Writing custom linter in Go
Buy me a coffeeBuy me a coffee


Writing linters is simple. I was surprised how it’s easy to write a Go linter. Today, we’ll write a linter that will calculate the cyclomatic complexity of the Go code.

What is cyclomatic complexity?

Cyclomatic complexity is a software metric used to indicate the complexity of a program. ref

The idea is simple - every time we find any control flow statements we increase the complexity by one. I know I oversimplified it a bit but I don’t want to overwhelm you with unnecessary details.

There are a few steps we should follow to write our custom linter. Firstly, we can create a test that will check if our linter works or not. Let’s put him into pkg/analyzer/analyzer_test.go file.

package analyzer

import (
	"os"
	"path/filepath"
	"testing"

	"golang.org/x/tools/go/analysis/analysistest"
)

func TestAll(t *testing.T) {
	wd, err := os.Getwd()
	if err != nil {
		t.Fatalf("Failed to get wd: %s", err)
	}

	testdata := filepath.Join(filepath.Dir(filepath.Dir(wd)), "testdata")
	analysistest.Run(t, testdata, NewAnalyzer(), "complexity")
}

The analysistest.Run() function is a helper that simplifies testing linters. What it does is running our linter on package complexity in testate folder. We use NewAnalyzer() function that will return instance of our analyzer. Let’s add it to pkg/analyzer/analyzer.go.

package analyzer

import "golang.org/x/tools/go/analysis"

//nolint:gochecknoglobals
var flagSet flag.FlagSet

func NewAnalyzer() *analysis.Analyzer {
	return &analysis.Analyzer{
		Name:  "cyclop",
		Doc:   "calculates cyclomatic complexity",
		Run:   run,
		Flags: flagSet,
	}
}

func run(pass *analysis.Pass) (interface{}, error) {
  return nil, nil
}

We will use the flag set to input parameters to the linter. In the NewAnalyzer() we define the analyzer with the name cyclop, description, and defined Run function. It accepts analysis.Pass struct that’s described in the docs. We’ll need only a few items from it right now.

To report an issue in the analyzed file we can use pass.Reportf() method. It accepts the position of the diagnostic and the message the user will see. The pass.Files is a slice of *ast.File, that is, a list of files within the package.

We’ll use the last one to iterate every file and check them one by one. When we spot an issue - we’ll report it. To do that, we have to iterate over those files using the loop below.

for _, f := range pass.Files {
		ast.Inspect(f, func(node ast.Node) bool {
		  // your code goes here
		}
}

This is the heart of our linter. The ast.Node is an interface with only two methods that are needed while reporting any issues.

type Node interface {
    Pos() token.Pos // position of first character belonging to the node
    End() token.Pos // position of first character immediately after the node
}

We’re interested only in functions or methods so, we have to cast this type into *ast.FuncDecl.

type FuncDecl struct {
    Doc  *CommentGroup // associated documentation; or nil
    Recv *FieldList    // receiver (methods); or nil (functions)
    Name *Ident        // function/method name
    Type *FuncType     // function signature: parameters, results, and position of "func" keyword
    Body *BlockStmt    // function body; or nil for external (non-Go) function
}

Our loop should look like the one below.

for _, f := range pass.Files {
		ast.Inspect(f, func(node ast.Node) bool {
    f, ok := node.(*ast.FuncDecl)
			if !ok {
				return true
			}
	})
}

It’s time to calculate the cyclomatic complexity. The new complexity function accepts the *ast.File. The algorithm increases the complexity every time it finds if, for, select or case statement as well as || or && operators in if statement.

func complexity(fn *ast.FuncDecl) int {
	v := complexityVisitor{Complexity:1}
	ast.Walk(&v, fn)
	return v.Complexity
}

type complexityVisitor struct {
	Complexity int
}

func (v *complexityVisitor) Visit(n ast.Node) ast.Visitor {
	switch n := n.(type) {
	case *ast.FuncDecl, *ast.IfStmt, *ast.ForStmt, *ast.RangeStmt, *ast.CaseClause, *ast.CommClause:
		v.Complexity++
	case *ast.BinaryExpr:
		if n.Op == token.LAND || n.Op == token.LOR {
			v.Complexity++
		}
	}
	return v
}

There’s ast.Walk(&v, fn) line that can be confusing. This function accepts a visitor that is called on every child node of the parent node we provide. We use a simple type assertion to determinate the type of the specific node.

When we have the code, we can write test cases. To do that, let’s create a new complexity package. Inside of complexity.go file put the code below.

package complexity

import "testing"

func highComplexity() { // want "calculated cyclomatic complexity for function"
	i := 1
	if i > 2 {
		if i > 2 {
		}
		if i > 2 {
		}
		if i > 2 {
		}
		if i > 2 {
		}
	} else {
		if i > 2 {
		}
		if i > 2 {
		}
		if i > 2 {
		}
		if i > 2 {
		}
	}

	if i > 2 {
	}
}

func noComplexity() {}

We created two functions. The noComplexity() function has 1 complexity because it has only one execution path. The linter shouldn’t report any problem. The highComplexity has a bunch of if statements that increase this metric. It should report an issue. Notice that we have a special comment that lets the test suite know about the diagnostic we expect to be reported. You can think about it as the then section in given-when-then approach. Everything that’s inside quotes is a regular expression that must match the diagnostic message. You can read more about it in the official docs.

When we run our tests, we should see success! The next step is to make the linter executable. All we have to do is create a file in cmd/cyclop/cyclop.go file and put the content below.

package main

import (
	"github.com/bkielbasa/cyclop/pkg/analyzer"
	"golang.org/x/tools/go/analysis/singlechecker"
)

func main() {
	singlechecker.Main(analyzer.NewAnalyzer())
}

The code above runs our analyzer as a standalone tool. And that’s it! We have a fully functional Go linter! The full (and a bit more complex) code is available on Github.

There’s one optional step you can make - add your linter to golangci-lint. Here’s an example for the linter we’ve built.

I hope the Go community will get a lot of awesome linters that will save us hours or days.

Tags: #golang #linter

See Also