inside versions/, it will assume it represents a specific version of Ruby that the user has installed.

The above function simply creates a sub-directory inside the RBENV_ROOT/versions directory, whose name corresponds to the function’s first argument. The -p flag makes any parent directories which don’t already exist.

The setup callback

Next code block:

setup() {
  mkdir -p "$RBENV_TEST_DIR"
  cd "$RBENV_TEST_DIR"
}

Another helper function, this time to create and navigte into the expected test directory.

When no RBENV Rubies are installed

Next block of code (and first test):

@test "no version selected" {
  assert [ ! -d "${RBENV_ROOT}/versions" ]
  run rbenv-version
  assert_success "system"
}

As a sanity check, we first assert that the $RBENV_ROOT/versions directory doesn’t exist. Then we run the command, and assert that the command falls back to printing system, signifying that the Ruby installation being used is the default one from the user’s machine.

When the origin is the RBENV_VERSION environment variable

Next test:

@test "set by RBENV_VERSION" {
  create_version "1.9.3"
  RBENV_VERSION=1.9.3 run rbenv-version
  assert_success "1.9.3 (set by RBENV_VERSION environment variable)"
}

This test creates a Ruby version in RBENV’s directory. It then runs the command with the RBENV_VERSION env var set to this version number, and asserts that:

  • the correct version number is output, and
  • RBENV_VERSION is the source of the version info.

When the origin is a .ruby-version file

Next test:

@test "set by local file" {
  create_version "1.9.3"
  cat > ".ruby-version" <<<"1.9.3"
  run rbenv-version
  assert_success "1.9.3 (set by ${PWD}/.ruby-version)"
}

Here we not only create a Ruby version directory, but we also add a .ruby-version file and set its contents to be the stringified Ruby version whose folder we just created. We then run the command, and assert that:

  • it succeeded, and that
  • the printed output includes the correct Ruby version and the correct source (the .ruby-version file we just created).

The line:

cat > ".ruby-version" <<<"1.9.3"

…says: “Read from STDIN and print what you receive to the file .ruby-version. At the same time, send the string “1.9.3” to STDIN.” The line <<<"1.9.3" is called a herestring, and is a way of sending text to stdin.

When the origin is RBENV’s global Ruby file

Final test:

@test "set by global file" {
  create_version "1.9.3"
  cat > "${RBENV_ROOT}/version" <<<"1.9.3"
  run rbenv-version
  assert_success "1.9.3 (set by ${RBENV_ROOT}/version)"
}

This test creates a Ruby version directory and a separate file named version inside the root directory. We then run the command, and assert that it was successful. We also assert that the printed output contains the correct version number and the correct source (the global version file).

That’s it for the test file.

Adding a new test

One thing I noticed is that the last two tests might not be covering all of our bases here. All they do is assert that we can pull the Ruby version from a local project file named .ruby-version, or that we can pull it from the global version file. If both those things exist, we don’t currently have test coverage documenting what should happen (i.e. that the local file should be used over the global file).

I create this test and run it to see if it passes:

@test "prefer local over global file" {
  create_version "1.9.3"
  create_version "3.0.0"
  cat > ".ruby-version" <<<"1.9.3"
  cat > "${RBENV_ROOT}/version" <<<"3.0.0"
  run rbenv-version
  assert_success "1.9.3 (set by ${PWD}/.ruby-version)"
}

It does indeed pass, so I create a PR to add this test, and a few days later the core team merges it.

Code

Now on to the code itself.

First block of code:

set -e
[ -n "$RBENV_DEBUG" ] && set -x

You know the drill:

set -e to exit after the first error Setting verbose mode when RBENV_DEBUG is passed.

Setting the version name and origin

Next block of code:

version_name="$(rbenv-version-name)"
version_origin="$(rbenv-version-origin)"

Here we run the rbenv version-name and rbenv version-origin commands, and store their output in properly-named variables. We’ll read through the code for these commands later. For now, we can just run them in the terminal to see what their output is:

bash-3.2$ rbenv version-name

2.7.5

bash-3.2$ rbenv version-origin

/Users/myusername/.rbenv/version

Pretty unsurprising IMHO, given what we already know.

Printing the version name (and possibly the origin)

Next / last block of code:

if [ "$version_origin" = "${RBENV_ROOT}/version" ] && [ ! -e "$version_origin" ]; then
  echo "$version_name"
else
  echo "$version_name (set by $version_origin)"
fi

Here we check whether:

  • the value of version_origin that was returned by the rbenv version-origin command is equal to the global version file, and
  • whether that file does not yet exist.

If both these things are true, we simply print the version number. If either is false, we print the version number plus its origin.

It’s a bit strange to me that we would check both that the origin is equal to ${RBENV_ROOT}/version and that the version_origin file does not yet exist. If the return value of the rbenv-version-origin command (which we stored in the version_origin variable) was ${RBENV_ROOT}/version, shouldn’t we be confident that the ${RBENV_ROOT}/version file exists?

I guess we’ll find out when we get to the rbenv-version-origin command.

At any rate, that’s the end of this file. On to the next one.