Welcome! This guide is intended to help developers that are new to the community navigate the process of making contributions to the Redpanda project. Please be sure to also checkout our code of conduct. We work hard to make this a welcoming community for all, and we're excited that you are here!
The basics:
-
Use the Issue Tracker to report bugs, crashes, performance issues, etc... Please include as much detail as possible.
-
Discussions is a great venue to ask questions, start a design discussion, or post an RFC.
-
We ask that developers sign our contributor license agreement. The process of signing the CLA is automated, and you'll be prompted instructions the first time you submit a pull request to the project.
We use a standard GitHub fork + pull request model for merging and reviewing
changes. New pull requests should be made against the upstream dev branch.
After a pull request is opened it will be reviewed, and merged after
passing continuous integration tests and being accepted by a project or
sub-system maintainer.
In addition to the content of the changes being made, reviews will consider three additional properties:
- Formatting of code and documentation
- Composition of commit history
- Content of commit messages
We care a great deal about the hygiene of the patches that are merged into the project, and we've designed our review process and standards to reflect that. However, we're always trying to improve, and welcome any suggestions and feedback.
Located in the root of the Redpanda project are style property files for various languages used in the project (e.g. clang-format for C/C++). Our continuous integration checks enforce these styles by verifying changes made in each pull request. All mainstream editors and IDEs should provide support (either natively or through a plugin) for integrating with code formatting tools that can consume the project property files. Configuring your editor to do this is highly recommended.
Styling rules are enforced at the granularity of the pull request, rather than on each commit. This is important because occasionally a small change may cause a large reformatting, which makes reviewing changes difficult. In this case it would be appropriate to split the code change and formatting into separate commits. This is usually unnecessary.
The ordering and structure of commits in a pull request is very important. This is because the way in which a large change is broken down into individual commits has a significant impact on the review process and later on the ability to efficiently find certain types of regressions. A good rule of thumb is that a set of changes is in good shape when the following properties hold:
- Each commit introduces a single logical change
- Each commit represents a tree that builds without errors
Combined, these two properties tend to produce a history with changes that are
easier to review (smaller, focused commits), and ensure that tools like git bisect are able to be applied to the tree.
Here is an example of commits in a hypothetical pull request:
b9f49db32 credits: opening sequence
b3032f211 scene: morty and summer enter the garage
60a49a8b9 scene: dog robots build space laser
a34b3eaa0 scene: morty turns into a glass jar
187547195 credit: closing sequence
After receiving feedback, it may be the case that the following additional changes are made:
b9f49db32 credits: opening sequence
b3032f211 scene: morty and summer enter the garage
60a49a8b9 scene: dog robots build space laser
a34b3eaa0 scene: morty turns into a glass jar
187547195 credit: closing sequence
// extra commits
20929ab22 add extra joke to garage scene
330fff211 add a new scene in living room
b0329aaa1 morty glass jar should also levitate
Before merging we expect that the extra commits be folded back into any previous commits that they modify, resulting in concise commit history that doesn't reflect the intermediate changes made based on feedback. In this example, the following is a reasonable final set of commits:
cbff210db Merge pull request #123 from rick/morty
b9f49db32 credits: opening sequence
b3032f211 scene: morty and summer enter the garage
60a49a8b9 scene: dog robots build space laser
a34b3eaa0 scene: morty turns into a glass jar
330fff211 scene: summer and snuffles watching tv
187547195 credit: closing sequence
It is common to force push to the pull request branch with the modified history.
In this case it can be useful to use the --keep-base option to git rebase
which will ensure a minimal change set for reviewers. However, depending on the
nature of pull request and feedback it may be useful to push extra commits and
clean up the history as a last step.
In addition to code formatting and commit history discussed above, the commit messages themselves (both the format and content being communicated) are also important. First let's take a look at formatting.
It may be surprising that we consider commit message formatting to be important enough to have its own section in the contributors guide. There are a couple reasons for this. First, in our experience we have found that adopting a standard for issues related to style and formatting help reduce cognitive load when consuming content like code or technical writing. But from a practical stand point, many developers use terminal based editors with restrictions on line width which is the primary dimension to the formatting guidelines of commit messages.
Here is a sample commit message and below we'll discuss formatting guidelines in the context of this message:
server: display redpanda logo in log when starting
When redpanda starts it logs lots of important messages like the active
configuration and details about the system resources. However, it fails
to do anything fun. This patch resolves this shortcoming by displaying
the following redpanda ascii art logo when the system starts.
___ ___ ___ ___ ___ ___
/\ \ /\__\ _____ /\ \ /\ \ /\ \ _____ /\ \
/::\ \ /:/ _/_ /::\ \ /::\ \ /::\ \ \:\ \ /::\ \ /::\ \
/:/\:\__\ /:/ /\__\ /:/\:\ \ /:/\:\__\ /:/\:\ \ \:\ \ /:/\:\ \ /:/\:\ \
/:/ /:/ / /:/ /:/ _/_ /:/ \:\__\ /:/ /:/ / /:/ /::\ \ _____\:\ \ /:/ \:\__\ /:/ /::\ \
/:/_/:/__/___ /:/_/:/ /\__\ /:/__/ \:|__| /:/_/:/ / /:/_/:/\:\__\ /::::::::\__\ /:/__/ \:|__| /:/_/:/\:\__\
\:\/:::::/ / \:\/:/ /:/ / \:\ \ /:/ / \:\/:/ / \:\/:/ \/__/ \:\~~\~~\/__/ \:\ \ /:/ / \:\/:/ \/__/
\::/~~/~~~~ \::/_/:/ / \:\ /:/ / \::/__/ \::/__/ \:\ \ \:\ /:/ / \::/__/
\:\~~\ \:\/:/ / \:\/:/ / \:\ \ \:\ \ \:\ \ \:\/:/ / \:\ \
\:\__\ \::/ / \::/ / \:\__\ \:\__\ \:\__\ \::/ / \:\__\
\/__/ \/__/ \/__/ \/__/ \/__/ \/__/ \/__/ \/__/
The logo is printed to standard out on start-up and when logging to a
normal log file. It is not displayed when logging to journald as the
formatting is not particularly nice.
The ascii art string literal is annotated with a compiler directive to
indicate that this string is rarely used and can be stored in a location
optimized for such data.
The format of the subject is area[/detail]: short description and the entire
line should be no more than 50 characters in length. The prefix: is helpful
for readers to quickly identify the relevant subsystem. We recommend taking
a look at recent Git history for examples of common prefix: choices.
There is no restriction on the length of the body, but lines should wrap at 72 characters. The exception to this rule is text that should be formatted verbatim like log messages, stack traces, and redpanda ascii art.
While there is no algorithm for writing the perfect commit message, we've found that working towards the following three goals is often a positive signal that a commit message is in good shape:
- Explain why the change is being made (the before, and after)
- Describe important / non-obvious technical details
- Require no extra context to understand
Good technical communication takes a lot of practice, and what makes sense to one person isn't always optimal for another reader. So our suggestion is to error on the side of more detail, and not stress about making it perfect. Read on below if you are curious for more details about these bullet points.
When composing a commit message it is important to consider that the content serves two distinct goals: optimizing for reviewer time and precision, and providing a historical narrative of changes. In the short term a commit message will be the first thing a reviewer consumes, and can have a dramatic impact on the review process, even for small changes which might require a lot of context to understand.
An important property of a commit message is that it is understandable when read in the future by developers that did not participate in the original design and review process. In complex systems like Redpanda it is not uncommon to review historical changes to sub-systems while debugging an issue or investigating regressions. But stumbling upon what looks like a critical change with no explanation can make this process slow and arduous. Since the chance of any one commit being the bottleneck in such a process are tiny, the only robust solution is to treat each commit as being likely involved in such a process.
A good rule of thumb is one paragraph for describing why the change is being made. This should generally stand independent of the physical change being made, and describe the state of the system being transitioned away from, why that state no longer desirable, and how the new state addresses those shortcomings.
It is much harder to decide what technical details are important or non-obvious. But, if you happened to think one aspect of the changes were interesting or challenging, reviewers and future debuggers will probably have the same--but magnified--reaction.
Finally, the commit message should stand on its own. This means that it should largely not depend on or refer to other commits in a series (e.g. "in the previous commit..." or "and in a later commit this will be implemented"), nor should it refer to out-of-band conversations (e.g. "in last week's meeting"). Refer to high-level context in the pull request description.