|
| 1 | +// Copyright 2025 The Go Authors. All rights reserved. |
| 2 | +// Use of this source code is governed by a BSD-style |
| 3 | +// license that can be found in the LICENSE file. |
| 4 | + |
| 5 | +package ast |
| 6 | + |
| 7 | +import ( |
| 8 | + "fmt" |
| 9 | + "go/token" |
| 10 | + "strconv" |
| 11 | + "strings" |
| 12 | + "unicode" |
| 13 | + "unicode/utf8" |
| 14 | +) |
| 15 | + |
| 16 | +// A Directive is a comment of this form: |
| 17 | +// |
| 18 | +// //tool:name args |
| 19 | +// |
| 20 | +// For example, this directive: |
| 21 | +// |
| 22 | +// //go:generate stringer -type Op -trimprefix Op |
| 23 | +// |
| 24 | +// would have Tool "go", Name "generate", and Args "stringer -type Op |
| 25 | +// -trimprefix Op". |
| 26 | +// |
| 27 | +// While Args does not have a strict syntax, by convention it is a |
| 28 | +// space-separated sequence of unquoted words, '"'-quoted Go strings, or |
| 29 | +// '`'-quoted raw strings. |
| 30 | +// |
| 31 | +// See https://go.dev/doc/comment#directives for specification. |
| 32 | +type Directive struct { |
| 33 | + Tool string |
| 34 | + Name string |
| 35 | + Args string // no leading or trailing whitespace |
| 36 | + |
| 37 | + // Slash is the position of the "//" at the beginning of the directive. |
| 38 | + Slash token.Pos |
| 39 | + |
| 40 | + // ArgsPos is the position where Args begins, based on the position passed |
| 41 | + // to ParseDirective. |
| 42 | + ArgsPos token.Pos |
| 43 | +} |
| 44 | + |
| 45 | +// ParseDirective parses a single comment line for a directive comment. |
| 46 | +// |
| 47 | +// If the line is not a directive comment, it returns false. |
| 48 | +// |
| 49 | +// The provided text must be a single line and should include the leading "//". |
| 50 | +// If the text does not start with "//", it returns false. |
| 51 | +// |
| 52 | +// The caller may provide a file position of the start of c. This will be used |
| 53 | +// to track the position of the arguments. This may be [Comment.Slash], |
| 54 | +// synthesized by the caller, or simply 0. If the caller passes 0, then the |
| 55 | +// positions are effectively byte offsets into the string c. |
| 56 | +func ParseDirective(pos token.Pos, c string) (Directive, bool) { |
| 57 | + // Fast path to eliminate most non-directive comments. Must be a line |
| 58 | + // comment starting with [a-z0-9] |
| 59 | + if !(len(c) >= 3 && c[0] == '/' && c[1] == '/' && isalnum(c[2])) { |
| 60 | + return Directive{}, false |
| 61 | + } |
| 62 | + |
| 63 | + buf := directiveScanner{c, pos} |
| 64 | + buf.skip(len("//")) |
| 65 | + |
| 66 | + // Check for a valid directive and parse tool part. |
| 67 | + // |
| 68 | + // This logic matches isDirective. (We could combine them, but isDirective |
| 69 | + // itself is duplicated in several places.) |
| 70 | + colon := strings.Index(buf.str, ":") |
| 71 | + if colon <= 0 || colon+1 >= len(buf.str) { |
| 72 | + return Directive{}, false |
| 73 | + } |
| 74 | + for i := 0; i <= colon+1; i++ { |
| 75 | + if i == colon { |
| 76 | + continue |
| 77 | + } |
| 78 | + if !isalnum(buf.str[i]) { |
| 79 | + return Directive{}, false |
| 80 | + } |
| 81 | + } |
| 82 | + tool := buf.take(colon) |
| 83 | + buf.skip(len(":")) |
| 84 | + |
| 85 | + // Parse name and args. |
| 86 | + name := buf.takeNonSpace() |
| 87 | + buf.skipSpace() |
| 88 | + argsPos := buf.pos |
| 89 | + args := strings.TrimRightFunc(buf.str, unicode.IsSpace) |
| 90 | + |
| 91 | + return Directive{tool, name, args, pos, argsPos}, true |
| 92 | +} |
| 93 | + |
| 94 | +func isalnum(b byte) bool { |
| 95 | + return 'a' <= b && b <= 'z' || '0' <= b && b <= '9' |
| 96 | +} |
| 97 | + |
| 98 | +func (d *Directive) Pos() token.Pos { return d.Slash } |
| 99 | +func (d *Directive) End() token.Pos { return token.Pos(int(d.ArgsPos) + len(d.Args)) } |
| 100 | + |
| 101 | +// A DirectiveArg is an argument to a directive comment. |
| 102 | +type DirectiveArg struct { |
| 103 | + // Arg is the parsed argument string. If the argument was a quoted string, |
| 104 | + // this is its unquoted form. |
| 105 | + Arg string |
| 106 | + // Pos is the position of the first character in this argument. |
| 107 | + Pos token.Pos |
| 108 | +} |
| 109 | + |
| 110 | +// ParseArgs parses a [Directive]'s arguments using the standard convention, |
| 111 | +// which is a sequence of tokens, where each token may be a bare word, or a |
| 112 | +// double quoted Go string, or a back quoted raw Go string. Each token must be |
| 113 | +// separated by one or more Unicode spaces. |
| 114 | +// |
| 115 | +// If the arguments do not conform to this syntax, it returns an error. |
| 116 | +func (d *Directive) ParseArgs() ([]DirectiveArg, error) { |
| 117 | + args := directiveScanner{d.Args, d.ArgsPos} |
| 118 | + |
| 119 | + list := []DirectiveArg{} |
| 120 | + for args.skipSpace(); args.str != ""; args.skipSpace() { |
| 121 | + var arg string |
| 122 | + argPos := args.pos |
| 123 | + |
| 124 | + switch args.str[0] { |
| 125 | + default: |
| 126 | + arg = args.takeNonSpace() |
| 127 | + |
| 128 | + case '`', '"': |
| 129 | + q, err := strconv.QuotedPrefix(args.str) |
| 130 | + if err != nil { // Always strconv.ErrSyntax |
| 131 | + return nil, fmt.Errorf("invalid quoted string in //%s:%s: %s", d.Tool, d.Name, args.str) |
| 132 | + } |
| 133 | + // Any errors will have been returned by QuotedPrefix |
| 134 | + arg, _ = strconv.Unquote(args.take(len(q))) |
| 135 | + |
| 136 | + // Check that the quoted string is followed by a space (or nothing) |
| 137 | + if args.str != "" { |
| 138 | + r, _ := utf8.DecodeRuneInString(args.str) |
| 139 | + if !unicode.IsSpace(r) { |
| 140 | + return nil, fmt.Errorf("invalid quoted string in //%s:%s: %s", d.Tool, d.Name, args.str) |
| 141 | + } |
| 142 | + } |
| 143 | + } |
| 144 | + |
| 145 | + list = append(list, DirectiveArg{arg, argPos}) |
| 146 | + } |
| 147 | + return list, nil |
| 148 | +} |
| 149 | + |
| 150 | +// directiveScanner is a helper for parsing directive comments while maintaining |
| 151 | +// position information. |
| 152 | +type directiveScanner struct { |
| 153 | + str string |
| 154 | + pos token.Pos |
| 155 | +} |
| 156 | + |
| 157 | +func (s *directiveScanner) skip(n int) { |
| 158 | + s.pos += token.Pos(n) |
| 159 | + s.str = s.str[n:] |
| 160 | +} |
| 161 | + |
| 162 | +func (s *directiveScanner) take(n int) string { |
| 163 | + res := s.str[:n] |
| 164 | + s.skip(n) |
| 165 | + return res |
| 166 | +} |
| 167 | + |
| 168 | +func (s *directiveScanner) takeNonSpace() string { |
| 169 | + i := strings.IndexFunc(s.str, unicode.IsSpace) |
| 170 | + if i == -1 { |
| 171 | + i = len(s.str) |
| 172 | + } |
| 173 | + return s.take(i) |
| 174 | +} |
| 175 | + |
| 176 | +func (s *directiveScanner) skipSpace() { |
| 177 | + trim := strings.TrimLeftFunc(s.str, unicode.IsSpace) |
| 178 | + s.skip(len(s.str) - len(trim)) |
| 179 | +} |
0 commit comments