| 1 | #include <errno.h> // errno
|
| 2 | #include <float.h> // DBL_MIN, DBL_MAX
|
| 3 | #include <math.h> // INFINITY
|
| 4 | #include <stdio.h> // required for readline/readline.h (man readline)
|
| 5 |
|
| 6 | #include "_build/detected-cpp-config.h"
|
| 7 | #include "mycpp/runtime.h"
|
| 8 | #ifdef HAVE_READLINE
|
| 9 | #include "cpp/frontend_pyreadline.h"
|
| 10 | #endif
|
| 11 |
|
| 12 | // Translation of Python's print().
|
| 13 | void print(BigStr* s) {
|
| 14 | fputs(s->data_, stdout); // print until first NUL
|
| 15 | fputc('\n', stdout);
|
| 16 | }
|
| 17 |
|
| 18 | BigStr* str(int i) {
|
| 19 | BigStr* s = OverAllocatedStr(kIntBufSize);
|
| 20 | int length = snprintf(s->data(), kIntBufSize, "%d", i);
|
| 21 | s->MaybeShrink(length);
|
| 22 | return s;
|
| 23 | }
|
| 24 |
|
| 25 | // TODO:
|
| 26 | // - Does libc depend on locale?
|
| 27 | BigStr* str(double d) {
|
| 28 | char buf[64]; // overestimate, but we use snprintf() to be safe
|
| 29 |
|
| 30 | // Problem:
|
| 31 | // %f prints 3.0000000 and 3.500000
|
| 32 | // %g prints 3 and 3.5
|
| 33 | //
|
| 34 | // We want 3.0 and 3.5, so add '.0' in some cases
|
| 35 |
|
| 36 | int n = sizeof(buf) - 2; // in case we add '.0'
|
| 37 |
|
| 38 | // %.9g digits for string that can be converted back to the same FLOAT
|
| 39 | // (not double)
|
| 40 | //
|
| 41 | // See mycpp/float_test.cc for round-tripping test
|
| 42 | //
|
| 43 | // https://stackoverflow.com/a/21162120
|
| 44 | // https://en.cppreference.com/w/cpp/types/numeric_limits/max_digits10
|
| 45 | int length = snprintf(buf, n, "%.9g", d);
|
| 46 |
|
| 47 | // %a is a hexfloat form, could use that somewhere
|
| 48 | // int length = snprintf(buf, n, "%a", d);
|
| 49 |
|
| 50 | if (strchr(buf, 'i') || strchr(buf, 'n')) { // inf, -inf, nan
|
| 51 | return StrFromC(buf);
|
| 52 | }
|
| 53 |
|
| 54 | if (!strchr(buf, '.')) { // 12345 -> 12345.0
|
| 55 | buf[length] = '.';
|
| 56 | buf[length + 1] = '0';
|
| 57 | buf[length + 2] = '\0';
|
| 58 | }
|
| 59 |
|
| 60 | return StrFromC(buf);
|
| 61 | }
|
| 62 |
|
| 63 | // Do we need this API? Or is mylib.InternedStr(BigStr* s, int start, int end)
|
| 64 | // better for getting values out of Token.line without allocating?
|
| 65 | //
|
| 66 | // e.g. mylib.InternedStr(tok.line, tok.start, tok.start+1)
|
| 67 | //
|
| 68 | // Also for SmallStr, we don't care about interning. Only for HeapStr.
|
| 69 |
|
| 70 | BigStr* intern(BigStr* s) {
|
| 71 | // TODO: put in table gHeap.interned_
|
| 72 | return s;
|
| 73 | }
|
| 74 |
|
| 75 | // Print quoted string. Called by StrFormat('%r').
|
| 76 | // TODO: consider using J8 notation instead, since error messages show that
|
| 77 | // string.
|
| 78 | BigStr* repr(BigStr* s) {
|
| 79 | // Worst case: \0 becomes 4 bytes as '\\x00', and then two quote bytes.
|
| 80 | int n = len(s);
|
| 81 | int upper_bound = n * 4 + 2;
|
| 82 |
|
| 83 | BigStr* result = OverAllocatedStr(upper_bound);
|
| 84 |
|
| 85 | // Single quote by default.
|
| 86 | char quote = '\'';
|
| 87 | if (memchr(s->data_, '\'', n) && !memchr(s->data_, '"', n)) {
|
| 88 | quote = '"';
|
| 89 | }
|
| 90 | char* p = result->data_;
|
| 91 |
|
| 92 | // From PyString_Repr()
|
| 93 | *p++ = quote;
|
| 94 | for (int i = 0; i < n; ++i) {
|
| 95 | unsigned char c = static_cast<unsigned char>(s->data_[i]);
|
| 96 | if (c == quote || c == '\\') {
|
| 97 | *p++ = '\\';
|
| 98 | *p++ = c;
|
| 99 | } else if (c == '\t') {
|
| 100 | *p++ = '\\';
|
| 101 | *p++ = 't';
|
| 102 | } else if (c == '\n') {
|
| 103 | *p++ = '\\';
|
| 104 | *p++ = 'n';
|
| 105 | } else if (c == '\r') {
|
| 106 | *p++ = '\\';
|
| 107 | *p++ = 'r';
|
| 108 | } else if (0x20 <= c && c < 0x80) {
|
| 109 | *p++ = c;
|
| 110 | } else {
|
| 111 | // Unprintable becomes \xff.
|
| 112 | // TODO: Consider \yff. This is similar to J8 strings, but we don't
|
| 113 | // decode UTF-8.
|
| 114 | sprintf(p, "\\x%02x", c & 0xff);
|
| 115 | p += 4;
|
| 116 | }
|
| 117 | }
|
| 118 | *p++ = quote;
|
| 119 | *p = '\0';
|
| 120 |
|
| 121 | int length = p - result->data_;
|
| 122 | result->MaybeShrink(length);
|
| 123 | return result;
|
| 124 | }
|
| 125 |
|
| 126 | // Helper functions that don't use exceptions.
|
| 127 |
|
| 128 | bool StringToInt(const char* s, int length, int base, int* result) {
|
| 129 | if (length == 0) {
|
| 130 | return false; // empty string isn't a valid integer
|
| 131 | }
|
| 132 |
|
| 133 | // Note: sizeof(int) is often 4 bytes on both 32-bit and 64-bit
|
| 134 | // sizeof(long) is often 4 bytes on both 32-bit but 8 bytes on 64-bit
|
| 135 | // static_assert(sizeof(long) == 8);
|
| 136 |
|
| 137 | char* pos; // mutated by strtol
|
| 138 |
|
| 139 | errno = 0;
|
| 140 | long v = strtol(s, &pos, base);
|
| 141 |
|
| 142 | if (errno == ERANGE) {
|
| 143 | switch (v) {
|
| 144 | case LONG_MIN:
|
| 145 | return false; // underflow of long, which may be 64 bits
|
| 146 | case LONG_MAX:
|
| 147 | return false; // overflow of long
|
| 148 | }
|
| 149 | }
|
| 150 |
|
| 151 | // It should ALSO fit in an int, not just a long
|
| 152 | if (v > INT_MAX) {
|
| 153 | return false;
|
| 154 | }
|
| 155 | if (v < INT_MIN) {
|
| 156 | return false;
|
| 157 | }
|
| 158 |
|
| 159 | const char* end = s + length;
|
| 160 | if (pos == end) {
|
| 161 | *result = v;
|
| 162 | return true; // strtol() consumed ALL characters.
|
| 163 | }
|
| 164 |
|
| 165 | while (pos < end) {
|
| 166 | if (!IsAsciiWhitespace(*pos)) {
|
| 167 | return false; // Trailing non-space
|
| 168 | }
|
| 169 | pos++;
|
| 170 | }
|
| 171 |
|
| 172 | *result = v;
|
| 173 | return true; // Trailing space is OK
|
| 174 | }
|
| 175 |
|
| 176 | bool StringToInt64(const char* s, int length, int base, int64_t* result) {
|
| 177 | if (length == 0) {
|
| 178 | return false; // empty string isn't a valid integer
|
| 179 | }
|
| 180 |
|
| 181 | // These should be the same type
|
| 182 | static_assert(sizeof(long long) == sizeof(int64_t));
|
| 183 |
|
| 184 | char* pos; // mutated by strtol
|
| 185 |
|
| 186 | errno = 0;
|
| 187 | long long v = strtoll(s, &pos, base);
|
| 188 |
|
| 189 | if (errno == ERANGE) {
|
| 190 | switch (v) {
|
| 191 | case LLONG_MIN:
|
| 192 | return false; // underflow
|
| 193 | case LLONG_MAX:
|
| 194 | return false; // overflow
|
| 195 | }
|
| 196 | }
|
| 197 |
|
| 198 | const char* end = s + length;
|
| 199 | if (pos == end) {
|
| 200 | *result = v;
|
| 201 | return true; // strtol() consumed ALL characters.
|
| 202 | }
|
| 203 |
|
| 204 | while (pos < end) {
|
| 205 | if (!IsAsciiWhitespace(*pos)) {
|
| 206 | return false; // Trailing non-space
|
| 207 | }
|
| 208 | pos++;
|
| 209 | }
|
| 210 |
|
| 211 | *result = v;
|
| 212 | return true; // Trailing space is OK
|
| 213 | }
|
| 214 |
|
| 215 | int to_int(BigStr* s, int base) {
|
| 216 | int i;
|
| 217 | if (StringToInt(s->data_, len(s), base, &i)) {
|
| 218 | return i; // truncated to int
|
| 219 | } else {
|
| 220 | throw Alloc<ValueError>();
|
| 221 | }
|
| 222 | }
|
| 223 |
|
| 224 | BigStr* chr(int i) {
|
| 225 | // NOTE: i should be less than 256, in which we could return an object from
|
| 226 | // GLOBAL_STR() pool, like StrIter
|
| 227 | auto result = NewStr(1);
|
| 228 | result->data_[0] = i;
|
| 229 | return result;
|
| 230 | }
|
| 231 |
|
| 232 | int ord(BigStr* s) {
|
| 233 | assert(len(s) == 1);
|
| 234 | // signed to unsigned conversion, so we don't get values like -127
|
| 235 | uint8_t c = static_cast<uint8_t>(s->data_[0]);
|
| 236 | return c;
|
| 237 | }
|
| 238 |
|
| 239 | bool to_bool(BigStr* s) {
|
| 240 | return len(s) != 0;
|
| 241 | }
|
| 242 |
|
| 243 | double to_float(int i) {
|
| 244 | return static_cast<double>(i);
|
| 245 | }
|
| 246 |
|
| 247 | double to_float(BigStr* s) {
|
| 248 | char* begin = s->data_;
|
| 249 | char* end = begin + len(s);
|
| 250 |
|
| 251 | errno = 0;
|
| 252 | double result = strtod(begin, &end);
|
| 253 |
|
| 254 | if (errno == ERANGE) { // error: overflow or underflow
|
| 255 | if (result >= HUGE_VAL) {
|
| 256 | return INFINITY;
|
| 257 | } else if (result <= -HUGE_VAL) {
|
| 258 | return -INFINITY;
|
| 259 | } else if (-DBL_MIN <= result && result <= DBL_MIN) {
|
| 260 | return 0.0;
|
| 261 | } else {
|
| 262 | FAIL("Invalid value after ERANGE");
|
| 263 | }
|
| 264 | }
|
| 265 | if (end == begin) { // error: not a floating point number
|
| 266 | throw Alloc<ValueError>();
|
| 267 | }
|
| 268 |
|
| 269 | return result;
|
| 270 | }
|
| 271 |
|
| 272 | // e.g. ('a' in 'abc')
|
| 273 | bool str_contains(BigStr* haystack, BigStr* needle) {
|
| 274 | // Common case
|
| 275 | if (len(needle) == 1) {
|
| 276 | return memchr(haystack->data_, needle->data_[0], len(haystack));
|
| 277 | }
|
| 278 |
|
| 279 | if (len(needle) > len(haystack)) {
|
| 280 | return false;
|
| 281 | }
|
| 282 |
|
| 283 | // General case. TODO: We could use a smarter substring algorithm.
|
| 284 |
|
| 285 | const char* end = haystack->data_ + len(haystack);
|
| 286 | const char* last_possible = end - len(needle);
|
| 287 | const char* p = haystack->data_;
|
| 288 |
|
| 289 | while (p <= last_possible) {
|
| 290 | if (memcmp(p, needle->data_, len(needle)) == 0) {
|
| 291 | return true;
|
| 292 | }
|
| 293 | p++;
|
| 294 | }
|
| 295 | return false;
|
| 296 | }
|
| 297 |
|
| 298 | BigStr* str_repeat(BigStr* s, int times) {
|
| 299 | // Python allows -1 too, and Oil used that
|
| 300 | if (times <= 0) {
|
| 301 | return kEmptyString;
|
| 302 | }
|
| 303 | int len_ = len(s);
|
| 304 | int new_len = len_ * times;
|
| 305 | BigStr* result = NewStr(new_len);
|
| 306 |
|
| 307 | char* dest = result->data_;
|
| 308 | for (int i = 0; i < times; i++) {
|
| 309 | memcpy(dest, s->data_, len_);
|
| 310 | dest += len_;
|
| 311 | }
|
| 312 | return result;
|
| 313 | }
|
| 314 |
|
| 315 | // for os_path.join()
|
| 316 | // NOTE(Jesse): Perfect candidate for BoundedBuffer
|
| 317 | BigStr* str_concat3(BigStr* a, BigStr* b, BigStr* c) {
|
| 318 | int a_len = len(a);
|
| 319 | int b_len = len(b);
|
| 320 | int c_len = len(c);
|
| 321 |
|
| 322 | int new_len = a_len + b_len + c_len;
|
| 323 | BigStr* result = NewStr(new_len);
|
| 324 | char* pos = result->data_;
|
| 325 |
|
| 326 | memcpy(pos, a->data_, a_len);
|
| 327 | pos += a_len;
|
| 328 |
|
| 329 | memcpy(pos, b->data_, b_len);
|
| 330 | pos += b_len;
|
| 331 |
|
| 332 | memcpy(pos, c->data_, c_len);
|
| 333 |
|
| 334 | assert(pos + c_len == result->data_ + new_len);
|
| 335 |
|
| 336 | return result;
|
| 337 | }
|
| 338 |
|
| 339 | BigStr* str_concat(BigStr* a, BigStr* b) {
|
| 340 | int a_len = len(a);
|
| 341 | int b_len = len(b);
|
| 342 | int new_len = a_len + b_len;
|
| 343 | BigStr* result = NewStr(new_len);
|
| 344 | char* buf = result->data_;
|
| 345 |
|
| 346 | memcpy(buf, a->data_, a_len);
|
| 347 | memcpy(buf + a_len, b->data_, b_len);
|
| 348 |
|
| 349 | return result;
|
| 350 | }
|
| 351 |
|
| 352 | //
|
| 353 | // Comparators
|
| 354 | //
|
| 355 |
|
| 356 | bool str_equals(BigStr* left, BigStr* right) {
|
| 357 | // Fast path for identical strings. String deduplication during GC could
|
| 358 | // make this more likely. String interning could guarantee it, allowing us
|
| 359 | // to remove memcmp().
|
| 360 | if (left == right) {
|
| 361 | return true;
|
| 362 | }
|
| 363 |
|
| 364 | // TODO: It would be nice to remove this condition, but I think we need MyPy
|
| 365 | // strict None checking for it
|
| 366 | if (left == nullptr || right == nullptr) {
|
| 367 | return false;
|
| 368 | }
|
| 369 |
|
| 370 | if (left->len_ != right->len_) {
|
| 371 | return false;
|
| 372 | }
|
| 373 |
|
| 374 | return memcmp(left->data_, right->data_, left->len_) == 0;
|
| 375 | }
|
| 376 |
|
| 377 | bool maybe_str_equals(BigStr* left, BigStr* right) {
|
| 378 | if (left && right) {
|
| 379 | return str_equals(left, right);
|
| 380 | }
|
| 381 |
|
| 382 | if (!left && !right) {
|
| 383 | return true; // None == None
|
| 384 | }
|
| 385 |
|
| 386 | return false; // one is None and one is a BigStr*
|
| 387 | }
|
| 388 |
|
| 389 | bool items_equal(BigStr* left, BigStr* right) {
|
| 390 | return str_equals(left, right);
|
| 391 | }
|
| 392 |
|
| 393 | bool keys_equal(BigStr* left, BigStr* right) {
|
| 394 | return items_equal(left, right);
|
| 395 | }
|
| 396 |
|
| 397 | bool items_equal(Tuple2<int, int>* t1, Tuple2<int, int>* t2) {
|
| 398 | return (t1->at0() == t2->at0()) && (t1->at1() == t2->at1());
|
| 399 | }
|
| 400 |
|
| 401 | bool keys_equal(Tuple2<int, int>* t1, Tuple2<int, int>* t2) {
|
| 402 | return items_equal(t1, t2);
|
| 403 | }
|
| 404 |
|
| 405 | bool items_equal(Tuple2<BigStr*, int>* t1, Tuple2<BigStr*, int>* t2) {
|
| 406 | return items_equal(t1->at0(), t2->at0()) && (t1->at1() == t2->at1());
|
| 407 | }
|
| 408 |
|
| 409 | bool keys_equal(Tuple2<BigStr*, int>* t1, Tuple2<BigStr*, int>* t2) {
|
| 410 | return items_equal(t1, t2);
|
| 411 | }
|
| 412 |
|
| 413 | bool str_equals_c(BigStr* s, const char* c_string, int c_len) {
|
| 414 | // Needs SmallStr change
|
| 415 | if (len(s) == c_len) {
|
| 416 | return memcmp(s->data_, c_string, c_len) == 0;
|
| 417 | } else {
|
| 418 | return false;
|
| 419 | }
|
| 420 | }
|
| 421 |
|
| 422 | bool str_equals0(const char* c_string, BigStr* s) {
|
| 423 | int n = strlen(c_string);
|
| 424 | if (len(s) == n) {
|
| 425 | return memcmp(s->data_, c_string, n) == 0;
|
| 426 | } else {
|
| 427 | return false;
|
| 428 | }
|
| 429 | }
|
| 430 |
|
| 431 | int hash(BigStr* s) {
|
| 432 | return s->hash(fnv1);
|
| 433 | }
|
| 434 |
|
| 435 | int max(int a, int b) {
|
| 436 | return std::max(a, b);
|
| 437 | }
|
| 438 |
|
| 439 | int min(int a, int b) {
|
| 440 | return std::min(a, b);
|
| 441 | }
|
| 442 |
|
| 443 | int max(List<int>* elems) {
|
| 444 | int n = len(elems);
|
| 445 | if (n < 1) {
|
| 446 | throw Alloc<ValueError>();
|
| 447 | }
|
| 448 |
|
| 449 | int ret = elems->at(0);
|
| 450 | for (int i = 0; i < n; ++i) {
|
| 451 | int cand = elems->at(i);
|
| 452 | if (cand > ret) {
|
| 453 | ret = cand;
|
| 454 | }
|
| 455 | }
|
| 456 |
|
| 457 | return ret;
|
| 458 | }
|