Hello, and welcome! Whether you are looking for help, trying to report a bug, thinking about getting involved in the project, or about to submit a patch, this document is for you! It intends to be both an entry point for newcomers to the community (with various technical backgrounds), and a guide/reference for contributors and maintainers.
Consult the Table of Contents below, and jump to the desired section.
- Where to seek for help?
- Where to report bugs?
- Where to submit feature requests?
- Contributing
- Code style
If you are a Kong Enterprise customer, you may contact the Enterprise Support channels by opening an Enterprise support ticket on https://support.konghq.com.
If you are experiencing a P1 issue, please call at the 24/7 Enterprise Support phone line for immediate assistance, as published in the Customer Success Reference Guide.
If you are interested in becoming a Kong Enterprise customer, please to visit https://konghq.com/kong-enterprise-edition/ or contact us at sales@konghq.com.
There are multiple channels where you can get answers from the community or the maintainers of this project:
- Our public forum, Kong Nation is great for asking questions, giving advice, and staying up-to-date with the latest announcements. Kong Nation is frequented by Kong maintainers.
- Two chat channels are used by the community, but are rarely visited by Kong maintainers:
Please avoid opening GitHub issues for general questions or help, as those should be reserved for actual bug reports. The Kong community is welcoming and more than willing to assist you on those channels!
Feel free to submit an issue on the GitHub repository, we would be grateful to hear about it! Please make sure that you respect the GitHub issue template, and include:
- A summary of the issue
- A list of steps to help reproduce the issue
- The version of Kong that you encountered the issue with
- Your Kong configuration, or the parts that are relevant to your issue
If you wish, you are more than welcome to propose a patch to fix the issue! See the Submit a patch section for more information on how to best do so.
You can submit an issue for feature requests. Please make sure to add as much detail as you can when doing so.
You are also welcome to propose patches adding new features. See the section on Submitting a patch for details.
We welcome contributions of all kinds, there is no need to do code to be helpful! All of the following tasks are noble and worthy contributions that you can make without coding:
- Reporting a bug (see the report bugs section)
- Helping other members of the community on the support channels
- Fixing a typo in the code
- Fixing a typo in the documentation at https://docs.konghq.com (see the documentation contribution section)
- Providing your feedback on the proposed features and designs
- Reviewing Pull Requests
If you wish to contribute code (features or bug fixes), see the Submitting a patch section.
The documentation hosted at https://docs.konghq.com is open source and built with Jekyll. You are very welcome to propose changes to it (correct typos, add examples or clarifications...) and contribute to the Kong Hub!
The repository is also hosted on GitHub at: https://github.com/Kong/docs.konghq.com/
We do not accept new plugins into the core repository. The plugins that are currently part of this repository are there because of historical reasons, but will be pushed into separate repositories in the foreseeable future.
If you are interested in writing a new plugin for your own needs, you should begin by reading the Plugin Development Guide.
If you already wrote a plugin, and are thinking about making it available to the community, we strongly encourage you to host it on a publicly available repository (like GitHub), and to distribute it via LuaRocks. A good resource on how to do so is the Distribution Section of the Plugin Development Guide.
To give visibility to your plugin, we advise that you:
- Add your plugin to the Kong Hub
- Create a post in the Announcements category of Kong Nation
Feel free to contribute fixes or minor features, we love to receive Pull Requests! If you are planning to develop a larger feature, come talk to us first!
When contributing, please follow the guidelines provided in this document. They will cover topics such as the different Git branches we use, the commit message format to use, or the appropriate code style.
Once you have read them, and you feel that you are ready to submit your Pull Request, be sure to verify a few things:
- Your commit history is clean: changes are atomic and the git message format was respected
- Rebase your work on top of the base branch (seek help online on how to use
git rebase
; this is important to ensure your commit history is clean and linear) - The static linting is succeeding: run
make lint
, orluacheck .
(see the development documentation for additional details) - The tests are passing: run
make test
,make test-all
, or whichever is appropriate for your change - Do not update CHANGELOG.md yourself. Your change will be included therein due time if it is accepted, no worries!
If the above guidelines are respected, your Pull Request has all its chances to be considered and will be reviewed by a maintainer.
If you are asked to update your patch by a reviewer, please do so! Remember: you are responsible for pushing your patch forward. If you contributed it, you are probably the one in need of it. You must be ready to apply changes to it if necessary.
If your Pull Request was accepted and fixes a bug, adds functionality, or makes it significantly easier to use or understand Kong, congratulations! You are now an official contributor to Kong. Get in touch with us to receive your very own Contributor T-shirt!
Your change will be included in the subsequent release Changelog, and we will not forget to include your name if you are an external contributor. 😉
If you have write access to the GitHub repository, please follow the following naming scheme when pushing your branch(es):
feat/foo-bar
for new featuresfix/foo-bar
for bug fixestests/foo-bar
when the change concerns only the test suiterefactor/foo-bar
when refactoring code without any behavior changestyle/foo-bar
when addressing some style issuedocs/foo-bar
for updates to the README.md, this file, or similar documents
When submitting patches, it is important that you organize your commits in logical units of work. You are free to propose a patch with one or many commits, as long as their atomicity is respected. This means that no unrelated changes should be included in a commit.
For example: you are writing a patch to fix a bug, but in your endeavour, you spot another bug. Do not fix both bugs in the same commit! Finish your work on the initial bug, propose your patch, and come back to the second bug later on. This is also valid for unrelated style fixes, refactors, etc...
You should use your best judgment when facing such decisions. A good approach for this is to put yourself in the shoes of the person who will review your patch: will they understand your changes and reasoning just by reading your commit history? Will they find unrelated changes in a particular commit? They shouldn't!
Writing meaningful commit messages that follow our commit message format will also help you respect this mantra (see the below section).
To maintain a healthy Git history, we ask of you that you write your commit messages as follows:
- The tense of your message must be present
- Your message must be prefixed by a type, and a scope
- The header of your message should not be longer than 50 characters
- A blank line should be included between the header and the body
- The body of your message should not contain lines longer than 72 characters
Here is a template of what your commit message should look like:
<type>(<scope>) <subject>
<BLANK LINE>
<body>
<BLANK LINE>
<footer>
The type of your commit indicates what type of change this commit is about. The accepted types are:
- feat: A new feature
- fix: A bug fix
- hotfix: An urgent bug fix during a release process
- tests: A change that is purely related to the test suite only (fixing a test, adding a test, improving its reliability, etc...)
- docs: Changes to the README.md, this file, or other such documents
- style: Changes that do not affect the meaning of the code (white-space trimming, formatting, etc...)
- perf: A code change that significantly improves performance
- refactor: A code change that neither fixes a bug nor adds a feature, and
is too big to be considered just
perf
- chore: Maintenance changes related to code cleaning that isn't considered part of a refactor, build process updates, dependency bumps, or auxiliary tools and libraries updates (LuaRocks, Travis-ci, etc...).
The scope is the part of the codebase that is affected by your change. Choosing it is at your discretion, but here are some of the most frequent ones:
- proxy: A change that affects the proxying of requests
- router: A change that affects the router, which matches a request to the desired configured API
- admin: A change to the Admin API
- balancer: Changes related to the internal Load Balancer
- core: Changes affecting a large part of the core, and touching many parts
such as
proxy
,balancer
,dns
- dns: Changes related to internal DNS resolution
- dao: A change related to the DAO, the interface to the datastores
- cli: Changes to the CLI
- cache: Changes to the configuration entities caching (datastore entities)
- deps: When updating dependencies (to be used with the
chore
prefix) - conf: Configuration-related changes (new values, improvements...)
<plugin-name>
: This could bebasic-auth
, orldap
for example*
: When the change affects too many parts of the codebase at once (this should be rare and avoided)
Your subject should contain a succinct description of the change. It should be written so that:
- It uses the present, imperative tense: "fix typo", and not "fixed" or "fixes"
- It is not capitalized: "fix typo", and not "Fix typo"
- It does not include a period. 😄
The body of your commit message should contain a detailed description of your changes. Ideally, if the change is significant, you should explain its motivation, the chosen implementation, and justify it.
As previously mentioned, lines in the commit messages should not exceed 72 characters.
The footer is the ideal place to link to related material about the change: related GitHub issues, Pull Requests, fixed bug reports, etc...
Here are a few examples of good commit messages to take inspiration from:
fix(admin) send HTTP 405 on unsupported method
The appropriate status code when the request method is not supported
on an endpoint it 405. We previously used to send HTTP 404, which
is not appropriate. This updates the Admin API helpers to properly
return 405 on such user errors.
* return 405 when the method is not supported in the Admin API helpers
* add a new test case in the Admin API test suite
Fix #678
Or:
tests(proxy) add a new test case for URI encoding
When proxying upstream, the URI sent by Kong should be the one
received from the client, even if it was percent-encoded.
This adds a new test case which was missing, to ensure it is
the case.
As mentioned in the guidelines to submit a patch, the linter must succeed. We use Luacheck to statically lint our Lua code. You can lint the code like so:
$ make lint
Or:
$ luacheck .
We use busted to write our tests. Your patch must include the related test updates or additions, in the appropriate test suite.
spec/01-unit
gathers our unit tests (to test a given Lua module or function)spec/02-integration
contains tests that start Kong (connected to a running database), execute Admin API and proxy requests against it, and verify the outputspec/03-plugins
contains tests (both unit and integration) for the bundled plugins (those plugins still live in the core repository as of now, but will eventually be externalized)
A few guidelines when writing tests:
- Make sure to use appropriate
describe
andit
blocks, so it's obvious to what is being tested exactly - Ensure the atomicity of your tests: no test should be asserting two unrelated behaviors at the same time
- Run tests related to the datastore against all supported databases
And a few recommendations, when asserting types:
-- bad
assert.Nil(foo)
assert.True(bar)
-- good
assert.is_nil(foo)
assert.is_true(bar)
Comparing tables:
-- bad (most of the time)
assert.equal(t1, t2)
-- good
assert.same(t1, t2)
We write code for the LuaJIT interpreter, not Lua-PUC. As such, you should follow the LuaJIT best practices:
-
Do not instantiate global variables
-
Consult the LuaJIT wiki
-
Follow the Performance Guide recommendations
-
Do not use NYI functions on hot code paths
-
Prefer using the FFI over traditional bindings via the Lua C API
-
Avoid table rehash by pre-allocating the slots of your tables when possible
-- bad local t = {} for i = 1, 100 do t[i] = i end -- good local new_tab = require "table.new" local t = new_tab(100, 0) for i = 1, 100 do t[i] = i end
-
Cache the globals used by your hot code paths
-- bad for i = 1, 100 do t[i] = math.random() end -- good local random = math.random for i = 1, 100 do t[i] = random() end
-
Cache the length and indices of your tables to avoid unnecessary CPU cycles
-- bad for i = 1, 100 do t[#t + 1] = other_tab[#other_tab] end -- good local n = 0 local n_other_tab = #other_tab for i = 1, 100 do n = n + 1 t[n] = other_tab[n_other_tab] end
And finally, most importantly: use your best judgment to design an efficient algorithm. Doing so will always be more performant than a poorly-designed algorithm, even following all the performance tricks of the language you are using. 😄
If your Pull Request to Kong/kong was accepted, and it fixes a bug, adds functionality, or makes it significantly easier to use or understand Kong, congratulations! You are eligible to receive the very special Contributor T-shirt! Go ahead and fill out the Contributors Submissions form.
Proudly wear your T-shirt and show it to us by tagging @thekonginc on Twitter!
In order to ensure a healthy and consistent codebase, we ask of you that you respect the adopted code style. This section contains a non-exhaustive list of preferred styles for writing Lua. It is opinionated, but follows the code styles of OpenResty and, by association, Nginx. OpenResty or Nginx contributors should find themselves at ease when contributing to Kong.
- No line should be longer than 80 characters
- Indentation should consist of 2 spaces
When you are unsure about the style to adopt, please browse other parts of the codebase to find a similar case, and stay consistent with it.
You might also notice places in the codebase where the described style is not respected. This is due to legacy code. Contributions to update the code to the recommended style are welcome!
When writing a module (a Lua file), separate logical blocks of code with two blank lines:
local foo = require "kong.foo"
local _M = {}
function _M.bar()
-- do thing...
end
function _M.baz()
-- do thing...
end
return _M
When naming a variable or function, do use snake_case:
-- bad
local myString = "hello world"
-- good
local my_string = "hello world"
When assigning a constant variable, do give it an uppercase name:
-- bad
local max_len = 100
-- good
local MAX_LEN = 100
Use the constructor syntax, and do include a trailing comma:
-- bad
local t = {}
t.foo = "hello"
t.bar = "world"
-- good
local t = {
foo = "hello",
bar = "world", -- note the trailing comma
}
On single-line constructors, do include spaces around curly-braces and assignments:
-- bad
local t = {foo="hello",bar="world"}
-- good
local t = { foo = "hello", bar = "world" }
Do favor the use of double quotes in all Lua code (plain files and
*_by_lua_block
directives):
-- bad
local str = 'hello'
-- good
local str = "hello"
If a string contains double quotes, do favor long bracket strings:
-- bad
local str = "message: \"hello\""
-- good
local str = [[message: "hello"]]
When using the concatenation operator, do insert spaces around it:
-- bad
local str = "hello ".."world"
-- good
local str = "hello " .. "world"
Prefer the function syntax over variable syntax:
-- bad
local foo = function()
end
-- good
local function foo()
end
Perform validation early and return as early as possible:
-- bad
local function check_name(name)
local valid = #name > 3
valid = valid and #name < 30
-- other validations
return valid
end
-- good
local function check_name(name)
if #name <= 3 or #name >= 30 then
return false
end
-- other validations
return true
end
Follow the return values conventions: Lua supports multiple return values, and
by convention, handles recoverable errors by returning nil
plus a string
describing the error:
-- bad
local function check()
local ok, err = do_thing()
if not ok then
return false, { message = err }
end
return true
end
-- good
local function check()
local ok, err = do_thing()
if not ok then
return nil, "could not do thing: " .. err
end
return true
end
When a function call makes a line go over 80 characters, do align the overflowing arguments to the first one:
-- bad
local str = string.format("SELECT * FROM users WHERE first_name = '%s'", first_name)
-- good
local str = string.format("SELECT * FROM users WHERE first_name = '%s'",
first_name)
Avoid writing 1-liner conditions, do indent the child branch:
-- bad
if err then return nil, err end
-- good
if err then
return nil, err
end
When testing the assignment of a value, do use shortcuts, unless you
care about the difference between nil
and false
:
-- bad
if str ~= nil then
end
-- good
if str then
end
When creating multiple branches that span multiple lines, do include a
blank line above the elseif
and else
statements:
-- bad
if foo then
do_stuff()
keep_doing_stuff()
elseif bar then
do_other_stuff()
keep_doing_other_stuff()
else
error()
end
-- good
if thing then
do_stuff()
keep_doing_stuff()
elseif bar then
do_other_stuff()
keep_doing_other_stuff()
else
error()
end
For one-line blocks, blank lines are not necessary:
--- good
if foo then
do_stuff()
else
error("failed!")
end
Note in the correct "long" example that if some branches are long, then all
branches are created with the preceding blank line (including the one-liner
else
case).
When a branch returns, do not create subsequent branches, but write the rest of your logic on the parent branch:
-- bad
if not str then
return nil, "bad value"
else
do_thing(str)
end
-- good
if not str then
return nil, "bad value"
end
do_thing(str)
When assigning a value or returning from a function, do use ternaries if it makes the code more readable:
-- bad
local foo
if bar then
foo = "hello"
else
foo = "world"
end
-- good
local foo = bar and "hello" or "world"
When an expression makes a line longer than 80 characters, do align the expression on the following lines:
-- bad
if thing_one < 1 and long_and_complicated_function(arg1, arg2) < 10 or thing_two > 10 then
end
-- good
if thing_one < 1 and long_and_complicated_function(arg1, arg2) < 10
or thing_two > 10
then
end