package stylecheck // import "honnef.co/go/tools/stylecheck" import ( "fmt" "go/ast" "go/constant" "go/token" "go/types" "strconv" "strings" "unicode" "unicode/utf8" "honnef.co/go/tools/lint" . "honnef.co/go/tools/lint/lintdsl" "honnef.co/go/tools/ssa" "golang.org/x/tools/go/types/typeutil" ) type Checker struct { CheckGenerated bool } func NewChecker() *Checker { return &Checker{} } func (*Checker) Name() string { return "stylecheck" } func (*Checker) Prefix() string { return "ST" } func (c *Checker) Init(prog *lint.Program) {} func (c *Checker) Checks() []lint.Check { return []lint.Check{ {ID: "ST1000", FilterGenerated: false, Fn: c.CheckPackageComment}, {ID: "ST1001", FilterGenerated: true, Fn: c.CheckDotImports}, // {ID: "ST1002", FilterGenerated: true, Fn: c.CheckBlankImports}, {ID: "ST1003", FilterGenerated: true, Fn: c.CheckNames}, // {ID: "ST1004", FilterGenerated: false, Fn: nil, }, {ID: "ST1005", FilterGenerated: false, Fn: c.CheckErrorStrings}, {ID: "ST1006", FilterGenerated: false, Fn: c.CheckReceiverNames}, // {ID: "ST1007", FilterGenerated: true, Fn: c.CheckIncDec}, {ID: "ST1008", FilterGenerated: false, Fn: c.CheckErrorReturn}, // {ID: "ST1009", FilterGenerated: false, Fn: c.CheckUnexportedReturn}, // {ID: "ST1010", FilterGenerated: false, Fn: c.CheckContextFirstArg}, {ID: "ST1011", FilterGenerated: false, Fn: c.CheckTimeNames}, {ID: "ST1012", FilterGenerated: false, Fn: c.CheckErrorVarNames}, {ID: "ST1013", FilterGenerated: true, Fn: c.CheckHTTPStatusCodes}, {ID: "ST1015", FilterGenerated: true, Fn: c.CheckDefaultCaseOrder}, {ID: "ST1016", FilterGenerated: false, Fn: c.CheckReceiverNamesIdentical}, {ID: "ST1017", FilterGenerated: true, Fn: c.CheckYodaConditions}, } } func (c *Checker) CheckPackageComment(j *lint.Job) { // - At least one file in a non-main package should have a package comment // // - The comment should be of the form // "Package x ...". This has a slight potential for false // positives, as multiple files can have package comments, in // which case they get appended. But that doesn't happen a lot in // the real world. for _, pkg := range j.Program.InitialPackages { if pkg.Name == "main" { continue } hasDocs := false for _, f := range pkg.Syntax { if IsInTest(j, f) { continue } if f.Doc != nil && len(f.Doc.List) > 0 { hasDocs = true prefix := "Package " + f.Name.Name + " " if !strings.HasPrefix(strings.TrimSpace(f.Doc.Text()), prefix) { j.Errorf(f.Doc, `package comment should be of the form "%s..."`, prefix) } f.Doc.Text() } } if !hasDocs { for _, f := range pkg.Syntax { if IsInTest(j, f) { continue } j.Errorf(f, "at least one file in a package should have a package comment") } } } } func (c *Checker) CheckDotImports(j *lint.Job) { for _, pkg := range j.Program.InitialPackages { for _, f := range pkg.Syntax { imports: for _, imp := range f.Imports { path := imp.Path.Value path = path[1 : len(path)-1] for _, w := range pkg.Config.DotImportWhitelist { if w == path { continue imports } } if imp.Name != nil && imp.Name.Name == "." && !IsInTest(j, f) { j.Errorf(imp, "should not use dot imports") } } } } } func (c *Checker) CheckBlankImports(j *lint.Job) { fset := j.Program.Fset() for _, f := range j.Program.Files { if IsInMain(j, f) || IsInTest(j, f) { continue } // Collect imports of the form `import _ "foo"`, i.e. with no // parentheses, as their comment will be associated with the // (paren-free) GenDecl, not the import spec itself. // // We don't directly process the GenDecl so that we can // correctly handle the following: // // import _ "foo" // import _ "bar" // // where only the first import should get flagged. skip := map[ast.Spec]bool{} ast.Inspect(f, func(node ast.Node) bool { switch node := node.(type) { case *ast.File: return true case *ast.GenDecl: if node.Tok != token.IMPORT { return false } if node.Lparen == token.NoPos && node.Doc != nil { skip[node.Specs[0]] = true } return false } return false }) for i, imp := range f.Imports { pos := fset.Position(imp.Pos()) if !IsBlank(imp.Name) { continue } // Only flag the first blank import in a group of imports, // or don't flag any of them, if the first one is // commented if i > 0 { prev := f.Imports[i-1] prevPos := fset.Position(prev.Pos()) if pos.Line-1 == prevPos.Line && IsBlank(prev.Name) { continue } } if imp.Doc == nil && imp.Comment == nil && !skip[imp] { j.Errorf(imp, "a blank import should be only in a main or test package, or have a comment justifying it") } } } } func (c *Checker) CheckIncDec(j *lint.Job) { // TODO(dh): this can be noisy for function bodies that look like this: // x += 3 // ... // x += 2 // ... // x += 1 fn := func(node ast.Node) bool { assign, ok := node.(*ast.AssignStmt) if !ok || (assign.Tok != token.ADD_ASSIGN && assign.Tok != token.SUB_ASSIGN) { return true } if (len(assign.Lhs) != 1 || len(assign.Rhs) != 1) || !IsIntLiteral(assign.Rhs[0], "1") { return true } suffix := "" switch assign.Tok { case token.ADD_ASSIGN: suffix = "++" case token.SUB_ASSIGN: suffix = "--" } j.Errorf(assign, "should replace %s with %s%s", Render(j, assign), Render(j, assign.Lhs[0]), suffix) return true } for _, f := range j.Program.Files { ast.Inspect(f, fn) } } func (c *Checker) CheckErrorReturn(j *lint.Job) { fnLoop: for _, fn := range j.Program.InitialFunctions { sig := fn.Type().(*types.Signature) rets := sig.Results() if rets == nil || rets.Len() < 2 { continue } if rets.At(rets.Len()-1).Type() == types.Universe.Lookup("error").Type() { // Last return type is error. If the function also returns // errors in other positions, that's fine. continue } for i := rets.Len() - 2; i >= 0; i-- { if rets.At(i).Type() == types.Universe.Lookup("error").Type() { j.Errorf(rets.At(i), "error should be returned as the last argument") continue fnLoop } } } } // CheckUnexportedReturn checks that exported functions on exported // types do not return unexported types. func (c *Checker) CheckUnexportedReturn(j *lint.Job) { for _, fn := range j.Program.InitialFunctions { if fn.Synthetic != "" || fn.Parent() != nil { continue } if !ast.IsExported(fn.Name()) || IsInMain(j, fn) || IsInTest(j, fn) { continue } sig := fn.Type().(*types.Signature) if sig.Recv() != nil && !ast.IsExported(Dereference(sig.Recv().Type()).(*types.Named).Obj().Name()) { continue } res := sig.Results() for i := 0; i < res.Len(); i++ { if named, ok := DereferenceR(res.At(i).Type()).(*types.Named); ok && !ast.IsExported(named.Obj().Name()) && named != types.Universe.Lookup("error").Type() { j.Errorf(fn, "should not return unexported type") } } } } func (c *Checker) CheckReceiverNames(j *lint.Job) { for _, pkg := range j.Program.InitialPackages { for _, m := range pkg.SSA.Members { if T, ok := m.Object().(*types.TypeName); ok && !T.IsAlias() { ms := typeutil.IntuitiveMethodSet(T.Type(), nil) for _, sel := range ms { fn := sel.Obj().(*types.Func) recv := fn.Type().(*types.Signature).Recv() if Dereference(recv.Type()) != T.Type() { // skip embedded methods continue } if recv.Name() == "self" || recv.Name() == "this" { j.Errorf(recv, `receiver name should be a reflection of its identity; don't use generic names such as "this" or "self"`) } if recv.Name() == "_" { j.Errorf(recv, "receiver name should not be an underscore, omit the name if it is unused") } } } } } } func (c *Checker) CheckReceiverNamesIdentical(j *lint.Job) { for _, pkg := range j.Program.InitialPackages { for _, m := range pkg.SSA.Members { names := map[string]int{} var firstFn *types.Func if T, ok := m.Object().(*types.TypeName); ok && !T.IsAlias() { ms := typeutil.IntuitiveMethodSet(T.Type(), nil) for _, sel := range ms { fn := sel.Obj().(*types.Func) recv := fn.Type().(*types.Signature).Recv() if Dereference(recv.Type()) != T.Type() { // skip embedded methods continue } if firstFn == nil { firstFn = fn } if recv.Name() != "" && recv.Name() != "_" { names[recv.Name()]++ } } } if len(names) > 1 { var seen []string for name, count := range names { seen = append(seen, fmt.Sprintf("%dx %q", count, name)) } j.Errorf(firstFn, "methods on the same type should have the same receiver name (seen %s)", strings.Join(seen, ", ")) } } } } func (c *Checker) CheckContextFirstArg(j *lint.Job) { // TODO(dh): this check doesn't apply to test helpers. Example from the stdlib: // func helperCommandContext(t *testing.T, ctx context.Context, s ...string) (cmd *exec.Cmd) { fnLoop: for _, fn := range j.Program.InitialFunctions { if fn.Synthetic != "" || fn.Parent() != nil { continue } params := fn.Signature.Params() if params.Len() < 2 { continue } if types.TypeString(params.At(0).Type(), nil) == "context.Context" { continue } for i := 1; i < params.Len(); i++ { param := params.At(i) if types.TypeString(param.Type(), nil) == "context.Context" { j.Errorf(param, "context.Context should be the first argument of a function") continue fnLoop } } } } func (c *Checker) CheckErrorStrings(j *lint.Job) { fnNames := map[*ssa.Package]map[string]bool{} for _, fn := range j.Program.InitialFunctions { m := fnNames[fn.Package()] if m == nil { m = map[string]bool{} fnNames[fn.Package()] = m } m[fn.Name()] = true } for _, fn := range j.Program.InitialFunctions { if IsInTest(j, fn) { // We don't care about malformed error messages in tests; // they're usually for direct human consumption, not part // of an API continue } for _, block := range fn.Blocks { instrLoop: for _, ins := range block.Instrs { call, ok := ins.(*ssa.Call) if !ok { continue } if !IsCallTo(call.Common(), "errors.New") && !IsCallTo(call.Common(), "fmt.Errorf") { continue } k, ok := call.Common().Args[0].(*ssa.Const) if !ok { continue } s := constant.StringVal(k.Value) if len(s) == 0 { continue } switch s[len(s)-1] { case '.', ':', '!', '\n': j.Errorf(call, "error strings should not end with punctuation or a newline") } idx := strings.IndexByte(s, ' ') if idx == -1 { // single word error message, probably not a real // error but something used in tests or during // debugging continue } word := s[:idx] first, n := utf8.DecodeRuneInString(word) if !unicode.IsUpper(first) { continue } for _, c := range word[n:] { if unicode.IsUpper(c) { // Word is probably an initialism or // multi-word function name continue instrLoop } } word = strings.TrimRightFunc(word, func(r rune) bool { return unicode.IsPunct(r) }) if fnNames[fn.Package()][word] { // Word is probably the name of a function in this package continue } // First word in error starts with a capital // letter, and the word doesn't contain any other // capitals, making it unlikely to be an // initialism or multi-word function name. // // It could still be a proper noun, though. j.Errorf(call, "error strings should not be capitalized") } } } } func (c *Checker) CheckTimeNames(j *lint.Job) { suffixes := []string{ "Sec", "Secs", "Seconds", "Msec", "Msecs", "Milli", "Millis", "Milliseconds", "Usec", "Usecs", "Microseconds", "MS", "Ms", } fn := func(T types.Type, names []*ast.Ident) { if !IsType(T, "time.Duration") && !IsType(T, "*time.Duration") { return } for _, name := range names { for _, suffix := range suffixes { if strings.HasSuffix(name.Name, suffix) { j.Errorf(name, "var %s is of type %v; don't use unit-specific suffix %q", name.Name, T, suffix) break } } } } for _, f := range j.Program.Files { ast.Inspect(f, func(node ast.Node) bool { switch node := node.(type) { case *ast.ValueSpec: T := TypeOf(j, node.Type) fn(T, node.Names) case *ast.FieldList: for _, field := range node.List { T := TypeOf(j, field.Type) fn(T, field.Names) } } return true }) } } func (c *Checker) CheckErrorVarNames(j *lint.Job) { for _, f := range j.Program.Files { for _, decl := range f.Decls { gen, ok := decl.(*ast.GenDecl) if !ok || gen.Tok != token.VAR { continue } for _, spec := range gen.Specs { spec := spec.(*ast.ValueSpec) if len(spec.Names) != len(spec.Values) { continue } for i, name := range spec.Names { val := spec.Values[i] if !IsCallToAST(j, val, "errors.New") && !IsCallToAST(j, val, "fmt.Errorf") { continue } prefix := "err" if name.IsExported() { prefix = "Err" } if !strings.HasPrefix(name.Name, prefix) { j.Errorf(name, "error var %s should have name of the form %sFoo", name.Name, prefix) } } } } } } var httpStatusCodes = map[int]string{ 100: "StatusContinue", 101: "StatusSwitchingProtocols", 102: "StatusProcessing", 200: "StatusOK", 201: "StatusCreated", 202: "StatusAccepted", 203: "StatusNonAuthoritativeInfo", 204: "StatusNoContent", 205: "StatusResetContent", 206: "StatusPartialContent", 207: "StatusMultiStatus", 208: "StatusAlreadyReported", 226: "StatusIMUsed", 300: "StatusMultipleChoices", 301: "StatusMovedPermanently", 302: "StatusFound", 303: "StatusSeeOther", 304: "StatusNotModified", 305: "StatusUseProxy", 307: "StatusTemporaryRedirect", 308: "StatusPermanentRedirect", 400: "StatusBadRequest", 401: "StatusUnauthorized", 402: "StatusPaymentRequired", 403: "StatusForbidden", 404: "StatusNotFound", 405: "StatusMethodNotAllowed", 406: "StatusNotAcceptable", 407: "StatusProxyAuthRequired", 408: "StatusRequestTimeout", 409: "StatusConflict", 410: "StatusGone", 411: "StatusLengthRequired", 412: "StatusPreconditionFailed", 413: "StatusRequestEntityTooLarge", 414: "StatusRequestURITooLong", 415: "StatusUnsupportedMediaType", 416: "StatusRequestedRangeNotSatisfiable", 417: "StatusExpectationFailed", 418: "StatusTeapot", 422: "StatusUnprocessableEntity", 423: "StatusLocked", 424: "StatusFailedDependency", 426: "StatusUpgradeRequired", 428: "StatusPreconditionRequired", 429: "StatusTooManyRequests", 431: "StatusRequestHeaderFieldsTooLarge", 451: "StatusUnavailableForLegalReasons", 500: "StatusInternalServerError", 501: "StatusNotImplemented", 502: "StatusBadGateway", 503: "StatusServiceUnavailable", 504: "StatusGatewayTimeout", 505: "StatusHTTPVersionNotSupported", 506: "StatusVariantAlsoNegotiates", 507: "StatusInsufficientStorage", 508: "StatusLoopDetected", 510: "StatusNotExtended", 511: "StatusNetworkAuthenticationRequired", } func (c *Checker) CheckHTTPStatusCodes(j *lint.Job) { for _, pkg := range j.Program.InitialPackages { whitelist := map[string]bool{} for _, code := range pkg.Config.HTTPStatusCodeWhitelist { whitelist[code] = true } fn := func(node ast.Node) bool { call, ok := node.(*ast.CallExpr) if !ok { return true } var arg int switch CallNameAST(j, call) { case "net/http.Error": arg = 2 case "net/http.Redirect": arg = 3 case "net/http.StatusText": arg = 0 case "net/http.RedirectHandler": arg = 1 default: return true } lit, ok := call.Args[arg].(*ast.BasicLit) if !ok { return true } if whitelist[lit.Value] { return true } n, err := strconv.Atoi(lit.Value) if err != nil { return true } s, ok := httpStatusCodes[n] if !ok { return true } j.Errorf(lit, "should use constant http.%s instead of numeric literal %d", s, n) return true } for _, f := range pkg.Syntax { ast.Inspect(f, fn) } } } func (c *Checker) CheckDefaultCaseOrder(j *lint.Job) { fn := func(node ast.Node) bool { stmt, ok := node.(*ast.SwitchStmt) if !ok { return true } list := stmt.Body.List for i, c := range list { if c.(*ast.CaseClause).List == nil && i != 0 && i != len(list)-1 { j.Errorf(c, "default case should be first or last in switch statement") break } } return true } for _, f := range j.Program.Files { ast.Inspect(f, fn) } } func (c *Checker) CheckYodaConditions(j *lint.Job) { fn := func(node ast.Node) bool { cond, ok := node.(*ast.BinaryExpr) if !ok { return true } if cond.Op != token.EQL && cond.Op != token.NEQ { return true } if _, ok := cond.X.(*ast.BasicLit); !ok { return true } if _, ok := cond.Y.(*ast.BasicLit); ok { // Don't flag lit == lit conditions, just in case return true } j.Errorf(cond, "don't use Yoda conditions") return true } for _, f := range j.Program.Files { ast.Inspect(f, fn) } }