2016-02-16 22:09:00 -05:00
# Gotchas
The purpose of this guide is to document potential "gotchas" that contributors
might encounter or should avoid during development of GitLab CE and EE.
2017-02-16 08:42:40 -05:00
## Do not assert against the absolute value of a sequence-generated attribute
2016-11-10 09:24:23 -05:00
Consider the following factory:
```ruby
2017-12-13 19:13:44 -05:00
FactoryBot.define do
2016-11-10 09:24:23 -05:00
factory :label do
sequence(:title) { |n| "label#{n}" }
end
end
```
Consider the following API spec:
```ruby
2019-08-29 03:56:52 -04:00
require 'spec_helper'
2016-11-10 09:24:23 -05:00
describe API::Labels do
it 'creates a first label' do
create(:label)
get api("/projects/#{project.id}/labels", user)
expect(response).to have_http_status(200)
expect(json_response.first['name']).to eq('label1')
end
it 'creates a second label' do
create(:label)
get api("/projects/#{project.id}/labels", user)
expect(response).to have_http_status(200)
expect(json_response.first['name']).to eq('label1')
end
end
```
When run, this spec doesn't do what we might expect:
```sh
1) API::API reproduce sequence issue creates a second label
Failure/Error: expect(json_response.first['name']).to eq('label1')
expected: "label1"
got: "label2"
(compared using ==)
```
2019-07-19 16:27:05 -04:00
This is because FactoryBot sequences are not reset for each example.
2016-11-10 09:24:23 -05:00
Please remember that sequence-generated values exist only to avoid having to
explicitly set attributes that have a uniqueness constraint when using a factory.
### Solution
If you assert against a sequence-generated attribute's value, you should set it
explicitly. Also, the value you set shouldn't match the sequence pattern.
For instance, using our `:label` factory, writing `create(:label, title: 'foo')`
is ok, but `create(:label, title: 'label1')` is not.
Following is the fixed API spec:
```ruby
2019-08-29 03:56:52 -04:00
require 'spec_helper'
2016-11-10 09:24:23 -05:00
describe API::Labels do
it 'creates a first label' do
create(:label, title: 'foo')
get api("/projects/#{project.id}/labels", user)
expect(response).to have_http_status(200)
expect(json_response.first['name']).to eq('foo')
end
it 'creates a second label' do
create(:label, title: 'bar')
get api("/projects/#{project.id}/labels", user)
expect(response).to have_http_status(200)
expect(json_response.first['name']).to eq('bar')
end
end
```
2018-12-03 08:12:53 -05:00
## Avoid using `expect_any_instance_of` or `allow_any_instance_of` in RSpec
2018-06-20 09:29:15 -04:00
### Why
2018-11-13 01:07:16 -05:00
- Because it is not isolated therefore it might be broken at times.
- Because it doesn't work whenever the method we want to stub was defined
2018-06-20 09:29:15 -04:00
in a prepended module, which is very likely the case in EE. We could see
error like this:
2019-07-09 03:16:17 -04:00
```
1.1) Failure/Error: expect_any_instance_of(ApplicationSetting).to receive_messages(messages)
Using `any_instance` to stub a method (elasticsearch_indexing) that has been defined on a prepended module (EE::ApplicationSetting) is not supported.
```
2018-06-20 09:29:15 -04:00
2019-10-11 17:05:59 -04:00
### Alternative: `expect_next_instance_of` or `allow_next_instance_of`
2018-06-20 09:29:15 -04:00
Instead of writing:
```ruby
# Don't do this:
2018-12-03 08:12:53 -05:00
expect_any_instance_of(Project).to receive(:add_import_job)
2019-10-11 17:05:59 -04:00
2019-10-15 14:06:01 -04:00
# Don't do this:
2019-10-11 17:05:59 -04:00
allow_any_instance_of(Project).to receive(:add_import_job)
2018-06-20 09:29:15 -04:00
```
We could write:
```ruby
# Do this:
expect_next_instance_of(Project) do |project|
expect(project).to receive(:add_import_job)
end
2019-10-11 17:05:59 -04:00
2019-10-15 14:06:01 -04:00
# Do this:
2019-10-11 17:05:59 -04:00
allow_next_instance_of(Project) do |project|
allow(project).to receive(:add_import_job)
end
2018-06-20 09:29:15 -04:00
```
2019-10-15 14:06:01 -04:00
If we also want to initialize the instance with some particular arguments, we
2019-10-11 17:05:59 -04:00
could also pass it like:
2018-06-20 09:29:15 -04:00
```ruby
# Do this:
expect_next_instance_of(MergeRequests::RefreshService, project, user) do |refresh_service|
expect(refresh_service).to receive(:execute).with(oldrev, newrev, ref)
end
```
This would expect the following:
```ruby
# Above expects:
refresh_service = MergeRequests::RefreshService.new(project, user)
refresh_service.execute(oldrev, newrev, ref)
```
2017-02-16 08:42:40 -05:00
## Do not `rescue Exception`
2016-02-16 22:09:00 -05:00
2019-09-27 08:06:07 -04:00
See ["Why is it bad style to `rescue Exception => e` in Ruby?" ](https://stackoverflow.com/questions/10048173/why-is-it-bad-style-to-rescue-exception-e-in-ruby ).
2016-02-16 22:09:00 -05:00
_**Note:** This rule is [enforced automatically by
2019-11-21 10:06:17 -05:00
Rubocop](https://gitlab.com/gitlab-org/gitlab-foss/blob/8-4-stable/.rubocop.yml#L911-914)._
2016-02-16 22:09:00 -05:00
2017-02-16 08:42:40 -05:00
## Do not use inline JavaScript in views
2016-02-16 22:09:00 -05:00
2016-10-27 13:09:58 -04:00
Using the inline `:javascript` Haml filters comes with a
2016-08-04 06:30:27 -04:00
performance overhead. Using inline JavaScript is not a good way to structure your code and should be avoided.
2016-02-16 22:09:00 -05:00
2019-09-18 14:06:14 -04:00
_**Note:** We've [removed these two filters ](https://gitlab.com/gitlab-org/gitlab/blob/master/config/initializers/hamlit.rb )
2016-02-17 15:35:28 -05:00
in an initializer._
2016-02-16 22:09:00 -05:00
### Further reading
2019-09-27 08:06:07 -04:00
- Stack Overflow: [Why you should not write inline JavaScript ](https://softwareengineering.stackexchange.com/questions/86589/why-should-i-avoid-inline-scripting )
2019-11-15 19:06:14 -05:00
## Auto loading
Rails auto-loading on `development` differs from the load policy in the `production` environment.
In development mode, `config.eager_load` is set to `false` , which means classes
are loaded as needed. With the classic Rails autoloader, it is known that this can lead to
[Rails resolving the wrong class ](https://guides.rubyonrails.org/v5.2/autoloading_and_reloading_constants.html#when-constants-aren-t-missed-relative-references )
if the class name is ambiguous. This can be fixed by specifying the complete namespace to the class.
### Error prone example
```ruby
# app/controllers/application_controller.rb
class ApplicationController < ActionController::Base
...
end
# app/controllers/projects/application_controller.rb
class Projects::ApplicationController < ApplicationController
...
private
def project
...
end
end
# app/controllers/projects/submodule/some_controller.rb
module Projects
module Submodule
class SomeController < ApplicationController
def index
@some_id = project.id
end
end
end
end
```
In this case, if for any reason the top level `ApplicationController`
is loaded but `Projects::ApplicationController` is not, `ApplicationController`
would be resolved to `::ApplicationController` and then the `project` method will
be undefined and we will get an error.
#### Solution
```ruby
# app/controllers/projects/submodule/some_controller.rb
module Projects
module Submodule
class SomeController < Projects::ApplicationController
def index
@some_id = project.id
end
end
end
end
```
By specifying `Projects::` , we tell Rails exactly what class we are referring
to and we would avoid the issue.
NOTE: **Note:**
This problem will disappear as soon as we upgrade to Rails 6 and use the Zeitwerk autoloader.
### Further reading
- Rails Guides: [Autoloading and Reloading Constants (Classic Mode) ](https://guides.rubyonrails.org/autoloading_and_reloading_constants_classic_mode.html )
- Ruby Constant lookup: [Everything you ever wanted to know about constant lookup in Ruby ](http://cirw.in/blog/constant-lookup )
- Rails 6 and Zeitwerk autoloader: [Understanding Zeitwerk in Rails 6 ](https://medium.com/cedarcode/understanding-zeitwerk-in-rails-6-f168a9f09a1f )