Update V8 to version 4.6.61.
[chromium-blink-merge.git] / tools / gn / string_utils.cc
blob6d3f87774a04be5d3ff4b6aa46f651e84da92409
1 // Copyright (c) 2013 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
5 #include "tools/gn/string_utils.h"
7 #include "tools/gn/err.h"
8 #include "tools/gn/input_file.h"
9 #include "tools/gn/parser.h"
10 #include "tools/gn/scope.h"
11 #include "tools/gn/token.h"
12 #include "tools/gn/tokenizer.h"
13 #include "tools/gn/value.h"
15 namespace {
17 // Constructs an Err indicating a range inside a string. We assume that the
18 // token has quotes around it that are not counted by the offset.
19 Err ErrInsideStringToken(const Token& token, size_t offset, size_t size,
20 const std::string& msg,
21 const std::string& help = std::string()) {
22 // The "+1" is skipping over the " at the beginning of the token.
23 int int_offset = static_cast<int>(offset);
24 Location begin_loc(token.location().file(),
25 token.location().line_number(),
26 token.location().char_offset() + int_offset + 1,
27 token.location().byte() + int_offset + 1);
28 Location end_loc(
29 token.location().file(),
30 token.location().line_number(),
31 token.location().char_offset() + int_offset + 1 + static_cast<int>(size),
32 token.location().byte() + int_offset + 1 + static_cast<int>(size));
33 return Err(LocationRange(begin_loc, end_loc), msg, help);
36 // Notes about expression interpolation. This is based loosly on Dart but is
37 // slightly less flexible. In Dart, seeing the ${ in a string is something
38 // the toplevel parser knows about, and it will recurse into the block
39 // treating it as a first-class {...} block. So even things like this work:
40 // "hello ${"foo}"*2+"bar"}" => "hello foo}foo}bar"
41 // (you can see it did not get confused by the nested strings or the nested "}"
42 // inside the block).
44 // This is cool but complicates the parser for almost no benefit for this
45 // non-general-purpose programming language. The main reason expressions are
46 // supported here at all are to support "${scope.variable}" and "${list[0]}",
47 // neither of which have any of these edge-cases.
49 // In this simplified approach, we search for the terminating '}' and execute
50 // the result. This means we can't support any expressions with embedded '}'
51 // or '"'. To keep people from getting confusing about what's supported and
52 // what's not, only identifier and accessor expressions are allowed (neither
53 // of these run into any of these edge-cases).
54 bool AppendInterpolatedExpression(Scope* scope,
55 const Token& token,
56 const char* input,
57 size_t begin_offset,
58 size_t end_offset,
59 std::string* output,
60 Err* err) {
61 SourceFile empty_source_file; // Prevent most vexing parse.
62 InputFile input_file(empty_source_file);
63 input_file.SetContents(
64 std::string(&input[begin_offset], end_offset - begin_offset));
66 // Tokenize.
67 std::vector<Token> tokens = Tokenizer::Tokenize(&input_file, err);
68 if (err->has_error()) {
69 // The error will point into our temporary buffer, rewrite it to refer
70 // to the original token. This will make the location information less
71 // precise, but generally there won't be complicated things in string
72 // interpolations.
73 *err = ErrInsideStringToken(token, begin_offset, end_offset - begin_offset,
74 err->message(), err->help_text());
75 return false;
78 // Parse.
79 scoped_ptr<ParseNode> node = Parser::ParseExpression(tokens, err);
80 if (err->has_error()) {
81 // Rewrite error as above.
82 *err = ErrInsideStringToken(token, begin_offset, end_offset - begin_offset,
83 err->message(), err->help_text());
84 return false;
86 if (!(node->AsIdentifier() || node->AsAccessor())) {
87 *err = ErrInsideStringToken(token, begin_offset, end_offset - begin_offset,
88 "Invalid string interpolation.",
89 "The thing inside the ${} must be an identifier ${foo},\n"
90 "a scope access ${foo.bar}, or a list access ${foo[0]}.");
91 return false;
94 // Evaluate.
95 Value result = node->Execute(scope, err);
96 if (err->has_error()) {
97 // Rewrite error as above.
98 *err = ErrInsideStringToken(token, begin_offset, end_offset - begin_offset,
99 err->message(), err->help_text());
100 return false;
103 output->append(result.ToString(false));
104 return true;
107 bool AppendInterpolatedIdentifier(Scope* scope,
108 const Token& token,
109 const char* input,
110 size_t begin_offset,
111 size_t end_offset,
112 std::string* output,
113 Err* err) {
114 base::StringPiece identifier(&input[begin_offset],
115 end_offset - begin_offset);
116 const Value* value = scope->GetValue(identifier, true);
117 if (!value) {
118 // We assume the input points inside the token.
119 *err = ErrInsideStringToken(
120 token, identifier.data() - token.value().data() - 1, identifier.size(),
121 "Undefined identifier in string expansion.",
122 std::string("\"") + identifier + "\" is not currently in scope.");
123 return false;
126 output->append(value->ToString(false));
127 return true;
130 // Handles string interpolations: $identifier and ${expression}
132 // |*i| is the index into |input| of the $. This will be updated to point to
133 // the last character consumed on success. The token is the original string
134 // to blame on failure.
136 // On failure, returns false and sets the error. On success, appends the
137 // result of the interpolation to |*output|.
138 bool AppendStringInterpolation(Scope* scope,
139 const Token& token,
140 const char* input, size_t size,
141 size_t* i,
142 std::string* output,
143 Err* err) {
144 size_t dollars_index = *i;
145 (*i)++;
146 if (*i == size) {
147 *err = ErrInsideStringToken(token, dollars_index, 1, "$ at end of string.",
148 "I was expecting an identifier or {...} after the $.");
149 return false;
152 if (input[*i] == '{') {
153 // Bracketed expression.
154 (*i)++;
155 size_t begin_offset = *i;
157 // Find the closing } and check for non-identifier chars. Don't need to
158 // bother checking for the more-restricted first character of an identifier
159 // since the {} unambiguously denotes the range, and identifiers with
160 // invalid names just won't be found later.
161 bool has_non_ident_chars = false;
162 while (*i < size && input[*i] != '}') {
163 has_non_ident_chars |= Tokenizer::IsIdentifierContinuingChar(input[*i]);
164 (*i)++;
166 if (*i == size) {
167 *err = ErrInsideStringToken(token, dollars_index, *i - dollars_index,
168 "Unterminated ${...");
169 return false;
172 // In the common case, the thing inside the {} will actually be a
173 // simple identifier. Avoid all the complicated parsing of accessors
174 // in this case.
175 if (!has_non_ident_chars) {
176 return AppendInterpolatedIdentifier(scope, token, input, begin_offset,
177 *i, output, err);
179 return AppendInterpolatedExpression(scope, token, input, begin_offset, *i,
180 output, err);
183 // Simple identifier.
184 // The first char of an identifier is more restricted.
185 if (!Tokenizer::IsIdentifierFirstChar(input[*i])) {
186 *err = ErrInsideStringToken(
187 token, dollars_index, *i - dollars_index + 1,
188 "$ not followed by an identifier char.",
189 "It you want a literal $ use \"\\$\".");
190 return false;
192 size_t begin_offset = *i;
193 (*i)++;
195 // Find the first non-identifier char following the string.
196 while (*i < size && Tokenizer::IsIdentifierContinuingChar(input[*i]))
197 (*i)++;
198 size_t end_offset = *i;
199 (*i)--; // Back up to mark the last character consumed.
200 return AppendInterpolatedIdentifier(scope, token, input, begin_offset,
201 end_offset, output, err);
204 } // namespace
206 bool ExpandStringLiteral(Scope* scope,
207 const Token& literal,
208 Value* result,
209 Err* err) {
210 DCHECK(literal.type() == Token::STRING);
211 DCHECK(literal.value().size() > 1); // Should include quotes.
212 DCHECK(result->type() == Value::STRING); // Should be already set.
214 // The token includes the surrounding quotes, so strip those off.
215 const char* input = &literal.value().data()[1];
216 size_t size = literal.value().size() - 2;
218 std::string& output = result->string_value();
219 output.reserve(size);
220 for (size_t i = 0; i < size; i++) {
221 if (input[i] == '\\') {
222 if (i < size - 1) {
223 switch (input[i + 1]) {
224 case '\\':
225 case '"':
226 case '$':
227 output.push_back(input[i + 1]);
228 i++;
229 continue;
230 default: // Everything else has no meaning: pass the literal.
231 break;
234 output.push_back(input[i]);
235 } else if (input[i] == '$') {
236 if (!AppendStringInterpolation(scope, literal, input, size, &i,
237 &output, err))
238 return false;
239 } else {
240 output.push_back(input[i]);
243 return true;
246 std::string RemovePrefix(const std::string& str, const std::string& prefix) {
247 CHECK(str.size() >= prefix.size() &&
248 str.compare(0, prefix.size(), prefix) == 0);
249 return str.substr(prefix.size());
252 void TrimTrailingSlash(std::string* str) {
253 if (!str->empty()) {
254 DCHECK((*str)[str->size() - 1] == '/');
255 str->resize(str->size() - 1);