How to create a custom flavor

[BD103]

Hi! I've been working on improving the Bevy Engine's website CI. We use megalinter only for DJLint, Markdownlint, and Editorconfig Checker. It seems a bit silly to pull in a 3.5 GB image for only three linters, which takes about 2.5 minutes in CI.

How would I create my own flavor that only contained the above linters? I'd most likely fork this repo, but where would I go from there? I looked over the contributing page on the website, but could not find any resources on this. (Also the amount of files in the repo is intimidating 😅)

[echoix]

Well, it depends on what your strengths are. There's a hard way, where you would manipulate docker images, since individual linters are published as images, so it would be possible to combine them by copying files around, and adjusting the env vars that are set at the end of the dockerfile. Editorconfig-checker is an easy one, since it is a single file copied from a docker image. That's if you are more comfortable in docker.

Another way would be to run megalinter three individual linters back to back.
If you are confortable in docker, in another way, it could be technically possible to repack the megalinter images with eStargz compression/indexing, that would allow if using containerd storage backend to not have to pull the container completely when running, and download on the fly the parts needed. So since you don't use all of the image, not all would be downloaded. When using actions on github that are based on docker images, Github pulls the container before the workflow starts, so containerd isn't available when it is pulled. So using mega-linter-runner might be needed (it is simply a wrapper that helps giving all the arguments to the docker call).

Finally, the more durable way, but needs to be maintained like a downstream fork, is adding a flavor here:

megalinter/megalinter/flavor_factory.py

Lines 41 to 68 in 3efbff3

	def list_megalinter_flavors():
	flavors = {
	"all": {"label": "MegaLinter for any type of project"},
	"ci_light": {
	"label": "Optimized for CI items (Dockerfile, Jenkinsfile, JSON/YAML schemas, XML)"
	},
	"cupcake": {"label": "MegaLinter for the most commonly used languages"},
	"c_cpp": {"label": "Optimized for pure C/C++ projects"},
	"documentation": {"label": "Optimized for documentation projects"},
	"dotnet": {"label": "Optimized for C, C++, C# or VB based projects"},
	"dotnetweb": {
	"label": "Optimized for C, C++, C# or VB based projects with JS/TS"
	},
	"formatters": {"label": "Contains only formatters"},
	"go": {"label": "Optimized for GO based projects"},
	"java": {"label": "Optimized for JAVA based projects"},
	"javascript": {
	"label": "Optimized for JAVASCRIPT or TYPESCRIPT based projects"
	},
	"php": {"label": "Optimized for PHP based projects"},
	"python": {"label": "Optimized for PYTHON based projects"},
	"ruby": {"label": "Optimized for RUBY based projects"},
	"rust": {"label": "Optimized for RUST based projects"},
	"salesforce": {"label": "Optimized for Salesforce based projects"},
	"security": {"label": "Optimized for security", "strict": True},
	"swift": {"label": "Optimized for SWIFT based projects"},
	"terraform": {"label": "Optimized for TERRAFORM based projects"},
	}

then adding that flavor name in descriptor_flavors at the descriptor-level (the complete language/group), or at the linter-level, like

megalinter/megalinter/descriptors/repository.megalinter-descriptor.yml

Lines 1 to 17 in 3efbff3

	descriptor_id: REPOSITORY
	descriptor_type: other
	descriptor_flavors:
	- all_flavors # Applicable to CI in any language project
	- ci_light
	- cupcake
	lint_all_files: true
	linters:
	# CHECKOV
	- linter_name: checkov
	name: REPOSITORY_CHECKOV
	can_output_sarif: true
	descriptor_flavors:
	- all_flavors # Applicable to CI in any language project
	- cupcake
	- terraform
	- security

Then each time, just run ./build.sh and all the generated files get synced from these descriptors (including the dockerfiles). You could now build the images from the dockerfile of your flavor. Note that you'll probably have a hard time updating from upstream at each update with a lot of conflicts, but only the 3 added lines for your three linters and the added line in the descriptor would need to be kept, the rest would be generated back.

[BD103]

Thank you! I'm going to try that :)