Contributing to Meshery CLI

mesheryctl is written in Golang or the Go Programming Language. For development use Go version 1.19+.

Mechanics of Contributing

Building mesheryctl

The /mesheryctl folder contains the complete code for mesheryctl. Fork and clone the Meshery repo. cd mesheryctl to change directory mesheryctl’s source.

After making changes, run make in the mesheryctl folder to build the binary. You can then use the binary by, say, ./mesheryctl system start.

Framework

mesheryctl uses the Cobra framework. A good first-step towards contributing to mesheryctl would be to familiarise yourself with the Cobra concepts. For manipulating config files, mesheryctl uses Viper.

Model and Configuration Data

A central struct is maintained in the mesheryctl/internal/cli/root/config/config.go file. These are updated and should be used for getting the Meshery configuration.

Updates to this central struct is made through updates in Context with setter functions. The changes made in this central struct are reflected back in the Meshery configuration file (.meshery/config.yaml).

Logging

For logs, mesheryctl uses Logrus. Going through the docs and understanding the different log-levels will help a lot.

Linting

mesheryctl uses golangci-lint. See the .github/workflow/ci.yaml for syntax used during Meshery’s build process.

Unit Tests

Unit test code coverage reports can be found in the CodeCov logs. Note: GitHub login may be required for access.

Documentation

The documentation pages for mesheryctl reference are made with the help of the Cobra Golang framework and use of GitHub Actions. Refer to Contributing to mesheryctl documentation for details.

Meshery CLI Style Guide

These guidelines are a collection of principles and conventions that need to be followed while designing mesheryctl commands. mesheryctl might be the interface that the users first have with Meshery. As such, mesheryctl needs to provide a great UX.

The following principles should be taken in mind while designing mesheryctl commands-

Design Principles

1. Consistency is quality.

• Consistency of interaction drives a quality user experience. Whether that experience is delightful or painful is a related, but separate consideration. Meshery’s behavior of user interactions should be consistent even when their user experience is poor.

2. Intuitive user experiences feel natural to users.

• When being designed, each of Meshery’s user experiences should be examined first from the user’s perspective. Design user experiences that are familiar.

3. Design for brevity.

• Avoid long commands with chained series of flags.

4. Design with automated testing in mind.

• Provide possibility to specify output format as json (-o json) for easy inspection of command response.

Part of delivering a great user experience is providing intuitive interfaces. In the case of mesheryctl takes inspiration from and delivers similar user experiences as popular CLIs do in this ecosystem, like kubectl and docker. Here is relevant kubectl information to reference - Kubectl SIG CLI Community Meeting Minutes, contributing to kubectl, code.

Command structure and command behavior should be designed in such a way that they are intuitive. Users should ideally be able to understand what a command is used for without having to extensively go through the documentation. For example, mesheryctl pattern apply -f <pattern name> requires no further clarification as it is evident that the command will apply the pattern specified.

Consistency is key when designing intuitive interfaces. Although mesheryctl perf run -f <performance profile name> may sound more intuitive, users who are experienced in using the CLI will prefer the consistant verb apply over run. This will also ensure a consistent command language making memorizing easier.

Flags

Consistency should also be enforced when chaining commands and using flags. For example, if mesheryctl pattern has a list and view command and has an -all and --output flag, then, similar commands like mesheryctl perf should also support the same commands and flags and provide a consistent user experience.

Rational defaults overriden with flags

Default behaviour should be optimised for what users will need to do most of the time.

These assumed defaults should be easily overriden by the user with flags.

For example, mesheryctl system context create <context name> assumes a default platform for the created context. But this can be easily overriden with the --platform flag.

User Experience: GUI vs CLI

Ideally, all functionaly provided in Meshery UI should be available to users via CLI (in mesheryctl). Meshery strives for parity of functionality between it’s two clients. For example, viewing a performance profile in the GUI and with mesheryctl system perf view <profile name> in the CLI should show the same data.

Command line interfaces offer less context to the user, which makes them inherently less intuitive compared to graphical user interfaces. Both of these user interfaces, however, are the same in that they are both clients of Meshery Server. Both clients are a user experience and as such, to be attended to in this way. The following considerations should be accounted for when designing mesheryctl experiences:

• Provide only relevant output. Use “debug” logs that can be accessed with --verbose flag wherever needed.
• Add headers to output to give context to the user.
• As mentioned above, similar commands should behave similarly.
• Confirm steps for risky commands. For example, use the AskForConfirmation function which will prompt the user to type in “yes” or “no” to continue.
• Anticipate user actions. If the user creates a new context with mesheryctl system context create then the next action might be mesheryctl system start to start Meshery ot mesheryctl system context switch to switch context names.
• Anticipate user errors. For example, if the user types in mesheryctl system satrt, using the inbuilt features with the cobra library, we can correct it to start and alert the user.

Designing Commands

If you are working on a new command or adding a new feature on an existing command, it is recommended to setup a design spec so that other contributors can weigh in on the design before you start to code.Broader features should have a design spec made in Google Doc. Check this Google Doc out, which has detailed information on creating a Design Specification. For small changes, communicating over the issue tracker or the discussions will be helpful.

When designing for the command line interface, ask and consider the following questions.

What the command does

• What makes sense to do from a terminal? What doesn’t?
• What might people want to automate?
• What is the default behavior? What flags might you need to change that behavior?
• What might people try and fail to do and how can you anticipate that?

What the command is called

• What should be the command language? (mesheryctl <command> <subcommand> [args] [flags] [value])
• What should be a command vs a flag?
• How can you align the language of the new command with the existing commands?

What are the command outputs

• How can you make the GUI and the CLI outputs similar?
• What should be outputted normally and what falls into debug logs?

How you would explain your command

You will need to provide a short and long description of the command for the help pages and also for the Meshery Documentation.

Writing unit tests and integration tests for mesheryctl

Unit tests and integration tests are essential to make each mesheryctl release robust and fault tolerant.

Below you will find guidelines to write unit tests and integration tests and examples of how they are implemented in mesheryctl.

Unit tests: Individual components are tested.

Integration tests: Individual components are combined and tested as a group.

Contributing to mesheryctl documentation

The Meshery CLI Reference is autogenerated based on docs sections in each of mesheryctl’s Golang files. Meshery CLI Reference pages are updated each time a change to its respective mesheryctl Golang file is merged into the project’s master branch. This approach to documentation facilitates a single source of truth for CLI syntax and command behavior, which results in higher quality reference and a reduction in the toil involved in keeping documentation up-to-date. To contribute to the Meshery CLI Reference, follow these steps:

• Go to the required command file in which the documentation has to be created/updated (mainly under /mesheryctl/internal/cli/root/…)
• Then, edit the Cobra macro variables present in the each file. An example is given below for reference.

var startCmd = &cobra.Command{
Use:   "start",
Short: "Start Meshery",
Long:  'Start Meshery and each of its service mesh components.',
Args:  cobra.NoArgs,
Example:```
// Start meshery
mesheryctl system start

// To create a new context for in-cluster Kubernetes deployments and set the new context as your current-context
mesheryctl system context create k8s -p kubernetes -s
```,
Annotations: linkScreenshot,
...

The variables present in above sample will be used in creating the doc pages for the specific command

Also, if the screenshot is present in the command, an Annotation macro variable (of map[string]string type) containing the link and the caption has to be added at the bottom of the Examples field in the command file. The image file has to be included in the docs/assets folder in PNG format. The screenshot field is given for reference below

var linkDocPatternApply = map[string]string{
	"link":    "![pattern-apply-usage](/assets/img/mesheryctl/patternApply.png)",
	"caption": "Usage of mesheryctl pattern apply",
}
...
Example:```
// apply a pattern file
mesheryctl pattern apply -f [file | URL]

// deploy a saved pattern
mesheryctl pattern apply [pattern-name]
```,
Annotations: linkDocPatternApply,
...

NOTE: It is advised not to modify the changes in docs folder, rather should be done in mesheryctl folder as the changes will get overwritten by the script.

Adding New/Removing Existing commands in the reference index

Though the command page is generated automatically by the Cobra CLI library, there are chances where the command does not appear in the reference index page. In such cases, the command details must be manually added to the reference index YAML file. This is generally done by editing the below two files:

• cmds.yml - The YAML file containing the data about the commands
• mesheryctl-commands.md - The markdown page of the command reference documentation

References

• jarcoal/httpmock
• Unit testing CLIs in Go
• How to test CLI commands with Go and Cobra
• Contributing to Meshery API Swagger Documentation

Key principles

The following key principles should be taken to mind when writing tests:

1. Golang’s standard library will be used to write tests.
2. The tests should cover all possible use cases and not just the happy paths.
3. Integration tests should contain the keyword “Integration” in the title and should be marked to be skipped under unit testing. (See below)
4. Fixtures are mock/raw data to use(for e.g. API response to mock an HTTP call).
5. Testdata is the expected response of mesheryctl commands or functions.
6. The mock data and the expected responses are stored in the golden files.
7. Table formatted tests are performed on functions and commands.
8. mesheryctl/pkg/utils/fixtures/validate.version.github.golden file needs to be updated regularly.
9. The version in utils.NewTestHelper() should be updated regularly.
10. Golden files should be updated synchronously as API responses, mesheryctl outputs are updated.

Marking integration tests under unit tests

Since there is no straightforward way to mark unit tests and integration tests differently. Here we use the --short flag while running tests to differentiate between unit tests and integration tests.

func TestPreflightCmdIntegration(t *testing.T) {
   // skipping this integration test with --short flag
   if testing.Short() {
       t.Skip("skipping integration test")
   }
}

In the above code sample, the test is marked with “Integration” in the title and if a --short flag is passed with the command, this test is skipped.

Running tests in GitHub workflows

Unit tests:

go test --short ./... -race -coverprofile=coverage.txt -covermode=atomic

Integration tests:

go test -run Integration ./... -race -coverprofile=coverage.txt -covermode=atomic

To update golden files with the test output use the --update flag:

var update = flag.Bool("update", false, "update golden files")

To grab console logs - fmt.Println() (check mesheryctl/internal/cli/root/perf/view_test.go )

To grab Logrus logs - logrus.Info() (check mesheryctl/internal/cli/root/perf/apply_test.go )

Suggested Reading