The following are the coding style and standards for the code written and maintained by the ManageIQ team.
- High Level Guidelines
- Ruby Style Guide
- Documentation
- Logging
- Commits
- Pull Requests and Branches
- Gems
- Error and Issue Reporting
- Be consistent.
- These guides describes general guidelines to follow for new code. For existing code, stay consistent with the conventions of the code you are changing.
- Prefer readability over performance and conciseness when the performance difference is minimal.
- If you can convince others why we should violate a guideline, then violate the guideline.
- These guides are living documents and are subject to change. Feel free to comment or submit pull requests for changes, additions, or removals.
- We follow, with some exceptions, the original Ruby style guide. Any changes we have that deviate from the default style guide are enumerated in the .rubocop_base.yml file, which is inherited by most projects in the ManageIQ organization.
We use yardoc to create inline code documentation. For now documentation is scarce but we encourage contributors to add it whenever possible. Until more of the codebase is documented, we limit the created documentation to those files that are well documented. Please try to document methods that are used by external teams, like providers. These are the integration API for them and should be documented best.
To view the documentation online visit rubydoc.info. For a local
version you can run bundle exec yard
and view the html docs in /doc/index.html
. When writing documentation you
should add your source files to /.yardopts
and run bundle exec yard server -r
. This will create a local
documentation server which regenerates the docs on each request.
-
Log messages should be prefixed consistently.
# in a class method (notice the . notation) $log.info("MIQ(#{self.name}.#{__method__}) The rest of the log message.") # in an instance method (notice the # notation) $log.info("MIQ(#{self.class.name}##{__method__}) The rest of the log message.")
-
If the same log prefix will be used many times within the same method, consider using a variable named log_prefix.
log_prefix = "MIQ(#{self.class.name}##{__method__})" ... $log.info("#{log_prefix} The rest of the log message.")
-
When wrapping a block of code in logging, consider using the same string with ellipses, to make it easier to search them in the logs. As a corollary, avoid using words that would not make sense on both ends, such as "Starting". Attach any extra information after the latter message using
" - #{extra_info}"
$log.info("#{log_prefix} Refreshing EMS #{name}...") ... $log.info("#{log_prefix} Refreshing EMS #{name}...Complete") # with extra information, such as timings or counts $log.info("#{log_prefix} Refreshing EMS #{name}...") ... $log.info("#{log_prefix} Refreshing EMS #{name}...Complete - Timings: #{timings.inspect}")
- Write a good commit message. The format for a commit message is as follows.
Also read more on writing good commit messages.
- A short summary of the commit under 72 characters. Do not use a ticket number as the subject alone.
- A blank line.
- An optional body of text, preferably wrapped at 72 characters. The line length is flexible, particularly in cases where data such as tables or URLs are being copied. The body's purpose is to convey more detailed information about the commit, especially to someone who may need to search the code history for changes. Feel free to include URLs, Git SHA references to other commits, and even raw data to make the purpose of the commit clearer (e.g. "Was broken by commit 0f3a459b").
- A blank line if you've written a body.
- References to any Bugzilla tickets or Github Issues, one per line if
there are multiple.
- Bugzilla tickets should be in the form of a full URL to the ticket.
- Github issues should be of the form "Issue #n", where n is the issue number.
- Each commit should have its own unique subject. Do not use the same subject for a series of commits in a branch of work.
- Keep commits small by committing often and only include related changes and tests together.
- You may be doing too much in a commit if...
- You can't get the subject of the commit message under 72 characters
- You are using a lot of "ands" or "ors" in the commit message
- You find yourself using lots of bullet points to enumerate all the work the commit does.
- Squash (combine) commits to keep logical units of changes grouped together in the same commit. Don't squash unrelated changes.
- Avoid mixing logical changes with style changes of unrelated code. Keep them as separate commits. If the style change makes it hard to see the logical change, do the style change in a different pull request.
- Avoid mixing code changes with code relocation. Keep them as separate commits. If the moved method has high churn, perhaps move the method in a separate pull request.
-
Use descriptive names for feature branches as they are included in the Git history.
# bad faster_tests rework_amazon_code # good remove_test_duplication_for_performance change_ems_amazon_to_be_region_specific
-
Write a good pull request message. By default, Github will use your branch name as the title. Adjust the title if this is not appropriate for the pull request.
- See writing a good commit message for information about writing a good pull request message, exchanging the word "subject" with "title".
-
All pull requests should have tests or mention that there are existing tests that cover the code changes.
-
Avoid large pull requests (e.g. 1000+ lines changed not counting any generated files such as test data).
- Consider if the changes should be done in separate pull request.
- Use the number of lines added/removed as an indicator of possible code smells.
-
Try to avoid having a commit with, for example, a spelling mistake, that is fixed in a subsequent commit in the same pull request. Use
git rebase -i
to clean up those commits to keep the history clean. -
If a pull request involves UI changes, consider adding a before/after set of screenshots to show what has changed visually.
-
Use @-mentions to request reviews from specific people.
-
When you add new commits to a pull request, be sure to @-mention others so they know you are ready for a new review.
-
Under no circumstances should customer names or customer related information be referenced in Github issues, error reports, commits, or pull requests.
-
For UI errors, the error message and stack trace are usually in production.log. A snippet from there with the entire UI transaction is needed, including the error message and the stack trace. A UI transaction starts with something that looks like
[----] I, [2013-08-22T04:39:11.910803 #24340:3fd36e0349dc] INFO -- : Started GET "/ems_infra/show/7" for 127.0.0.1 at 2013-08-22 00:39:11 -0400 [----] I, [2013-08-22T04:39:11.929926 #24340:3fd36e0349dc] INFO -- : Processing by EmsInfraController#show as HTML
and ends with something that looks like
[----] I, [2013-08-22T04:39:12.127578 #24340:3fd36e0349dc] INFO -- : Rendered layouts/_global_footer.html.erb (0.1ms) [----] I, [2013-08-22T04:39:12.127794 #24340:3fd36e0349dc] INFO -- : Completed 200 OK in 198ms (Views: 110.1ms | ActiveRecord: 15.4ms)
Everything in between with the same PID (the number between the # and : symbols in the log line header) is important.
-
For non-UI errors (or errors that appear in the UI but are really backend errors), the error message is usually in the evm.log. Typically, we need more information than just the error message and stack trace, so it is helpful to have some extra context lines above and below the error. The amount is hard to quantify, so we will have to build up a list of things as time moves forward.
Most items are handled by a queue worker of some type, so for those it is most useful to have:
- The MiqQueue.put or MiqQueue.merge line from the worker that placed the item on the queue.
- Context around why the item was placed on the queue. For example, if some operation caused a policy item to be placed on the queue, is useful to have the log lines around what that operation was that triggered a policy check. This trail going backwards usually continues until you get a UI action or a scheduler action.
- The MiqQueue.get line from the worker that picked up the queue item.
- All work from the PID of the worker that picked up the item, up to and including the error message and stack trace.
- On occasion, incidental work done by other workers in the same time frame. Since all logs are UTC based, it makes it easier to coordinate log times from multiple appliances, regardless of where the logs live. This type of information is usually needed when there are environmental or coincidental errors occurring across all or part of the system.
-
Screenshots of the "UI error screen" are not useful at all, as all it shows is the exact error that is in the logs, but without all of the rest of the required information listed above.
-
For bugs on an old version of the product, it is helpful to know if the behavior is repeatable on the latest version on that z-stream. In addition, it is helpful to know if the behavior is repeatable on the latest of the next release, or even upstream.
When extracting code into a new gem or creating a new gem:
- Follow SemVer.
- Follow the internal guide for a creating a new gem.
- Create a CONTRIBUTING.md that references these guides. Project-specific changes or additions to the guides can be put here.
- Create a LICENSE.txt with an appropriate license.
Note after the changes in this section, you will need git push -f
if you have
already pushed them before.
-
Reword/squashing/reordering a commit
To modify with recent commit in current branch, first do
git rebase -i origin branch-name
. To modify a specific commit, usegit rebase -i SOME_COMMIT_ID^
instead. git will popup a vi window to let you do modification on commits, press:wq
after done.-
Reword a commit
Change the
pick
before the commit you want to reword toedit
and edit its message in popup vi window. -
Squashing commits
Change the
pick
before the commit you want to squach tosquash
and edit the commit message after squash in a following popup vi window. A commit will be squashed with its previous commit. -
Reordering commits
Reordering them in this vi window will reorder the commits.
-
-
Amend a commit
You can commit first and rebase it in the previous section. Or if you want to amend most recent commit, you can:
git commit some_file --amend
. -
Deleting a commit
You can delete commits by delete corresponding lines in
git rebase
. Or if you want to delete most recent commit, you cangit reset --hard HEAD^
. If you want to go back to a specific commit and delete commits after that, usegit reset --hard commit-hash
. -
Uncommit a file from an existing commit
git reset HEAD^ path/to/file/to/revert git commit --amend
- Rails style guide see https://github.com/bbatsov/rails-style-guide
- Bugzilla how-to
- how to copy commit details / clone a ticket
This work is licensed under a Creative Commons Attribution 3.0 Unported License