1 | #include "absl/strings/internal/str_format/parser.h" |
2 | |
3 | #include <assert.h> |
4 | #include <string.h> |
5 | #include <wchar.h> |
6 | #include <cctype> |
7 | #include <cstdint> |
8 | |
9 | #include <algorithm> |
10 | #include <initializer_list> |
11 | #include <limits> |
12 | #include <ostream> |
13 | #include <string> |
14 | #include <unordered_set> |
15 | |
16 | namespace absl { |
17 | namespace str_format_internal { |
18 | |
19 | using CC = ConversionChar::Id; |
20 | using LM = LengthMod::Id; |
21 | ABSL_CONST_INIT const ConvTag kTags[256] = { |
22 | {}, {}, {}, {}, {}, {}, {}, {}, // 00-07 |
23 | {}, {}, {}, {}, {}, {}, {}, {}, // 08-0f |
24 | {}, {}, {}, {}, {}, {}, {}, {}, // 10-17 |
25 | {}, {}, {}, {}, {}, {}, {}, {}, // 18-1f |
26 | {}, {}, {}, {}, {}, {}, {}, {}, // 20-27 |
27 | {}, {}, {}, {}, {}, {}, {}, {}, // 28-2f |
28 | {}, {}, {}, {}, {}, {}, {}, {}, // 30-37 |
29 | {}, {}, {}, {}, {}, {}, {}, {}, // 38-3f |
30 | {}, CC::A, {}, CC::C, {}, CC::E, CC::F, CC::G, // @ABCDEFG |
31 | {}, {}, {}, {}, LM::L, {}, {}, {}, // HIJKLMNO |
32 | {}, {}, {}, CC::S, {}, {}, {}, {}, // PQRSTUVW |
33 | CC::X, {}, {}, {}, {}, {}, {}, {}, // XYZ[\]^_ |
34 | {}, CC::a, {}, CC::c, CC::d, CC::e, CC::f, CC::g, // `abcdefg |
35 | LM::h, CC::i, LM::j, {}, LM::l, {}, CC::n, CC::o, // hijklmno |
36 | CC::p, LM::q, {}, CC::s, LM::t, CC::u, {}, {}, // pqrstuvw |
37 | CC::x, {}, LM::z, {}, {}, {}, {}, {}, // xyz{|}! |
38 | {}, {}, {}, {}, {}, {}, {}, {}, // 80-87 |
39 | {}, {}, {}, {}, {}, {}, {}, {}, // 88-8f |
40 | {}, {}, {}, {}, {}, {}, {}, {}, // 90-97 |
41 | {}, {}, {}, {}, {}, {}, {}, {}, // 98-9f |
42 | {}, {}, {}, {}, {}, {}, {}, {}, // a0-a7 |
43 | {}, {}, {}, {}, {}, {}, {}, {}, // a8-af |
44 | {}, {}, {}, {}, {}, {}, {}, {}, // b0-b7 |
45 | {}, {}, {}, {}, {}, {}, {}, {}, // b8-bf |
46 | {}, {}, {}, {}, {}, {}, {}, {}, // c0-c7 |
47 | {}, {}, {}, {}, {}, {}, {}, {}, // c8-cf |
48 | {}, {}, {}, {}, {}, {}, {}, {}, // d0-d7 |
49 | {}, {}, {}, {}, {}, {}, {}, {}, // d8-df |
50 | {}, {}, {}, {}, {}, {}, {}, {}, // e0-e7 |
51 | {}, {}, {}, {}, {}, {}, {}, {}, // e8-ef |
52 | {}, {}, {}, {}, {}, {}, {}, {}, // f0-f7 |
53 | {}, {}, {}, {}, {}, {}, {}, {}, // f8-ff |
54 | }; |
55 | |
56 | namespace { |
57 | |
58 | bool CheckFastPathSetting(const UnboundConversion& conv) { |
59 | bool should_be_basic = !conv.flags.left && // |
60 | !conv.flags.show_pos && // |
61 | !conv.flags.sign_col && // |
62 | !conv.flags.alt && // |
63 | !conv.flags.zero && // |
64 | (conv.width.value() == -1) && |
65 | (conv.precision.value() == -1); |
66 | if (should_be_basic != conv.flags.basic) { |
67 | fprintf(stderr, |
68 | "basic=%d left=%d show_pos=%d sign_col=%d alt=%d zero=%d " |
69 | "width=%d precision=%d\n" , |
70 | conv.flags.basic, conv.flags.left, conv.flags.show_pos, |
71 | conv.flags.sign_col, conv.flags.alt, conv.flags.zero, |
72 | conv.width.value(), conv.precision.value()); |
73 | } |
74 | return should_be_basic == conv.flags.basic; |
75 | } |
76 | |
77 | template <bool is_positional> |
78 | const char *ConsumeConversion(const char *pos, const char *const end, |
79 | UnboundConversion *conv, int *next_arg) { |
80 | const char* const original_pos = pos; |
81 | char c; |
82 | // Read the next char into `c` and update `pos`. Returns false if there are |
83 | // no more chars to read. |
84 | #define ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR() \ |
85 | do { \ |
86 | if (ABSL_PREDICT_FALSE(pos == end)) return nullptr; \ |
87 | c = *pos++; \ |
88 | } while (0) |
89 | |
90 | const auto parse_digits = [&] { |
91 | int digits = c - '0'; |
92 | // We do not want to overflow `digits` so we consume at most digits10 |
93 | // digits. If there are more digits the parsing will fail later on when the |
94 | // digit doesn't match the expected characters. |
95 | int num_digits = std::numeric_limits<int>::digits10; |
96 | for (;;) { |
97 | if (ABSL_PREDICT_FALSE(pos == end)) break; |
98 | c = *pos++; |
99 | if (!std::isdigit(c)) break; |
100 | --num_digits; |
101 | if (ABSL_PREDICT_FALSE(!num_digits)) break; |
102 | digits = 10 * digits + c - '0'; |
103 | } |
104 | return digits; |
105 | }; |
106 | |
107 | if (is_positional) { |
108 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
109 | if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; |
110 | conv->arg_position = parse_digits(); |
111 | assert(conv->arg_position > 0); |
112 | if (ABSL_PREDICT_FALSE(c != '$')) return nullptr; |
113 | } |
114 | |
115 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
116 | |
117 | // We should start with the basic flag on. |
118 | assert(conv->flags.basic); |
119 | |
120 | // Any non alpha character makes this conversion not basic. |
121 | // This includes flags (-+ #0), width (1-9, *) or precision (.). |
122 | // All conversion characters and length modifiers are alpha characters. |
123 | if (c < 'A') { |
124 | conv->flags.basic = false; |
125 | |
126 | for (; c <= '0';) { |
127 | // FIXME: We might be able to speed this up reusing the lookup table from |
128 | // above. It might require changing Flags to be a plain integer where we |
129 | // can |= a value. |
130 | switch (c) { |
131 | case '-': |
132 | conv->flags.left = true; |
133 | break; |
134 | case '+': |
135 | conv->flags.show_pos = true; |
136 | break; |
137 | case ' ': |
138 | conv->flags.sign_col = true; |
139 | break; |
140 | case '#': |
141 | conv->flags.alt = true; |
142 | break; |
143 | case '0': |
144 | conv->flags.zero = true; |
145 | break; |
146 | default: |
147 | goto flags_done; |
148 | } |
149 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
150 | } |
151 | flags_done: |
152 | |
153 | if (c <= '9') { |
154 | if (c >= '0') { |
155 | int maybe_width = parse_digits(); |
156 | if (!is_positional && c == '$') { |
157 | if (ABSL_PREDICT_FALSE(*next_arg != 0)) return nullptr; |
158 | // Positional conversion. |
159 | *next_arg = -1; |
160 | conv->flags = Flags(); |
161 | conv->flags.basic = true; |
162 | return ConsumeConversion<true>(original_pos, end, conv, next_arg); |
163 | } |
164 | conv->width.set_value(maybe_width); |
165 | } else if (c == '*') { |
166 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
167 | if (is_positional) { |
168 | if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; |
169 | conv->width.set_from_arg(parse_digits()); |
170 | if (ABSL_PREDICT_FALSE(c != '$')) return nullptr; |
171 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
172 | } else { |
173 | conv->width.set_from_arg(++*next_arg); |
174 | } |
175 | } |
176 | } |
177 | |
178 | if (c == '.') { |
179 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
180 | if (std::isdigit(c)) { |
181 | conv->precision.set_value(parse_digits()); |
182 | } else if (c == '*') { |
183 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
184 | if (is_positional) { |
185 | if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; |
186 | conv->precision.set_from_arg(parse_digits()); |
187 | if (c != '$') return nullptr; |
188 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
189 | } else { |
190 | conv->precision.set_from_arg(++*next_arg); |
191 | } |
192 | } else { |
193 | conv->precision.set_value(0); |
194 | } |
195 | } |
196 | } |
197 | |
198 | auto tag = GetTagForChar(c); |
199 | |
200 | if (ABSL_PREDICT_FALSE(!tag.is_conv())) { |
201 | if (ABSL_PREDICT_FALSE(!tag.is_length())) return nullptr; |
202 | |
203 | // It is a length modifier. |
204 | using str_format_internal::LengthMod; |
205 | LengthMod length_mod = tag.as_length(); |
206 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
207 | if (c == 'h' && length_mod.id() == LengthMod::h) { |
208 | conv->length_mod = LengthMod::FromId(LengthMod::hh); |
209 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
210 | } else if (c == 'l' && length_mod.id() == LengthMod::l) { |
211 | conv->length_mod = LengthMod::FromId(LengthMod::ll); |
212 | ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); |
213 | } else { |
214 | conv->length_mod = length_mod; |
215 | } |
216 | tag = GetTagForChar(c); |
217 | if (ABSL_PREDICT_FALSE(!tag.is_conv())) return nullptr; |
218 | } |
219 | |
220 | assert(CheckFastPathSetting(*conv)); |
221 | (void)(&CheckFastPathSetting); |
222 | |
223 | conv->conv = tag.as_conv(); |
224 | if (!is_positional) conv->arg_position = ++*next_arg; |
225 | return pos; |
226 | } |
227 | |
228 | } // namespace |
229 | |
230 | const char *ConsumeUnboundConversion(const char *p, const char *end, |
231 | UnboundConversion *conv, int *next_arg) { |
232 | if (*next_arg < 0) return ConsumeConversion<true>(p, end, conv, next_arg); |
233 | return ConsumeConversion<false>(p, end, conv, next_arg); |
234 | } |
235 | |
236 | struct ParsedFormatBase::ParsedFormatConsumer { |
237 | explicit ParsedFormatConsumer(ParsedFormatBase *parsedformat) |
238 | : parsed(parsedformat), data_pos(parsedformat->data_.get()) {} |
239 | |
240 | bool Append(string_view s) { |
241 | if (s.empty()) return true; |
242 | |
243 | size_t text_end = AppendText(s); |
244 | |
245 | if (!parsed->items_.empty() && !parsed->items_.back().is_conversion) { |
246 | // Let's extend the existing text run. |
247 | parsed->items_.back().text_end = text_end; |
248 | } else { |
249 | // Let's make a new text run. |
250 | parsed->items_.push_back({false, text_end, {}}); |
251 | } |
252 | return true; |
253 | } |
254 | |
255 | bool ConvertOne(const UnboundConversion &conv, string_view s) { |
256 | size_t text_end = AppendText(s); |
257 | parsed->items_.push_back({true, text_end, conv}); |
258 | return true; |
259 | } |
260 | |
261 | size_t AppendText(string_view s) { |
262 | memcpy(data_pos, s.data(), s.size()); |
263 | data_pos += s.size(); |
264 | return static_cast<size_t>(data_pos - parsed->data_.get()); |
265 | } |
266 | |
267 | ParsedFormatBase *parsed; |
268 | char* data_pos; |
269 | }; |
270 | |
271 | ParsedFormatBase::ParsedFormatBase(string_view format, bool allow_ignored, |
272 | std::initializer_list<Conv> convs) |
273 | : data_(format.empty() ? nullptr : new char[format.size()]) { |
274 | has_error_ = !ParseFormatString(format, ParsedFormatConsumer(this)) || |
275 | !MatchesConversions(allow_ignored, convs); |
276 | } |
277 | |
278 | bool ParsedFormatBase::MatchesConversions( |
279 | bool allow_ignored, std::initializer_list<Conv> convs) const { |
280 | std::unordered_set<int> used; |
281 | auto add_if_valid_conv = [&](int pos, char c) { |
282 | if (static_cast<size_t>(pos) > convs.size() || |
283 | !Contains(convs.begin()[pos - 1], c)) |
284 | return false; |
285 | used.insert(pos); |
286 | return true; |
287 | }; |
288 | for (const ConversionItem &item : items_) { |
289 | if (!item.is_conversion) continue; |
290 | auto &conv = item.conv; |
291 | if (conv.precision.is_from_arg() && |
292 | !add_if_valid_conv(conv.precision.get_from_arg(), '*')) |
293 | return false; |
294 | if (conv.width.is_from_arg() && |
295 | !add_if_valid_conv(conv.width.get_from_arg(), '*')) |
296 | return false; |
297 | if (!add_if_valid_conv(conv.arg_position, conv.conv.Char())) return false; |
298 | } |
299 | return used.size() == convs.size() || allow_ignored; |
300 | } |
301 | |
302 | } // namespace str_format_internal |
303 | } // namespace absl |
304 | |