| 1 | //===-- lib/runtime/edit-input.cpp ------------------------------*- C++ -*-===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "edit-input.h" |
| 10 | #include "flang-rt/runtime/namelist.h" |
| 11 | #include "flang-rt/runtime/utf.h" |
| 12 | #include "flang/Common/optional.h" |
| 13 | #include "flang/Common/real.h" |
| 14 | #include "flang/Common/uint128.h" |
| 15 | #include "flang/Runtime/freestanding-tools.h" |
| 16 | #include <algorithm> |
| 17 | #include <cfenv> |
| 18 | |
| 19 | namespace Fortran::runtime::io { |
| 20 | RT_OFFLOAD_API_GROUP_BEGIN |
| 21 | |
| 22 | // Checks that a list-directed input value has been entirely consumed and |
| 23 | // doesn't contain unparsed characters before the next value separator. |
| 24 | static inline RT_API_ATTRS bool IsCharValueSeparator( |
| 25 | const DataEdit &edit, char32_t ch) { |
| 26 | char32_t comma{ |
| 27 | edit.modes.editingFlags & decimalComma ? char32_t{';'} : char32_t{','}}; |
| 28 | return ch == ' ' || ch == '\t' || ch == comma || ch == '/' || |
| 29 | (edit.IsNamelist() && (ch == '&' || ch == '$')); |
| 30 | } |
| 31 | |
| 32 | static RT_API_ATTRS bool CheckCompleteListDirectedField( |
| 33 | IoStatementState &io, const DataEdit &edit) { |
| 34 | if (edit.IsListDirected()) { |
| 35 | std::size_t byteCount; |
| 36 | if (auto ch{io.GetCurrentChar(byteCount)}) { |
| 37 | if (IsCharValueSeparator(edit, *ch)) { |
| 38 | return true; |
| 39 | } else { |
| 40 | const auto &connection{io.GetConnectionState()}; |
| 41 | io.GetIoErrorHandler().SignalError(IostatBadListDirectedInputSeparator, |
| 42 | "invalid character (0x%x) after list-directed input value, " |
| 43 | "at column %d in record %d" , |
| 44 | static_cast<unsigned>(*ch), |
| 45 | static_cast<int>(connection.positionInRecord + 1), |
| 46 | static_cast<int>(connection.currentRecordNumber)); |
| 47 | return false; |
| 48 | } |
| 49 | } else { |
| 50 | return true; // end of record: ok |
| 51 | } |
| 52 | } else { |
| 53 | return true; |
| 54 | } |
| 55 | } |
| 56 | |
| 57 | static inline RT_API_ATTRS char32_t GetSeparatorChar(const DataEdit &edit) { |
| 58 | return edit.modes.editingFlags & decimalComma ? char32_t{';'} : char32_t{','}; |
| 59 | } |
| 60 | |
| 61 | template <int LOG2_BASE> |
| 62 | static RT_API_ATTRS bool EditBOZInput( |
| 63 | IoStatementState &io, const DataEdit &edit, void *n, std::size_t bytes) { |
| 64 | // Skip leading white space & zeroes |
| 65 | Fortran::common::optional<int> remaining{io.CueUpInput(edit)}; |
| 66 | auto start{io.GetConnectionState().positionInRecord}; |
| 67 | Fortran::common::optional<char32_t> next{io.NextInField(remaining, edit)}; |
| 68 | if (next.value_or('?') == '0') { |
| 69 | do { |
| 70 | start = io.GetConnectionState().positionInRecord; |
| 71 | next = io.NextInField(remaining, edit); |
| 72 | } while (next && *next == '0'); |
| 73 | } |
| 74 | // Count significant digits after any leading white space & zeroes |
| 75 | int digits{0}; |
| 76 | int significantBits{0}; |
| 77 | const char32_t comma{GetSeparatorChar(edit)}; |
| 78 | for (; next; next = io.NextInField(remaining, edit)) { |
| 79 | char32_t ch{*next}; |
| 80 | if (ch == ' ' || ch == '\t') { |
| 81 | if (edit.modes.editingFlags & blankZero) { |
| 82 | ch = '0'; // BZ mode - treat blank as if it were zero |
| 83 | } else { |
| 84 | continue; |
| 85 | } |
| 86 | } |
| 87 | if (ch >= '0' && ch <= '1') { |
| 88 | } else if (LOG2_BASE >= 3 && ch >= '2' && ch <= '7') { |
| 89 | } else if (LOG2_BASE >= 4 && ch >= '8' && ch <= '9') { |
| 90 | } else if (LOG2_BASE >= 4 && ch >= 'A' && ch <= 'F') { |
| 91 | } else if (LOG2_BASE >= 4 && ch >= 'a' && ch <= 'f') { |
| 92 | } else if (ch == comma) { |
| 93 | break; // end non-list-directed field early |
| 94 | } else { |
| 95 | io.GetIoErrorHandler().SignalError( |
| 96 | "Bad character '%lc' in B/O/Z input field" , ch); |
| 97 | return false; |
| 98 | } |
| 99 | if (digits++ == 0) { |
| 100 | if (ch >= '0' && ch <= '1') { |
| 101 | significantBits = 1; |
| 102 | } else if (ch >= '2' && ch <= '3') { |
| 103 | significantBits = 2; |
| 104 | } else if (ch >= '4' && ch <= '7') { |
| 105 | significantBits = 3; |
| 106 | } else { |
| 107 | significantBits = 4; |
| 108 | } |
| 109 | } else { |
| 110 | significantBits += LOG2_BASE; |
| 111 | } |
| 112 | } |
| 113 | auto significantBytes{static_cast<std::size_t>(significantBits + 7) / 8}; |
| 114 | if (significantBytes > bytes) { |
| 115 | io.GetIoErrorHandler().SignalError(IostatBOZInputOverflow, |
| 116 | "B/O/Z input of %d digits overflows %zd-byte variable" , digits, bytes); |
| 117 | return false; |
| 118 | } |
| 119 | // Reset to start of significant digits |
| 120 | io.HandleAbsolutePosition(start); |
| 121 | remaining.reset(); |
| 122 | // Make a second pass now that the digit count is known |
| 123 | std::memset(n, 0, bytes); |
| 124 | int increment{isHostLittleEndian ? -1 : 1}; |
| 125 | auto *data{reinterpret_cast<unsigned char *>(n) + |
| 126 | (isHostLittleEndian ? significantBytes - 1 : bytes - significantBytes)}; |
| 127 | int bitsAfterFirstDigit{(digits - 1) * LOG2_BASE}; |
| 128 | int shift{bitsAfterFirstDigit & 7}; |
| 129 | if (shift + (significantBits - bitsAfterFirstDigit) > 8) { |
| 130 | shift = shift - 8; // misaligned octal |
| 131 | } |
| 132 | while (digits > 0) { |
| 133 | char32_t ch{io.NextInField(remaining, edit).value_or(' ')}; |
| 134 | int digit{0}; |
| 135 | if (ch == ' ' || ch == '\t') { |
| 136 | if (edit.modes.editingFlags & blankZero) { |
| 137 | ch = '0'; // BZ mode - treat blank as if it were zero |
| 138 | } else { |
| 139 | continue; |
| 140 | } |
| 141 | } |
| 142 | --digits; |
| 143 | if (ch >= '0' && ch <= '9') { |
| 144 | digit = ch - '0'; |
| 145 | } else if (ch >= 'A' && ch <= 'F') { |
| 146 | digit = ch + 10 - 'A'; |
| 147 | } else if (ch >= 'a' && ch <= 'f') { |
| 148 | digit = ch + 10 - 'a'; |
| 149 | } else { |
| 150 | continue; |
| 151 | } |
| 152 | if (shift < 0) { |
| 153 | if (shift + LOG2_BASE > 0) { // misaligned octal |
| 154 | *data |= digit >> -shift; |
| 155 | } |
| 156 | shift += 8; |
| 157 | data += increment; |
| 158 | } |
| 159 | *data |= digit << shift; |
| 160 | shift -= LOG2_BASE; |
| 161 | } |
| 162 | return CheckCompleteListDirectedField(io, edit); |
| 163 | } |
| 164 | |
| 165 | static inline RT_API_ATTRS char32_t GetRadixPointChar(const DataEdit &edit) { |
| 166 | return edit.modes.editingFlags & decimalComma ? char32_t{','} : char32_t{'.'}; |
| 167 | } |
| 168 | |
| 169 | // Prepares input from a field, and returns the sign, if any, else '\0'. |
| 170 | static RT_API_ATTRS char ScanNumericPrefix(IoStatementState &io, |
| 171 | const DataEdit &edit, Fortran::common::optional<char32_t> &next, |
| 172 | Fortran::common::optional<int> &remaining, |
| 173 | IoStatementState::FastAsciiField *fastField = nullptr) { |
| 174 | remaining = io.CueUpInput(edit, fastField); |
| 175 | next = io.NextInField(remaining, edit, fastField); |
| 176 | char sign{'\0'}; |
| 177 | if (next) { |
| 178 | if (*next == '-' || *next == '+') { |
| 179 | sign = *next; |
| 180 | if (!edit.IsListDirected()) { |
| 181 | io.SkipSpaces(remaining, fastField); |
| 182 | } |
| 183 | next = io.NextInField(remaining, edit, fastField); |
| 184 | } |
| 185 | } |
| 186 | return sign; |
| 187 | } |
| 188 | |
| 189 | RT_API_ATTRS bool EditIntegerInput(IoStatementState &io, const DataEdit &edit, |
| 190 | void *n, int kind, bool isSigned) { |
| 191 | auto &handler{io.GetIoErrorHandler()}; |
| 192 | RUNTIME_CHECK(handler, kind >= 1 && !(kind & (kind - 1))); |
| 193 | if (!n) { |
| 194 | handler.Crash("Null address for integer input item" ); |
| 195 | } |
| 196 | switch (edit.descriptor) { |
| 197 | case DataEdit::ListDirected: |
| 198 | if (IsNamelistNameOrSlash(io)) { |
| 199 | return false; |
| 200 | } |
| 201 | break; |
| 202 | case 'G': |
| 203 | case 'I': |
| 204 | break; |
| 205 | case 'B': |
| 206 | return EditBOZInput<1>(io, edit, n, kind); |
| 207 | case 'O': |
| 208 | return EditBOZInput<3>(io, edit, n, kind); |
| 209 | case 'Z': |
| 210 | return EditBOZInput<4>(io, edit, n, kind); |
| 211 | case 'A': // legacy extension |
| 212 | return EditCharacterInput(io, edit, reinterpret_cast<char *>(n), kind); |
| 213 | default: |
| 214 | handler.SignalError(IostatErrorInFormat, |
| 215 | "Data edit descriptor '%c' may not be used with an INTEGER data item" , |
| 216 | edit.descriptor); |
| 217 | return false; |
| 218 | } |
| 219 | Fortran::common::optional<int> remaining; |
| 220 | Fortran::common::optional<char32_t> next; |
| 221 | auto fastField{io.GetUpcomingFastAsciiField()}; |
| 222 | char sign{ScanNumericPrefix(io, edit, next, remaining, &fastField)}; |
| 223 | if (sign == '-' && !isSigned) { |
| 224 | handler.SignalError("Negative sign in UNSIGNED input field" ); |
| 225 | return false; |
| 226 | } |
| 227 | common::uint128_t value{0}; |
| 228 | bool any{!!sign}; |
| 229 | bool overflow{false}; |
| 230 | const char32_t comma{GetSeparatorChar(edit)}; |
| 231 | static constexpr auto maxu128{~common::uint128_t{0}}; |
| 232 | for (; next; next = io.NextInField(remaining, edit, &fastField)) { |
| 233 | char32_t ch{*next}; |
| 234 | if (ch == ' ' || ch == '\t') { |
| 235 | if (edit.modes.editingFlags & blankZero) { |
| 236 | ch = '0'; // BZ mode - treat blank as if it were zero |
| 237 | } else { |
| 238 | continue; |
| 239 | } |
| 240 | } |
| 241 | int digit{0}; |
| 242 | if (ch >= '0' && ch <= '9') { |
| 243 | digit = ch - '0'; |
| 244 | } else if (ch == comma) { |
| 245 | break; // end non-list-directed field early |
| 246 | } else { |
| 247 | if (edit.modes.inNamelist && ch == GetRadixPointChar(edit)) { |
| 248 | // Ignore any fractional part that might appear in NAMELIST integer |
| 249 | // input, like a few other Fortran compilers do. |
| 250 | // TODO: also process exponents? Some compilers do, but they obviously |
| 251 | // can't just be ignored. |
| 252 | while ((next = io.NextInField(remaining, edit, &fastField))) { |
| 253 | if (*next < '0' || *next > '9') { |
| 254 | break; |
| 255 | } |
| 256 | } |
| 257 | if (!next || *next == comma) { |
| 258 | break; |
| 259 | } |
| 260 | } |
| 261 | handler.SignalError("Bad character '%lc' in INTEGER input field" , ch); |
| 262 | return false; |
| 263 | } |
| 264 | static constexpr auto maxu128OverTen{maxu128 / 10}; |
| 265 | static constexpr int maxLastDigit{ |
| 266 | static_cast<int>(maxu128 - (maxu128OverTen * 10))}; |
| 267 | overflow |= value >= maxu128OverTen && |
| 268 | (value > maxu128OverTen || digit > maxLastDigit); |
| 269 | value *= 10; |
| 270 | value += digit; |
| 271 | any = true; |
| 272 | } |
| 273 | if (!any && !remaining) { |
| 274 | handler.SignalError( |
| 275 | "Integer value absent from NAMELIST or list-directed input" ); |
| 276 | return false; |
| 277 | } |
| 278 | if (isSigned) { |
| 279 | auto maxForKind{common::uint128_t{1} << ((8 * kind) - 1)}; |
| 280 | overflow |= value >= maxForKind && (value > maxForKind || sign != '-'); |
| 281 | } else { |
| 282 | auto maxForKind{maxu128 >> (((16 - kind) * 8) + (isSigned ? 1 : 0))}; |
| 283 | overflow |= value >= maxForKind; |
| 284 | } |
| 285 | if (overflow) { |
| 286 | handler.SignalError(IostatIntegerInputOverflow, |
| 287 | "Decimal input overflows INTEGER(%d) variable" , kind); |
| 288 | return false; |
| 289 | } |
| 290 | if (sign == '-') { |
| 291 | value = -value; |
| 292 | } |
| 293 | if (any || !handler.InError()) { |
| 294 | // The value is stored in the lower order bits on big endian platform. |
| 295 | // For memcpy, shift the value to the highest order bits. |
| 296 | #if USING_NATIVE_INT128_T |
| 297 | auto shft{static_cast<int>(sizeof value - kind)}; |
| 298 | if (!isHostLittleEndian && shft >= 0) { |
| 299 | auto shifted{value << (8 * shft)}; |
| 300 | std::memcpy(n, &shifted, kind); |
| 301 | } else { |
| 302 | std::memcpy(n, &value, kind); // a blank field means zero |
| 303 | } |
| 304 | #else |
| 305 | auto shft{static_cast<int>(sizeof(value.low())) - kind}; |
| 306 | // For kind==8 (i.e. shft==0), the value is stored in low_ in big endian. |
| 307 | if (!isHostLittleEndian && shft >= 0) { |
| 308 | auto l{value.low() << (8 * shft)}; |
| 309 | std::memcpy(n, &l, kind); |
| 310 | } else { |
| 311 | std::memcpy(n, &value, kind); // a blank field means zero |
| 312 | } |
| 313 | #endif |
| 314 | io.GotChar(fastField.got()); |
| 315 | return true; |
| 316 | } else { |
| 317 | return false; |
| 318 | } |
| 319 | } |
| 320 | |
| 321 | // Parses a REAL input number from the input source as a normalized |
| 322 | // fraction into a supplied buffer -- there's an optional '-', a |
| 323 | // decimal point when the input is not hexadecimal, and at least one |
| 324 | // digit. Replaces blanks with zeroes where appropriate. |
| 325 | struct ScannedRealInput { |
| 326 | // Number of characters that (should) have been written to the |
| 327 | // buffer -- this can be larger than the buffer size, which |
| 328 | // indicates buffer overflow. Zero indicates an error. |
| 329 | int got{0}; |
| 330 | int exponent{0}; // adjusted as necessary; binary if isHexadecimal |
| 331 | bool isHexadecimal{false}; // 0X... |
| 332 | }; |
| 333 | static RT_API_ATTRS ScannedRealInput ScanRealInput( |
| 334 | char *buffer, int bufferSize, IoStatementState &io, const DataEdit &edit) { |
| 335 | Fortran::common::optional<int> remaining; |
| 336 | Fortran::common::optional<char32_t> next; |
| 337 | int got{0}; |
| 338 | Fortran::common::optional<int> radixPointOffset; |
| 339 | // The following lambda definition violates the conding style, |
| 340 | // but cuda-11.8 nvcc hits an internal error with the brace initialization. |
| 341 | auto Put = [&](char ch) -> void { |
| 342 | if (got < bufferSize) { |
| 343 | buffer[got] = ch; |
| 344 | } |
| 345 | ++got; |
| 346 | }; |
| 347 | char sign{ScanNumericPrefix(io, edit, next, remaining)}; |
| 348 | if (sign == '-') { |
| 349 | Put('-'); |
| 350 | } |
| 351 | bool bzMode{(edit.modes.editingFlags & blankZero) != 0}; |
| 352 | int exponent{0}; |
| 353 | if (!next || (!bzMode && *next == ' ') || |
| 354 | (!(edit.modes.editingFlags & decimalComma) && *next == ',')) { |
| 355 | if (!edit.IsListDirected() && !io.GetConnectionState().IsAtEOF()) { |
| 356 | // An empty/blank field means zero when not list-directed. |
| 357 | // A fixed-width field containing only a sign is also zero; |
| 358 | // this behavior isn't standard-conforming in F'2023 but it is |
| 359 | // required to pass FCVS. |
| 360 | Put('0'); |
| 361 | } |
| 362 | return {got, exponent, false}; |
| 363 | } |
| 364 | char32_t radixPointChar{GetRadixPointChar(edit)}; |
| 365 | char32_t first{*next >= 'a' && *next <= 'z' ? *next + 'A' - 'a' : *next}; |
| 366 | bool isHexadecimal{false}; |
| 367 | if (first == 'N' || first == 'I') { |
| 368 | // NaN or infinity - convert to upper case |
| 369 | // Subtle: a blank field of digits could be followed by 'E' or 'D', |
| 370 | for (; next && |
| 371 | ((*next >= 'a' && *next <= 'z') || (*next >= 'A' && *next <= 'Z')); |
| 372 | next = io.NextInField(remaining, edit)) { |
| 373 | if (*next >= 'a' && *next <= 'z') { |
| 374 | Put(*next - 'a' + 'A'); |
| 375 | } else { |
| 376 | Put(*next); |
| 377 | } |
| 378 | } |
| 379 | if (next && *next == '(') { // NaN(...) |
| 380 | Put('('); |
| 381 | int depth{1}; |
| 382 | while (true) { |
| 383 | next = io.NextInField(remaining, edit); |
| 384 | if (depth == 0) { |
| 385 | break; |
| 386 | } else if (!next) { |
| 387 | return {}; // error |
| 388 | } else if (*next == '(') { |
| 389 | ++depth; |
| 390 | } else if (*next == ')') { |
| 391 | --depth; |
| 392 | } |
| 393 | Put(*next); |
| 394 | } |
| 395 | } |
| 396 | } else if (first == radixPointChar || (first >= '0' && first <= '9') || |
| 397 | (bzMode && (first == ' ' || first == '\t')) || first == 'E' || |
| 398 | first == 'D' || first == 'Q') { |
| 399 | if (first == '0') { |
| 400 | next = io.NextInField(remaining, edit); |
| 401 | if (next && (*next == 'x' || *next == 'X')) { // 0X... |
| 402 | isHexadecimal = true; |
| 403 | next = io.NextInField(remaining, edit); |
| 404 | } else { |
| 405 | Put('0'); |
| 406 | } |
| 407 | } |
| 408 | // input field is normalized to a fraction |
| 409 | if (!isHexadecimal) { |
| 410 | Put('.'); |
| 411 | } |
| 412 | auto start{got}; |
| 413 | for (; next; next = io.NextInField(remaining, edit)) { |
| 414 | char32_t ch{*next}; |
| 415 | if (ch == ' ' || ch == '\t') { |
| 416 | if (isHexadecimal) { |
| 417 | return {}; // error |
| 418 | } else if (bzMode) { |
| 419 | ch = '0'; // BZ mode - treat blank as if it were zero |
| 420 | } else { |
| 421 | continue; // ignore blank in fixed field |
| 422 | } |
| 423 | } |
| 424 | if (ch == '0' && got == start && !radixPointOffset) { |
| 425 | // omit leading zeroes before the radix point |
| 426 | } else if (ch >= '0' && ch <= '9') { |
| 427 | Put(ch); |
| 428 | } else if (ch == radixPointChar && !radixPointOffset) { |
| 429 | // The radix point character is *not* copied to the buffer. |
| 430 | radixPointOffset = got - start; // # of digits before the radix point |
| 431 | } else if (isHexadecimal && ch >= 'A' && ch <= 'F') { |
| 432 | Put(ch); |
| 433 | } else if (isHexadecimal && ch >= 'a' && ch <= 'f') { |
| 434 | Put(ch - 'a' + 'A'); // normalize to capitals |
| 435 | } else { |
| 436 | break; |
| 437 | } |
| 438 | } |
| 439 | if (got == start) { |
| 440 | // Nothing but zeroes and maybe a radix point. F'2018 requires |
| 441 | // at least one digit, but F'77 did not, and a bare "." shows up in |
| 442 | // the FCVS suite. |
| 443 | Put('0'); // emit at least one digit |
| 444 | } |
| 445 | // In list-directed input, a bad exponent is not consumed. |
| 446 | auto nextBeforeExponent{next}; |
| 447 | auto startExponent{io.GetConnectionState().positionInRecord}; |
| 448 | bool hasGoodExponent{false}; |
| 449 | if (next) { |
| 450 | if (isHexadecimal) { |
| 451 | if (*next == 'p' || *next == 'P') { |
| 452 | next = io.NextInField(remaining, edit); |
| 453 | } else { |
| 454 | // The binary exponent is not optional in the standard. |
| 455 | return {}; // error |
| 456 | } |
| 457 | } else if (*next == 'e' || *next == 'E' || *next == 'd' || *next == 'D' || |
| 458 | *next == 'q' || *next == 'Q') { |
| 459 | // Optional exponent letter. Blanks are allowed between the |
| 460 | // optional exponent letter and the exponent value. |
| 461 | io.SkipSpaces(remaining); |
| 462 | next = io.NextInField(remaining, edit); |
| 463 | } |
| 464 | } |
| 465 | if (next && |
| 466 | (*next == '-' || *next == '+' || (*next >= '0' && *next <= '9') || |
| 467 | *next == ' ' || *next == '\t')) { |
| 468 | bool negExpo{*next == '-'}; |
| 469 | if (negExpo || *next == '+') { |
| 470 | next = io.NextInField(remaining, edit); |
| 471 | } |
| 472 | for (; next; next = io.NextInField(remaining, edit)) { |
| 473 | if (*next >= '0' && *next <= '9') { |
| 474 | hasGoodExponent = true; |
| 475 | if (exponent < 10000) { |
| 476 | exponent = 10 * exponent + *next - '0'; |
| 477 | } |
| 478 | } else if (*next == ' ' || *next == '\t') { |
| 479 | if (isHexadecimal) { |
| 480 | break; |
| 481 | } else if (bzMode) { |
| 482 | hasGoodExponent = true; |
| 483 | exponent = 10 * exponent; |
| 484 | } |
| 485 | } else { |
| 486 | break; |
| 487 | } |
| 488 | } |
| 489 | if (negExpo) { |
| 490 | exponent = -exponent; |
| 491 | } |
| 492 | } |
| 493 | if (!hasGoodExponent) { |
| 494 | if (isHexadecimal) { |
| 495 | return {}; // error |
| 496 | } |
| 497 | // There isn't a good exponent; do not consume it. |
| 498 | next = nextBeforeExponent; |
| 499 | io.HandleAbsolutePosition(startExponent); |
| 500 | // The default exponent is -kP, but the scale factor doesn't affect |
| 501 | // an explicit exponent. |
| 502 | exponent = -edit.modes.scale; |
| 503 | } |
| 504 | // Adjust exponent by number of digits before the radix point. |
| 505 | if (isHexadecimal) { |
| 506 | // Exponents for hexadecimal input are binary. |
| 507 | exponent += radixPointOffset.value_or(got - start) * 4; |
| 508 | } else if (radixPointOffset) { |
| 509 | exponent += *radixPointOffset; |
| 510 | } else { |
| 511 | // When no redix point (or comma) appears in the value, the 'd' |
| 512 | // part of the edit descriptor must be interpreted as the number of |
| 513 | // digits in the value to be interpreted as being to the *right* of |
| 514 | // the assumed radix point (13.7.2.3.2) |
| 515 | exponent += got - start - edit.digits.value_or(0); |
| 516 | } |
| 517 | } |
| 518 | // Consume the trailing ')' of a list-directed or NAMELIST complex |
| 519 | // input value. |
| 520 | if (edit.descriptor == DataEdit::ListDirectedImaginaryPart) { |
| 521 | if (next && (*next == ' ' || *next == '\t')) { |
| 522 | io.SkipSpaces(remaining); |
| 523 | next = io.NextInField(remaining, edit); |
| 524 | } |
| 525 | if (!next) { // NextInField fails on separators like ')' |
| 526 | std::size_t byteCount{0}; |
| 527 | next = io.GetCurrentChar(byteCount); |
| 528 | if (next && *next == ')') { |
| 529 | io.HandleRelativePosition(byteCount); |
| 530 | } |
| 531 | } |
| 532 | } else if (remaining) { |
| 533 | while (next && (*next == ' ' || *next == '\t')) { |
| 534 | next = io.NextInField(remaining, edit); |
| 535 | } |
| 536 | if (next && (*next != ',' || (edit.modes.editingFlags & decimalComma))) { |
| 537 | return {}; // error: unused nonblank character in fixed-width field |
| 538 | } |
| 539 | } |
| 540 | return {got, exponent, isHexadecimal}; |
| 541 | } |
| 542 | |
| 543 | static RT_API_ATTRS void RaiseFPExceptions( |
| 544 | decimal::ConversionResultFlags flags) { |
| 545 | #undef RAISE |
| 546 | #if defined(RT_DEVICE_COMPILATION) |
| 547 | Terminator terminator(__FILE__, __LINE__); |
| 548 | #define RAISE(e) \ |
| 549 | terminator.Crash( \ |
| 550 | "not implemented yet: raising FP exception in device code: %s", #e); |
| 551 | #else // !defined(RT_DEVICE_COMPILATION) |
| 552 | #ifdef feraisexcept // a macro in some environments; omit std:: |
| 553 | #define RAISE feraiseexcept |
| 554 | #else |
| 555 | #define RAISE std::feraiseexcept |
| 556 | #endif |
| 557 | #endif // !defined(RT_DEVICE_COMPILATION) |
| 558 | |
| 559 | // Some environment (e.g. emscripten, musl) don't define FE_OVERFLOW as allowed |
| 560 | // by c99 (but not c++11) :-/ |
| 561 | #if defined(FE_OVERFLOW) || defined(RT_DEVICE_COMPILATION) |
| 562 | if (flags & decimal::ConversionResultFlags::Overflow) { |
| 563 | RAISE(FE_OVERFLOW); |
| 564 | } |
| 565 | #endif |
| 566 | #if defined(FE_UNDERFLOW) || defined(RT_DEVICE_COMPILATION) |
| 567 | if (flags & decimal::ConversionResultFlags::Underflow) { |
| 568 | RAISE(FE_UNDERFLOW); |
| 569 | } |
| 570 | #endif |
| 571 | #if defined(FE_INEXACT) || defined(RT_DEVICE_COMPILATION) |
| 572 | if (flags & decimal::ConversionResultFlags::Inexact) { |
| 573 | RAISE(FE_INEXACT); |
| 574 | } |
| 575 | #endif |
| 576 | #if defined(FE_INVALID) || defined(RT_DEVICE_COMPILATION) |
| 577 | if (flags & decimal::ConversionResultFlags::Invalid) { |
| 578 | RAISE(FE_INVALID); |
| 579 | } |
| 580 | #endif |
| 581 | #undef RAISE |
| 582 | } |
| 583 | |
| 584 | // If no special modes are in effect and the form of the input value |
| 585 | // that's present in the input stream is acceptable to the decimal->binary |
| 586 | // converter without modification, this fast path for real input |
| 587 | // saves time by avoiding memory copies and reformatting of the exponent. |
| 588 | template <int PRECISION> |
| 589 | static RT_API_ATTRS bool TryFastPathRealDecimalInput( |
| 590 | IoStatementState &io, const DataEdit &edit, void *n) { |
| 591 | if (edit.modes.editingFlags & (blankZero | decimalComma)) { |
| 592 | return false; |
| 593 | } |
| 594 | if (edit.modes.scale != 0) { |
| 595 | return false; |
| 596 | } |
| 597 | const ConnectionState &connection{io.GetConnectionState()}; |
| 598 | if (connection.internalIoCharKind > 1) { |
| 599 | return false; // reading non-default character |
| 600 | } |
| 601 | const char *str{nullptr}; |
| 602 | std::size_t got{io.GetNextInputBytes(str)}; |
| 603 | if (got == 0 || str == nullptr || !connection.recordLength.has_value()) { |
| 604 | return false; // could not access reliably-terminated input stream |
| 605 | } |
| 606 | const char *p{str}; |
| 607 | std::int64_t maxConsume{ |
| 608 | std::min<std::int64_t>(got, edit.width.value_or(got))}; |
| 609 | const char *limit{str + maxConsume}; |
| 610 | decimal::ConversionToBinaryResult<PRECISION> converted{ |
| 611 | decimal::ConvertToBinary<PRECISION>(p, edit.modes.round, limit)}; |
| 612 | if (converted.flags & (decimal::Invalid | decimal::Overflow)) { |
| 613 | return false; |
| 614 | } |
| 615 | if (edit.digits.value_or(0) != 0) { |
| 616 | // Edit descriptor is Fw.d (or other) with d != 0, which |
| 617 | // implies scaling |
| 618 | const char *q{str}; |
| 619 | for (; q < limit; ++q) { |
| 620 | if (*q == '.' || *q == 'n' || *q == 'N') { |
| 621 | break; |
| 622 | } |
| 623 | } |
| 624 | if (q == limit) { |
| 625 | // No explicit decimal point, and not NaN/Inf. |
| 626 | return false; |
| 627 | } |
| 628 | } |
| 629 | if (edit.descriptor == DataEdit::ListDirectedImaginaryPart) { |
| 630 | // Need to consume a trailing ')', possibly with leading spaces |
| 631 | for (; p < limit && (*p == ' ' || *p == '\t'); ++p) { |
| 632 | } |
| 633 | if (p < limit && *p == ')') { |
| 634 | ++p; |
| 635 | } else { |
| 636 | return false; |
| 637 | } |
| 638 | } else if (edit.IsListDirected()) { |
| 639 | if (p < limit && !IsCharValueSeparator(edit, *p)) { |
| 640 | return false; |
| 641 | } |
| 642 | } else { |
| 643 | for (; p < limit && (*p == ' ' || *p == '\t'); ++p) { |
| 644 | } |
| 645 | if (edit.width && p < str + *edit.width) { |
| 646 | return false; // unconverted characters remain in fixed width field |
| 647 | } |
| 648 | } |
| 649 | // Success on the fast path! |
| 650 | *reinterpret_cast<decimal::BinaryFloatingPointNumber<PRECISION> *>(n) = |
| 651 | converted.binary; |
| 652 | io.HandleRelativePosition(p - str); |
| 653 | // Set FP exception flags |
| 654 | if (converted.flags != decimal::ConversionResultFlags::Exact) { |
| 655 | RaiseFPExceptions(converted.flags); |
| 656 | } |
| 657 | return true; |
| 658 | } |
| 659 | |
| 660 | template <int binaryPrecision> |
| 661 | RT_API_ATTRS decimal::ConversionToBinaryResult<binaryPrecision> |
| 662 | ConvertHexadecimal( |
| 663 | const char *&p, enum decimal::FortranRounding rounding, int expo) { |
| 664 | using RealType = decimal::BinaryFloatingPointNumber<binaryPrecision>; |
| 665 | using RawType = typename RealType::RawType; |
| 666 | bool isNegative{*p == '-'}; |
| 667 | constexpr RawType one{1}; |
| 668 | RawType signBit{0}; |
| 669 | if (isNegative) { |
| 670 | ++p; |
| 671 | signBit = one << (RealType::bits - 1); |
| 672 | } |
| 673 | RawType fraction{0}; |
| 674 | // Adjust the incoming binary P+/- exponent to shift the radix point |
| 675 | // to below the LSB and add in the bias. |
| 676 | expo += binaryPrecision - 1 + RealType::exponentBias; |
| 677 | // Input the fraction. |
| 678 | int roundingBit{0}; |
| 679 | int guardBit{0}; |
| 680 | for (; *p; ++p) { |
| 681 | fraction <<= 4; |
| 682 | expo -= 4; |
| 683 | if (*p >= '0' && *p <= '9') { |
| 684 | fraction |= *p - '0'; |
| 685 | } else if (*p >= 'A' && *p <= 'F') { |
| 686 | fraction |= *p - 'A' + 10; // data were normalized to capitals |
| 687 | } else { |
| 688 | break; |
| 689 | } |
| 690 | if (fraction >> binaryPrecision) { |
| 691 | while (fraction >> binaryPrecision) { |
| 692 | guardBit |= roundingBit; |
| 693 | roundingBit = (int)fraction & 1; |
| 694 | fraction >>= 1; |
| 695 | ++expo; |
| 696 | } |
| 697 | // Consume excess digits |
| 698 | while (*++p) { |
| 699 | if (*p == '0') { |
| 700 | } else if ((*p >= '1' && *p <= '9') || (*p >= 'A' && *p <= 'F')) { |
| 701 | guardBit = 1; |
| 702 | } else { |
| 703 | break; |
| 704 | } |
| 705 | } |
| 706 | break; |
| 707 | } |
| 708 | } |
| 709 | if (fraction) { |
| 710 | // Boost biased expo if too small |
| 711 | while (expo < 1) { |
| 712 | guardBit |= roundingBit; |
| 713 | roundingBit = (int)fraction & 1; |
| 714 | fraction >>= 1; |
| 715 | ++expo; |
| 716 | } |
| 717 | // Normalize |
| 718 | while (expo > 1 && !(fraction >> (binaryPrecision - 1))) { |
| 719 | fraction <<= 1; |
| 720 | --expo; |
| 721 | guardBit = roundingBit = 0; |
| 722 | } |
| 723 | } |
| 724 | // Rounding |
| 725 | bool increase{false}; |
| 726 | switch (rounding) { |
| 727 | case decimal::RoundNearest: // RN & RP |
| 728 | increase = roundingBit && (guardBit | ((int)fraction & 1)); |
| 729 | break; |
| 730 | case decimal::RoundUp: // RU |
| 731 | increase = !isNegative && (roundingBit | guardBit); |
| 732 | break; |
| 733 | case decimal::RoundDown: // RD |
| 734 | increase = isNegative && (roundingBit | guardBit); |
| 735 | break; |
| 736 | case decimal::RoundToZero: // RZ |
| 737 | break; |
| 738 | case decimal::RoundCompatible: // RC |
| 739 | increase = roundingBit != 0; |
| 740 | break; |
| 741 | } |
| 742 | if (increase) { |
| 743 | ++fraction; |
| 744 | if (fraction >> binaryPrecision) { |
| 745 | fraction >>= 1; |
| 746 | ++expo; |
| 747 | } |
| 748 | } |
| 749 | // Package & return result |
| 750 | constexpr RawType significandMask{(one << RealType::significandBits) - 1}; |
| 751 | int flags{(roundingBit | guardBit) ? decimal::Inexact : decimal::Exact}; |
| 752 | if (!fraction) { |
| 753 | expo = 0; |
| 754 | } else if (expo == 1 && !(fraction >> (binaryPrecision - 1))) { |
| 755 | expo = 0; // subnormal |
| 756 | flags |= decimal::Underflow; |
| 757 | } else if (expo >= RealType::maxExponent) { |
| 758 | if (rounding == decimal::RoundToZero || |
| 759 | (rounding == decimal::RoundDown && !isNegative) || |
| 760 | (rounding == decimal::RoundUp && isNegative)) { |
| 761 | expo = RealType::maxExponent - 1; // +/-HUGE() |
| 762 | fraction = significandMask; |
| 763 | } else { |
| 764 | expo = RealType::maxExponent; // +/-Inf |
| 765 | fraction = 0; |
| 766 | flags |= decimal::Overflow; |
| 767 | } |
| 768 | } else { |
| 769 | fraction &= significandMask; // remove explicit normalization unless x87 |
| 770 | } |
| 771 | return decimal::ConversionToBinaryResult<binaryPrecision>{ |
| 772 | RealType{static_cast<RawType>(signBit | |
| 773 | static_cast<RawType>(expo) << RealType::significandBits | fraction)}, |
| 774 | static_cast<decimal::ConversionResultFlags>(flags)}; |
| 775 | } |
| 776 | |
| 777 | template <int KIND> |
| 778 | RT_API_ATTRS bool EditCommonRealInput( |
| 779 | IoStatementState &io, const DataEdit &edit, void *n) { |
| 780 | constexpr int binaryPrecision{common::PrecisionOfRealKind(KIND)}; |
| 781 | if (TryFastPathRealDecimalInput<binaryPrecision>(io, edit, n)) { |
| 782 | return CheckCompleteListDirectedField(io, edit); |
| 783 | } |
| 784 | // Fast path wasn't available or didn't work; go the more general route |
| 785 | static constexpr int maxDigits{ |
| 786 | common::MaxDecimalConversionDigits(binaryPrecision)}; |
| 787 | static constexpr int bufferSize{maxDigits + 18}; |
| 788 | char buffer[bufferSize]; |
| 789 | auto scanned{ScanRealInput(buffer, maxDigits + 2, io, edit)}; |
| 790 | int got{scanned.got}; |
| 791 | if (got >= maxDigits + 2) { |
| 792 | io.GetIoErrorHandler().Crash("EditCommonRealInput: buffer was too small" ); |
| 793 | return false; |
| 794 | } |
| 795 | if (got == 0) { |
| 796 | const auto &connection{io.GetConnectionState()}; |
| 797 | io.GetIoErrorHandler().SignalError(IostatBadRealInput, |
| 798 | "Bad real input data at column %d of record %d" , |
| 799 | static_cast<int>(connection.positionInRecord + 1), |
| 800 | static_cast<int>(connection.currentRecordNumber)); |
| 801 | return false; |
| 802 | } |
| 803 | decimal::ConversionToBinaryResult<binaryPrecision> converted; |
| 804 | const char *p{buffer}; |
| 805 | if (scanned.isHexadecimal) { |
| 806 | buffer[got] = '\0'; |
| 807 | converted = ConvertHexadecimal<binaryPrecision>( |
| 808 | p, edit.modes.round, scanned.exponent); |
| 809 | } else { |
| 810 | bool {got > maxDigits}; |
| 811 | int exponent{scanned.exponent}; |
| 812 | if (exponent != 0) { |
| 813 | buffer[got++] = 'e'; |
| 814 | if (exponent < 0) { |
| 815 | buffer[got++] = '-'; |
| 816 | exponent = -exponent; |
| 817 | } |
| 818 | if (exponent > 9999) { |
| 819 | exponent = 9999; // will convert to +/-Inf |
| 820 | } |
| 821 | if (exponent > 999) { |
| 822 | int dig{exponent / 1000}; |
| 823 | buffer[got++] = '0' + dig; |
| 824 | int rest{exponent - 1000 * dig}; |
| 825 | dig = rest / 100; |
| 826 | buffer[got++] = '0' + dig; |
| 827 | rest -= 100 * dig; |
| 828 | dig = rest / 10; |
| 829 | buffer[got++] = '0' + dig; |
| 830 | buffer[got++] = '0' + (rest - 10 * dig); |
| 831 | } else if (exponent > 99) { |
| 832 | int dig{exponent / 100}; |
| 833 | buffer[got++] = '0' + dig; |
| 834 | int rest{exponent - 100 * dig}; |
| 835 | dig = rest / 10; |
| 836 | buffer[got++] = '0' + dig; |
| 837 | buffer[got++] = '0' + (rest - 10 * dig); |
| 838 | } else if (exponent > 9) { |
| 839 | int dig{exponent / 10}; |
| 840 | buffer[got++] = '0' + dig; |
| 841 | buffer[got++] = '0' + (exponent - 10 * dig); |
| 842 | } else { |
| 843 | buffer[got++] = '0' + exponent; |
| 844 | } |
| 845 | } |
| 846 | buffer[got] = '\0'; |
| 847 | converted = decimal::ConvertToBinary<binaryPrecision>(p, edit.modes.round); |
| 848 | if (hadExtra) { |
| 849 | converted.flags = static_cast<enum decimal::ConversionResultFlags>( |
| 850 | converted.flags | decimal::Inexact); |
| 851 | } |
| 852 | } |
| 853 | if (*p) { // unprocessed junk after value |
| 854 | const auto &connection{io.GetConnectionState()}; |
| 855 | io.GetIoErrorHandler().SignalError(IostatBadRealInput, |
| 856 | "Trailing characters after real input data at column %d of record %d" , |
| 857 | static_cast<int>(connection.positionInRecord + 1), |
| 858 | static_cast<int>(connection.currentRecordNumber)); |
| 859 | return false; |
| 860 | } |
| 861 | *reinterpret_cast<decimal::BinaryFloatingPointNumber<binaryPrecision> *>(n) = |
| 862 | converted.binary; |
| 863 | // Set FP exception flags |
| 864 | if (converted.flags != decimal::ConversionResultFlags::Exact) { |
| 865 | if (converted.flags & decimal::ConversionResultFlags::Overflow) { |
| 866 | io.GetIoErrorHandler().SignalError(IostatRealInputOverflow); |
| 867 | return false; |
| 868 | } |
| 869 | RaiseFPExceptions(converted.flags); |
| 870 | } |
| 871 | return CheckCompleteListDirectedField(io, edit); |
| 872 | } |
| 873 | |
| 874 | template <int KIND> |
| 875 | RT_API_ATTRS bool EditRealInput( |
| 876 | IoStatementState &io, const DataEdit &edit, void *n) { |
| 877 | switch (edit.descriptor) { |
| 878 | case DataEdit::ListDirected: |
| 879 | if (IsNamelistNameOrSlash(io)) { |
| 880 | return false; |
| 881 | } |
| 882 | return EditCommonRealInput<KIND>(io, edit, n); |
| 883 | case DataEdit::ListDirectedRealPart: |
| 884 | case DataEdit::ListDirectedImaginaryPart: |
| 885 | case 'F': |
| 886 | case 'E': // incl. EN, ES, & EX |
| 887 | case 'D': |
| 888 | case 'G': |
| 889 | return EditCommonRealInput<KIND>(io, edit, n); |
| 890 | case 'B': |
| 891 | return EditBOZInput<1>(io, edit, n, |
| 892 | common::BitsForBinaryPrecision(common::PrecisionOfRealKind(KIND)) >> 3); |
| 893 | case 'O': |
| 894 | return EditBOZInput<3>(io, edit, n, |
| 895 | common::BitsForBinaryPrecision(common::PrecisionOfRealKind(KIND)) >> 3); |
| 896 | case 'Z': |
| 897 | return EditBOZInput<4>(io, edit, n, |
| 898 | common::BitsForBinaryPrecision(common::PrecisionOfRealKind(KIND)) >> 3); |
| 899 | case 'A': // legacy extension |
| 900 | return EditCharacterInput(io, edit, reinterpret_cast<char *>(n), KIND); |
| 901 | default: |
| 902 | io.GetIoErrorHandler().SignalError(IostatErrorInFormat, |
| 903 | "Data edit descriptor '%c' may not be used for REAL input" , |
| 904 | edit.descriptor); |
| 905 | return false; |
| 906 | } |
| 907 | } |
| 908 | |
| 909 | // 13.7.3 in Fortran 2018 |
| 910 | RT_API_ATTRS bool EditLogicalInput( |
| 911 | IoStatementState &io, const DataEdit &edit, bool &x) { |
| 912 | switch (edit.descriptor) { |
| 913 | case DataEdit::ListDirected: |
| 914 | if (IsNamelistNameOrSlash(io)) { |
| 915 | return false; |
| 916 | } |
| 917 | break; |
| 918 | case 'L': |
| 919 | case 'G': |
| 920 | break; |
| 921 | default: |
| 922 | io.GetIoErrorHandler().SignalError(IostatErrorInFormat, |
| 923 | "Data edit descriptor '%c' may not be used for LOGICAL input" , |
| 924 | edit.descriptor); |
| 925 | return false; |
| 926 | } |
| 927 | Fortran::common::optional<int> remaining{io.CueUpInput(edit)}; |
| 928 | Fortran::common::optional<char32_t> next{io.NextInField(remaining, edit)}; |
| 929 | if (next && *next == '.') { // skip optional period |
| 930 | next = io.NextInField(remaining, edit); |
| 931 | } |
| 932 | if (!next) { |
| 933 | io.GetIoErrorHandler().SignalError("Empty LOGICAL input field" ); |
| 934 | return false; |
| 935 | } |
| 936 | switch (*next) { |
| 937 | case 'T': |
| 938 | case 't': |
| 939 | x = true; |
| 940 | break; |
| 941 | case 'F': |
| 942 | case 'f': |
| 943 | x = false; |
| 944 | break; |
| 945 | default: |
| 946 | io.GetIoErrorHandler().SignalError( |
| 947 | "Bad character '%lc' in LOGICAL input field" , *next); |
| 948 | return false; |
| 949 | } |
| 950 | if (remaining) { // ignore the rest of a fixed-width field |
| 951 | io.HandleRelativePosition(*remaining); |
| 952 | } else if (edit.descriptor == DataEdit::ListDirected) { |
| 953 | while (io.NextInField(remaining, edit)) { // discard rest of field |
| 954 | } |
| 955 | } |
| 956 | return CheckCompleteListDirectedField(io, edit); |
| 957 | } |
| 958 | |
| 959 | // See 13.10.3.1 paragraphs 7-9 in Fortran 2018 |
| 960 | template <typename CHAR> |
| 961 | static RT_API_ATTRS bool EditDelimitedCharacterInput( |
| 962 | IoStatementState &io, CHAR *x, std::size_t length, char32_t delimiter) { |
| 963 | bool result{true}; |
| 964 | while (true) { |
| 965 | std::size_t byteCount{0}; |
| 966 | auto ch{io.GetCurrentChar(byteCount)}; |
| 967 | if (!ch) { |
| 968 | if (io.AdvanceRecord()) { |
| 969 | continue; |
| 970 | } else { |
| 971 | result = false; // EOF in character value |
| 972 | break; |
| 973 | } |
| 974 | } |
| 975 | io.HandleRelativePosition(byteCount); |
| 976 | if (*ch == delimiter) { |
| 977 | auto next{io.GetCurrentChar(byteCount)}; |
| 978 | if (next && *next == delimiter) { |
| 979 | // Repeated delimiter: use as character value |
| 980 | io.HandleRelativePosition(byteCount); |
| 981 | } else { |
| 982 | break; // closing delimiter |
| 983 | } |
| 984 | } |
| 985 | if (length > 0) { |
| 986 | *x++ = static_cast<CHAR>(*ch); |
| 987 | --length; |
| 988 | } |
| 989 | } |
| 990 | Fortran::runtime::fill_n(x, length, ' '); |
| 991 | return result; |
| 992 | } |
| 993 | |
| 994 | template <typename CHAR> |
| 995 | static RT_API_ATTRS bool EditListDirectedCharacterInput( |
| 996 | IoStatementState &io, CHAR *x, std::size_t length, const DataEdit &edit) { |
| 997 | std::size_t byteCount{0}; |
| 998 | auto ch{io.GetCurrentChar(byteCount)}; |
| 999 | if (ch && (*ch == '\'' || *ch == '"')) { |
| 1000 | io.HandleRelativePosition(byteCount); |
| 1001 | return EditDelimitedCharacterInput(io, x, length, *ch); |
| 1002 | } |
| 1003 | if (IsNamelistNameOrSlash(io) || io.GetConnectionState().IsAtEOF()) { |
| 1004 | return false; |
| 1005 | } |
| 1006 | // Undelimited list-directed character input: stop at a value separator |
| 1007 | // or the end of the current record. |
| 1008 | while (auto ch{io.GetCurrentChar(byteCount)}) { |
| 1009 | bool isSep{false}; |
| 1010 | switch (*ch) { |
| 1011 | case ' ': |
| 1012 | case '\t': |
| 1013 | case '/': |
| 1014 | isSep = true; |
| 1015 | break; |
| 1016 | case '&': |
| 1017 | case '$': |
| 1018 | isSep = edit.IsNamelist(); |
| 1019 | break; |
| 1020 | case ',': |
| 1021 | isSep = !(edit.modes.editingFlags & decimalComma); |
| 1022 | break; |
| 1023 | case ';': |
| 1024 | isSep = !!(edit.modes.editingFlags & decimalComma); |
| 1025 | break; |
| 1026 | default: |
| 1027 | break; |
| 1028 | } |
| 1029 | if (isSep) { |
| 1030 | break; |
| 1031 | } |
| 1032 | if (length > 0) { |
| 1033 | *x++ = static_cast<CHAR>(*ch); |
| 1034 | --length; |
| 1035 | } else if (edit.IsNamelist()) { |
| 1036 | // GNU compatibility |
| 1037 | break; |
| 1038 | } |
| 1039 | io.HandleRelativePosition(byteCount); |
| 1040 | io.GotChar(byteCount); |
| 1041 | } |
| 1042 | Fortran::runtime::fill_n(x, length, ' '); |
| 1043 | return true; |
| 1044 | } |
| 1045 | |
| 1046 | template <typename CHAR> |
| 1047 | RT_API_ATTRS bool EditCharacterInput(IoStatementState &io, const DataEdit &edit, |
| 1048 | CHAR *x, std::size_t lengthChars) { |
| 1049 | switch (edit.descriptor) { |
| 1050 | case DataEdit::ListDirected: |
| 1051 | return EditListDirectedCharacterInput(io, x, lengthChars, edit); |
| 1052 | case 'A': |
| 1053 | case 'G': |
| 1054 | break; |
| 1055 | case 'B': |
| 1056 | return EditBOZInput<1>(io, edit, x, lengthChars * sizeof *x); |
| 1057 | case 'O': |
| 1058 | return EditBOZInput<3>(io, edit, x, lengthChars * sizeof *x); |
| 1059 | case 'Z': |
| 1060 | return EditBOZInput<4>(io, edit, x, lengthChars * sizeof *x); |
| 1061 | default: |
| 1062 | io.GetIoErrorHandler().SignalError(IostatErrorInFormat, |
| 1063 | "Data edit descriptor '%c' may not be used with a CHARACTER data item" , |
| 1064 | edit.descriptor); |
| 1065 | return false; |
| 1066 | } |
| 1067 | const ConnectionState &connection{io.GetConnectionState()}; |
| 1068 | std::size_t remainingChars{lengthChars}; |
| 1069 | // Skip leading characters. |
| 1070 | // Their bytes don't count towards INQUIRE(IOLENGTH=). |
| 1071 | std::size_t skipChars{0}; |
| 1072 | if (edit.width && *edit.width > 0) { |
| 1073 | remainingChars = *edit.width; |
| 1074 | if (remainingChars > lengthChars) { |
| 1075 | skipChars = remainingChars - lengthChars; |
| 1076 | } |
| 1077 | } |
| 1078 | // When the field is wider than the variable, we drop the leading |
| 1079 | // characters. When the variable is wider than the field, there can be |
| 1080 | // trailing padding or an EOR condition. |
| 1081 | const char *input{nullptr}; |
| 1082 | std::size_t readyBytes{0}; |
| 1083 | // Transfer payload bytes; these do count. |
| 1084 | while (remainingChars > 0) { |
| 1085 | if (readyBytes == 0) { |
| 1086 | readyBytes = io.GetNextInputBytes(input); |
| 1087 | if (readyBytes == 0 || |
| 1088 | (readyBytes < remainingChars && edit.modes.nonAdvancing)) { |
| 1089 | if (io.CheckForEndOfRecord(readyBytes, connection)) { |
| 1090 | if (readyBytes == 0) { |
| 1091 | // PAD='YES' and no more data |
| 1092 | Fortran::runtime::fill_n(x, lengthChars, ' '); |
| 1093 | return !io.GetIoErrorHandler().InError(); |
| 1094 | } else { |
| 1095 | // Do partial read(s) then pad on last iteration |
| 1096 | } |
| 1097 | } else { |
| 1098 | return !io.GetIoErrorHandler().InError(); |
| 1099 | } |
| 1100 | } |
| 1101 | } |
| 1102 | std::size_t chunkBytes; |
| 1103 | std::size_t chunkChars{1}; |
| 1104 | bool skipping{skipChars > 0}; |
| 1105 | if (connection.isUTF8) { |
| 1106 | chunkBytes = MeasureUTF8Bytes(*input); |
| 1107 | if (skipping) { |
| 1108 | --skipChars; |
| 1109 | } else if (auto ucs{DecodeUTF8(input)}) { |
| 1110 | if ((sizeof *x == 1 && *ucs > 0xff) || |
| 1111 | (sizeof *x == 2 && *ucs > 0xffff)) { |
| 1112 | *x++ = '?'; |
| 1113 | } else { |
| 1114 | *x++ = static_cast<CHAR>(*ucs); |
| 1115 | } |
| 1116 | --lengthChars; |
| 1117 | } else if (chunkBytes == 0) { |
| 1118 | // error recovery: skip bad encoding |
| 1119 | chunkBytes = 1; |
| 1120 | } |
| 1121 | } else if (connection.internalIoCharKind > 1) { |
| 1122 | // Reading from non-default character internal unit |
| 1123 | chunkBytes = connection.internalIoCharKind; |
| 1124 | if (skipping) { |
| 1125 | --skipChars; |
| 1126 | } else { |
| 1127 | char32_t buffer{0}; |
| 1128 | std::memcpy(&buffer, input, chunkBytes); |
| 1129 | if ((sizeof *x == 1 && buffer > 0xff) || |
| 1130 | (sizeof *x == 2 && buffer > 0xffff)) { |
| 1131 | *x++ = '?'; |
| 1132 | } else { |
| 1133 | *x++ = static_cast<CHAR>(buffer); |
| 1134 | } |
| 1135 | --lengthChars; |
| 1136 | } |
| 1137 | } else if constexpr (sizeof *x > 1) { |
| 1138 | // Read single byte with expansion into multi-byte CHARACTER |
| 1139 | chunkBytes = 1; |
| 1140 | if (skipping) { |
| 1141 | --skipChars; |
| 1142 | } else { |
| 1143 | *x++ = static_cast<unsigned char>(*input); |
| 1144 | --lengthChars; |
| 1145 | } |
| 1146 | } else { // single bytes -> default CHARACTER |
| 1147 | if (skipping) { |
| 1148 | chunkBytes = std::min<std::size_t>(a: skipChars, b: readyBytes); |
| 1149 | chunkChars = chunkBytes; |
| 1150 | skipChars -= chunkChars; |
| 1151 | } else { |
| 1152 | chunkBytes = std::min<std::size_t>(a: remainingChars, b: readyBytes); |
| 1153 | chunkBytes = std::min<std::size_t>(a: lengthChars, b: chunkBytes); |
| 1154 | chunkChars = chunkBytes; |
| 1155 | std::memcpy(x, input, chunkBytes); |
| 1156 | x += chunkBytes; |
| 1157 | lengthChars -= chunkChars; |
| 1158 | } |
| 1159 | } |
| 1160 | input += chunkBytes; |
| 1161 | remainingChars -= chunkChars; |
| 1162 | if (!skipping) { |
| 1163 | io.GotChar(chunkBytes); |
| 1164 | } |
| 1165 | io.HandleRelativePosition(chunkBytes); |
| 1166 | readyBytes -= chunkBytes; |
| 1167 | } |
| 1168 | // Pad the remainder of the input variable, if any. |
| 1169 | Fortran::runtime::fill_n(x, lengthChars, ' '); |
| 1170 | return CheckCompleteListDirectedField(io, edit); |
| 1171 | } |
| 1172 | |
| 1173 | template RT_API_ATTRS bool EditRealInput<2>( |
| 1174 | IoStatementState &, const DataEdit &, void *); |
| 1175 | template RT_API_ATTRS bool EditRealInput<3>( |
| 1176 | IoStatementState &, const DataEdit &, void *); |
| 1177 | template RT_API_ATTRS bool EditRealInput<4>( |
| 1178 | IoStatementState &, const DataEdit &, void *); |
| 1179 | template RT_API_ATTRS bool EditRealInput<8>( |
| 1180 | IoStatementState &, const DataEdit &, void *); |
| 1181 | template RT_API_ATTRS bool EditRealInput<10>( |
| 1182 | IoStatementState &, const DataEdit &, void *); |
| 1183 | // TODO: double/double |
| 1184 | template RT_API_ATTRS bool EditRealInput<16>( |
| 1185 | IoStatementState &, const DataEdit &, void *); |
| 1186 | |
| 1187 | template RT_API_ATTRS bool EditCharacterInput( |
| 1188 | IoStatementState &, const DataEdit &, char *, std::size_t); |
| 1189 | template RT_API_ATTRS bool EditCharacterInput( |
| 1190 | IoStatementState &, const DataEdit &, char16_t *, std::size_t); |
| 1191 | template RT_API_ATTRS bool EditCharacterInput( |
| 1192 | IoStatementState &, const DataEdit &, char32_t *, std::size_t); |
| 1193 | |
| 1194 | RT_OFFLOAD_API_GROUP_END |
| 1195 | } // namespace Fortran::runtime::io |
| 1196 | |