KAL is a Golang based linter for Kubernetes API types. It checks for common mistakes and enforces best practices. The rules implemented by KAL, are based on the Kubernetes API Conventions.
KAL is aimed at being an assistant to API review, by catching the mechanical elements of API review, and allowing reviewers to focus on the more complex aspects of API design.
KAL currently comes in two flavours, a standalone binary, and a golangci-lint plugin.
To install the standalone binary, run the following command:
go install github.com/JoelSpeed/kal/cmd/kal@latestThe standalone binary can be run with the following command:
kal path/to/api/typeskal currently accepts no complex configuration, and will run all checks considered to be default.
Individual linters can be disabled with the flag corresponding to the linter name. For example, to disable the commentstart linter, run the following command:
kal -commentstart=false path/to/api/typesWhere fixes are available, these can be applied automatically with the -fix flag.
Note, automatic fixing is currently only available via the standalone binary, and is not available via the golangci-lint plugin.
kal -fix path/to/api/typesOther standard Golang linter flags implemented by multichecker based linters are also supported.
To install the golangci-lint plugin, first you must have golangci-lint installed.
If you do not have golangci-lint installed, review the golangci-lint install guide.
You will need to create a .custom-gcl.yml file to describe the custom linters you want to run. The following is an example of a .custom-gcl.yml file:
version: v1.62.0
name: golangci-kal
destination: ./bin
plugins:
- module: 'github.com/JoelSpeed/kal'
version: 'v0.0.0' # Replace with the latest versionOnce you have created the custom configuration file, you can run the following command to build the custom golangci-kal binary:
golangci-lint customThe output binary will be a combination of the initial golangci-lint binary and the KAL linters.
This means that you can use any of the standard golangci-lint configuration or flags to run the binary, but may also include the KAL linters.
If you wish to only use the KAL linters, you can configure your .golangci.yml file to only run the KAL linters:
linters-settings:
custom:
kal:
type: "module"
description: KAL is the Kube-API-Linter and lints Kube like APIs based on API conventions and best practices.
settings:
linters: {}
lintersConfig: {}
linters:
disable-all: true
enable:
- kalThe settings for KAL are based on the GolangCIConfig struct and allow for finer control over the linter rules. The finer control over linter rules is not currently avaialable outside of the plugin based version of KAL.
If you wish to use the KAL linters in conjunction with other linters, you can enable the KAL linters in the .golangci.yml file by ensuring that kal is in the linters.enabled list.
To provide further configuration, add the custom.kal section to your linter-settings as per the example above.
Since VSCode already integrates with golangci-lint via the Go extension, you can use the golangci-kal binary as a linter in VSCode.
If your project authors are already using VSCode and have the configuration to lint their code when saving, this can be a seamless integration.
Ensure that your project setup includes building the golangci-kal binary, and then configure the go.lintTool and go.alternateTools settings in your project .vscode/settings.json file.
{
"go.lintTool": "golangci-lint",
"go.alternateTools": {
"golangci-lint": "${workspaceFolder}/bin/golangci-kal",
}
}Alternatively, you can also replace the binary with a script that runs the golangci-kal binary,
allowing for customisation or automatic copmilation of the project should it not already exist.
{
"go.lintTool": "golangci-lint",
"go.alternateTools": {
"golangci-lint": "${workspaceFolder}/hack/golangci-lint.sh",
}
}The conditions linter checks that Conditions fields in the API types are correctly formatted.
The Conditions field should be a slice of metav1.Condition with the following tags and markers:
// +listType=map
// +listMapKey=type
// +patchStrategy=merge
// +patchMergeKey=type
// +optional
Conditions []metav1.Condition `json:"conditions,omitempty" patchStrategy:"merge" patchMergeKey:"type" protobuf:"bytes,1,opt,name=conditions"`Conditions are idiomatically the first field within the status struct, and the linter will highlight when the Conditions are not the first field.
lintersConfig:
conditions:
isFirstField: Warn | Ignore # The policy for the Conditions field being the first field. Defaults to `Warn`.
useProtobuf: SuggestFix | Warn | Ignore # The policy for the protobuf tag on the Conditions field. Defaults to `SuggestFix`.The conditions linter can automatically fix the tags on the Conditions field.
When they do not match the expected format, the linter will suggest to update the tags to match the expected format.
For CRDs, protobuf tags are not expected. By setting the useProtobuf configuration to Ignore, the linter will not suggest to add the protobuf tag to the Conditions field tags.
The linter will also suggest to add missing markers. If any of the 5 markers in the example above are missing, the linter will suggest to add them directly above the field.
The commentstart linter checks that all comments in the API types start with the serialized form of the type they are commenting on.
This helps to ensure that generated documentation reflects the most common usage of the field, the serialized YAML form.
The commentstart linter can automatically fix comments that do not start with the serialized form of the type.
When the json tag is present, and matches the first word of the field comment in all but casing, the linter will suggest that the comment be updated to match the json tag.
The integers linter checks for usage of unsupported integer types.
Only int32 and int64 types should be used in APIs, and other integer types, including unsigned integers are forbidden.
## JSONTags
The jsontags linter checks that all fields in the API types have a json tag, and that those tags are correctly formatted.
The json tag for a field within a Kubernetes API type should use a camel case version of the field name.
The jsontags linter checks the tag name against the regex "^[a-z][a-z0-9]*(?:[A-Z][a-z0-9]*)*$" which allows consecutive upper case characters, to allow for acronyms, e.g. requestTTL.
lintersConfig:
jsonTags:
jsonTagRegex: "^[a-z][a-z0-9]*(?:[A-Z][a-z0-9]*)*$" # Provide a custom regex, which the json tag must match.The nobools linter checks that fields in the API types do not contain a bool type.
Booleans are limited and do not evolve well over time. It is recommended instead to create a string alias with meaningful values, as an enum.
The nophase linter checks that the fields in the API types don't contain a 'Phase', or any field which contains 'Phase' as a substring, e.g MachinePhase.
The optionalorrequired linter checks that all fields in the API types are either optional or required, and are marked explicitly as such.
The linter expects to find a comment marker // +optional or // +required within the comment for the field.
It also supports the // +kubebuilder:validation:Optional and // +kubebuilder:validation:Required markers, but will suggest to use the // +optional and // +required markers instead.
If you prefer to use the Kubebuilder markers instead, you can change the preference in the configuration.
lintersConfig:
optionalOrRequired:
preferredOptionalMarker: optional | kubebuilder:validation:Optional # The preferred optional marker to use, fixes will suggest to use this marker. Defaults to `optional`.
preferredRequiredMarker: required | kubebuilder:validation:Required # The preferred required marker to use, fixes will suggest to use this marker. Defaults to `required`.The optionalorrequired linter can automatically fix fields that are using the incorrect form of either the optional or required marker.
It will also remove the secondary marker where both the preferred and secondary marker are present on a field.
The requiredfields linter checks that fields that are marked as required, follow the convention of not being pointers,
and not having an omitempty value in their json tag.
lintersConfig:
requiredFields:
pointerPolicy: Warn | SuggestFix # The policy for pointers in required fields. Defaults to `SuggestFix`.The requiredfields linter can automatically fix fields that are marked as required, but are pointers.
It will suggest to remove the pointer from the field, and update the json tag to remove the omitempty value.
If you prefer not to suggest fixes for pointers in required fields, you can change the pointerPolicy to Warn.
The linter will then only suggest to remove the omitempty value from the json tag.
New linters can be added by following the New Linter guide.
KAL is licensed under the Apache License, Version 2.0. See LICENSE for the full license text.