What's this?

A rails repo created to help demonstrate an issue with qlty cli where both rubocop and standardrb plugins are unable to format/fix rails related offenses despite being auto-correctable.

Details

standard is a linter/formatter built on RuboCop. Standard's CLI is invoked by calling standardrb and qlty named it's plugin to match that.

The qlty standardrb plugin works great most of the time but if you add the standard-rails extra_package, any/all rubocop cops which have a NOTE: Required Rails version: of any kind like this one or this one, will fail to be formatted by the qlty standardrb plugin. The qlty standardrb plugin "linter" will be able to check/complain about it, but the "formatter" is unable to fix it.

This occurs because the rubocop source code expects to find Gemfile.lock and .bundle/config files in a directory above where the cop is run from in order to format/fix offenses that have a specific "target_rails_version", but qlty CLI executes the formatting inside of a folder at /tmp/qlty which prevents rubocop from formatting/fixing those offenses.

How to replicate

Clone this repo

git clone git@github.com:milesstanfield/qlty-test.git
cd qlty-test

try to format this file with standardrb plugin

qlty fmt app/models/foo.rb --filter=standardrb

Notice the check command finds a standard-rails issue that the fmt command didnt fix

qlty check app/models/foo.rb --no-formatters --filter=standardrb
# app/models/foo.rb:3:3
# 3:3  medium  Use `Rails.env.local?` instead.  standardrb:Rails/EnvLocal

But if you were to use the standardrb cli directly with the same commands the qlty plugin runs here, you'll find that it does fix it

# you'll need to locate the exact path to the 1.41.1-${identifier} you have locally and
# then replace the code below with that to use the standardrb executable that qlty has
# installed under the hood
~/.qlty/cache/tools/standardrb/1.41.1-56e214bd429f/gems/standard-1.41.1/exe/standardrb app/models/foo.rb --fix

It's fixed

qlty check app/models/foo.rb --no-formatters --filter=standardrb
# ✔ No issues
git diff
# -    Rails.env.development? || Rails.env.test?
# +    Rails.env.local?

You can also perform this same test with qlty rubocop plugin and have the same results

# this wont format it
qlty fmt app/models/foo.rb --filter=rubocop

# check to verify
qlty check app/models/foo.rb --no-formatters --filter=rubocop
# app/models/foo.rb:3:3
# 3:3  medium  Use `Rails.env.local?` instead.  standardrb:Rails/EnvLocal

# this _does_ format it
~/.qlty/cache/tools/rubocop/1.70.0-8d8faa68e17b/gems/rubocop-1.70.0/exe/rubocop --autocorrect app/models/foo.rb

# and now check to verify
qlty check app/models/foo.rb --no-formatters --filter=rubocop
# ✔ No issues

Temporary Solution

Adding the following to .qlty/qlty.toml solves the standardrb issue, but it's a temporary hack until the underlying issue is resolved by qlty (somehow)

[plugins.definitions.standardrb.drivers.format]
script = """
lockfile=$(echo \"${config_file}\" | sed \"s/.standard.yml/Gemfile.lock/\");
destination=${PWD%/*};
cp -rf $lockfile $destination;
mkdir -p ${destination}/.bundle;
touch ${destination}/.bundle;
standardrb ${target} --fix
"""

And similarly, this will resolve the rubocop issue though I did have to change the script command from rubocop --fix-layout ${target} to be rubocop --autocorrect ${target}. Honestly i suspect it should be --autocorrect but that's a different issue altogether imo.

[plugins.definitions.rubocop.drivers.format]
script = """
lockfile=$(echo \"${config_file}\" | sed \"s/.rubocop.yml/Gemfile.lock/\");
destination=${PWD%/*};
cp -rf $lockfile $destination;
mkdir -p ${destination}/.bundle;
touch ${destination}/.bundle;
rubocop --autocorrect ${target}
"""