Willy Tarreau | e11f727 | 2017-05-30 17:49:36 +0200 | [diff] [blame] | 1 | /* |
| 2 | * include/common/ist.h |
| 3 | * Very simple indirect string manipulation functions. |
| 4 | * |
| 5 | * Copyright (C) 2014-2017 Willy Tarreau - w@1wt.eu |
| 6 | * |
| 7 | * Permission is hereby granted, free of charge, to any person obtaining |
| 8 | * a copy of this software and associated documentation files (the |
| 9 | * "Software"), to deal in the Software without restriction, including |
| 10 | * without limitation the rights to use, copy, modify, merge, publish, |
| 11 | * distribute, sublicense, and/or sell copies of the Software, and to |
| 12 | * permit persons to whom the Software is furnished to do so, subject to |
| 13 | * the following conditions: |
| 14 | * |
| 15 | * The above copyright notice and this permission notice shall be |
| 16 | * included in all copies or substantial portions of the Software. |
| 17 | * |
| 18 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
| 19 | * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES |
| 20 | * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| 21 | * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT |
| 22 | * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
| 23 | * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
| 24 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
| 25 | * OTHER DEALINGS IN THE SOFTWARE. |
| 26 | */ |
| 27 | |
| 28 | #ifndef _COMMON_IST_H |
| 29 | #define _COMMON_IST_H |
| 30 | |
Christopher Faulet | 2076145 | 2018-06-06 16:33:53 +0200 | [diff] [blame] | 31 | #include <ctype.h> |
Willy Tarreau | e11f727 | 2017-05-30 17:49:36 +0200 | [diff] [blame] | 32 | #include <string.h> |
Willy Tarreau | a7280a1 | 2018-11-26 19:41:40 +0100 | [diff] [blame] | 33 | #include <unistd.h> |
Willy Tarreau | e11f727 | 2017-05-30 17:49:36 +0200 | [diff] [blame] | 34 | |
| 35 | #include <common/config.h> |
| 36 | |
| 37 | /* This string definition will most often be used to represent a read-only |
| 38 | * string returned from a function, based on the starting point and its length |
| 39 | * in bytes. No storage is provided, only a pointer and a length. The types |
| 40 | * here are important as we only want to have 2 native machine words there so |
| 41 | * that on modern architectures the compiler is capable of efficiently |
| 42 | * returning a register pair without having to allocate stack room from the |
| 43 | * caller. This is done with -freg-struct which is often enabled by default. |
| 44 | */ |
| 45 | struct ist { |
| 46 | char *ptr; |
| 47 | size_t len; |
| 48 | }; |
| 49 | |
Willy Tarreau | 2ba6727 | 2017-09-21 15:24:10 +0200 | [diff] [blame] | 50 | /* makes a constant ist from a constant string, for use in array declarations */ |
| 51 | #define IST(str) { .ptr = str "", .len = (sizeof str "") - 1 } |
| 52 | |
Willy Tarreau | e11f727 | 2017-05-30 17:49:36 +0200 | [diff] [blame] | 53 | /* makes an ist from a regular zero terminated string. Null has length 0. |
| 54 | * Constants are detected and replaced with constant initializers. Other values |
| 55 | * are measured by hand without strlen() as it's much cheaper and inlinable on |
| 56 | * small strings. The construct is complex because we must never call |
| 57 | * __builtin_strlen() with an expression otherwise it involves a real |
| 58 | * measurement. |
| 59 | */ |
| 60 | #if __GNUC__ >= 4 |
| 61 | // gcc >= 4 detects constant propagation of str through __x and resolves the |
| 62 | // length of constant strings easily. |
| 63 | #define ist(str) ({ \ |
| 64 | char *__x = (void *)(str); \ |
| 65 | (struct ist){ \ |
| 66 | .ptr = __x, \ |
| 67 | .len = __builtin_constant_p(str) ? \ |
| 68 | ((void *)str == (void *)0) ? 0 : \ |
| 69 | __builtin_strlen(__x) : \ |
| 70 | ({ \ |
| 71 | size_t __l = 0; \ |
| 72 | if (__x) for (__l--; __x[++__l]; ) ; \ |
| 73 | __l; \ |
| 74 | }) \ |
| 75 | }; \ |
| 76 | }) |
| 77 | #else |
| 78 | // gcc < 4 can't do this, and the side effect is a warning each time a NULL is |
| 79 | // passed to ist() due to the check on __builtin_strlen(). It doesn't have the |
| 80 | // ability to know that this code is never called. |
| 81 | #define ist(str) ({ \ |
| 82 | char *__x = (void *)(str); \ |
| 83 | (struct ist){ \ |
| 84 | .ptr = __x, \ |
| 85 | .len = __builtin_constant_p(str) ? \ |
| 86 | ((void *)str == (void *)0) ? 0 : \ |
| 87 | __builtin_strlen(str) : \ |
| 88 | ({ \ |
| 89 | size_t __l = 0; \ |
| 90 | if (__x) for (__l--; __x[++__l]; ) ; \ |
| 91 | __l; \ |
| 92 | }) \ |
| 93 | }; \ |
| 94 | }) |
| 95 | #endif |
| 96 | |
| 97 | /* makes an ist struct from a string and a length */ |
| 98 | static inline struct ist ist2(const void *ptr, size_t len) |
| 99 | { |
| 100 | return (struct ist){ .ptr = (char *)ptr, .len = len }; |
| 101 | } |
| 102 | |
Willy Tarreau | e67c4e5 | 2017-10-19 06:28:23 +0200 | [diff] [blame] | 103 | /* This function MODIFIES the string to add a zero AFTER the end, and returns |
| 104 | * the start pointer. The purpose is to use it on strings extracted by parsers |
| 105 | * from larger strings cut with delimiters that are not important and can be |
| 106 | * destroyed. It allows any such string to be used with regular string |
| 107 | * functions. It's also convenient to use with printf() to show data extracted |
| 108 | * from writable areas. The caller is obviously responsible for ensuring that |
| 109 | * the string is valid and that the first byte past the end is writable. If |
| 110 | * these conditions cannot be satisfied, use istpad() below instead. |
| 111 | */ |
| 112 | static inline char *ist0(struct ist ist) |
| 113 | { |
| 114 | ist.ptr[ist.len] = 0; |
| 115 | return ist.ptr; |
| 116 | } |
| 117 | |
Willy Tarreau | e11f727 | 2017-05-30 17:49:36 +0200 | [diff] [blame] | 118 | /* returns the length of the string */ |
| 119 | static inline size_t istlen(const struct ist ist) |
| 120 | { |
| 121 | return ist.len; |
| 122 | } |
| 123 | |
| 124 | /* skips to next character in the string, always stops at the end */ |
| 125 | static inline struct ist istnext(const struct ist ist) |
| 126 | { |
| 127 | struct ist ret = ist; |
| 128 | |
| 129 | if (ret.len) { |
| 130 | ret.len--; |
| 131 | ret.ptr++; |
| 132 | } |
| 133 | return ret; |
| 134 | } |
| 135 | |
| 136 | /* copies the contents from string <ist> to buffer <buf> and adds a trailing |
| 137 | * zero. The caller must ensure <buf> is large enough. |
| 138 | */ |
| 139 | static inline struct ist istpad(void *buf, const struct ist ist) |
| 140 | { |
| 141 | struct ist ret = { .ptr = buf, .len = ist.len }; |
| 142 | |
| 143 | for (ret.len = 0; ret.len < ist.len; ret.len++) |
| 144 | ret.ptr[ret.len] = ist.ptr[ret.len]; |
| 145 | |
| 146 | ret.ptr[ret.len] = 0; |
| 147 | return ret; |
| 148 | } |
| 149 | |
| 150 | /* trims string <ist> to no more than <size> characters. The string is |
| 151 | * returned. |
| 152 | */ |
| 153 | static inline struct ist isttrim(const struct ist ist, size_t size) |
| 154 | { |
| 155 | struct ist ret = ist; |
| 156 | |
| 157 | if (ret.len > size) |
| 158 | ret.len = size; |
| 159 | return ret; |
| 160 | } |
| 161 | |
| 162 | /* trims string <ist> to no more than <size>-1 characters and ensures that a |
| 163 | * zero is placed after <ist.len> (possibly reduced by one) and before <size>, |
| 164 | * unless <size> is already zero. The string is returned. This is mostly aimed |
| 165 | * at building printable strings that need to be zero-terminated. |
| 166 | */ |
| 167 | static inline struct ist istzero(const struct ist ist, size_t size) |
| 168 | { |
| 169 | struct ist ret = ist; |
| 170 | |
| 171 | if (!size) |
| 172 | ret.len = 0; |
| 173 | else { |
| 174 | if (ret.len > size - 1) |
| 175 | ret.len = size - 1; |
| 176 | ret.ptr[ret.len] = 0; |
| 177 | } |
| 178 | return ret; |
| 179 | } |
| 180 | |
| 181 | /* returns the ordinal difference between two strings : |
| 182 | * < 0 if ist1 < ist2 |
| 183 | * = 0 if ist1 == ist2 |
| 184 | * > 0 if ist1 > ist2 |
| 185 | */ |
| 186 | static inline int istdiff(const struct ist ist1, const struct ist ist2) |
| 187 | { |
| 188 | struct ist l = ist1; |
| 189 | struct ist r = ist2; |
| 190 | |
| 191 | do { |
| 192 | if (!l.len--) |
| 193 | return -r.len; |
| 194 | if (!r.len--) |
| 195 | return 1; |
| 196 | } while (*l.ptr++ == *r.ptr++); |
| 197 | |
| 198 | return *(unsigned char *)(l.ptr - 1) - *(unsigned char *)(r.ptr - 1); |
| 199 | } |
| 200 | |
| 201 | /* returns non-zero if <ist1> starts like <ist2> (empty strings do match) */ |
| 202 | static inline int istmatch(const struct ist ist1, const struct ist ist2) |
| 203 | { |
| 204 | struct ist l = ist1; |
| 205 | struct ist r = ist2; |
| 206 | |
| 207 | if (l.len < r.len) |
| 208 | return 0; |
| 209 | |
| 210 | while (r.len--) { |
| 211 | if (*l.ptr++ != *r.ptr++) |
| 212 | return 0; |
| 213 | } |
| 214 | return 1; |
| 215 | } |
| 216 | |
| 217 | /* returns non-zero if <ist1> starts like <ist2> on the first <count> |
| 218 | * characters (empty strings do match). |
| 219 | */ |
| 220 | static inline int istnmatch(const struct ist ist1, const struct ist ist2, size_t count) |
| 221 | { |
| 222 | struct ist l = ist1; |
| 223 | struct ist r = ist2; |
| 224 | |
| 225 | if (l.len > count) |
| 226 | l.len = count; |
| 227 | if (r.len > count) |
| 228 | r.len = count; |
| 229 | return istmatch(l, r); |
| 230 | } |
| 231 | |
| 232 | /* returns non-zero if <ist1> equals <ist2> (empty strings are equal) */ |
| 233 | static inline int isteq(const struct ist ist1, const struct ist ist2) |
| 234 | { |
| 235 | struct ist l = ist1; |
| 236 | struct ist r = ist2; |
| 237 | |
| 238 | if (l.len != r.len) |
| 239 | return 0; |
| 240 | |
| 241 | while (l.len--) { |
| 242 | if (*l.ptr++ != *r.ptr++) |
| 243 | return 0; |
| 244 | } |
| 245 | return 1; |
| 246 | } |
| 247 | |
Christopher Faulet | 2076145 | 2018-06-06 16:33:53 +0200 | [diff] [blame] | 248 | /* returns non-zero if <ist1> equals <ist2>, ignoring the case (empty strings are equal) */ |
| 249 | static inline int isteqi(const struct ist ist1, const struct ist ist2) |
| 250 | { |
| 251 | struct ist l = ist1; |
| 252 | struct ist r = ist2; |
| 253 | |
| 254 | if (l.len != r.len) |
| 255 | return 0; |
| 256 | |
| 257 | while (l.len--) { |
| 258 | if (tolower(*l.ptr) != tolower(*r.ptr)) |
| 259 | return 0; |
| 260 | l.ptr++; |
| 261 | r.ptr++; |
| 262 | } |
| 263 | return 1; |
| 264 | } |
| 265 | |
Willy Tarreau | e11f727 | 2017-05-30 17:49:36 +0200 | [diff] [blame] | 266 | /* returns non-zero if <ist1> equals <ist2> on the first <count> characters |
| 267 | * (empty strings are equal). |
| 268 | */ |
| 269 | static inline int istneq(const struct ist ist1, const struct ist ist2, size_t count) |
| 270 | { |
| 271 | struct ist l = ist1; |
| 272 | struct ist r = ist2; |
| 273 | |
| 274 | if (l.len > count) |
| 275 | l.len = count; |
| 276 | if (r.len > count) |
| 277 | r.len = count; |
| 278 | return isteq(l, r); |
| 279 | } |
| 280 | |
| 281 | /* copies <src> over <dst> for a maximum of <count> bytes. Returns the number |
| 282 | * of characters copied (src.len), or -1 if it does not fit. In all cases, the |
| 283 | * contents are copied prior to reporting an error, so that the destination |
| 284 | * at least contains a valid but truncated string. |
| 285 | */ |
| 286 | static inline ssize_t istcpy(struct ist *dst, const struct ist src, size_t count) |
| 287 | { |
| 288 | dst->len = 0; |
| 289 | |
| 290 | if (count > src.len) |
| 291 | count = src.len; |
| 292 | |
| 293 | while (dst->len < count) { |
| 294 | dst->ptr[dst->len] = src.ptr[dst->len]; |
| 295 | dst->len++; |
| 296 | } |
| 297 | |
| 298 | if (dst->len == src.len) |
| 299 | return src.len; |
| 300 | |
| 301 | return -1; |
| 302 | } |
| 303 | |
| 304 | /* copies <src> over <dst> for a maximum of <count> bytes. Returns the number |
| 305 | * of characters copied, or -1 if it does not fit. A (possibly truncated) valid |
| 306 | * copy of <src> is always left into <dst>, and a trailing \0 is appended as |
| 307 | * long as <count> is not null, even if that results in reducing the string by |
| 308 | * one character. |
| 309 | */ |
| 310 | static inline ssize_t istscpy(struct ist *dst, const struct ist src, size_t count) |
| 311 | { |
| 312 | dst->len = 0; |
| 313 | |
| 314 | if (!count) |
| 315 | goto fail; |
| 316 | |
| 317 | if (count > src.len) |
| 318 | count = src.len + 1; |
| 319 | |
| 320 | while (dst->len < count - 1) { |
| 321 | dst->ptr[dst->len] = src.ptr[dst->len]; |
| 322 | dst->len++; |
| 323 | } |
| 324 | |
| 325 | dst->ptr[dst->len] = 0; |
| 326 | if (dst->len == src.len) |
| 327 | return src.len; |
| 328 | fail: |
| 329 | return -1; |
| 330 | } |
| 331 | |
| 332 | /* appends <src> after <dst> for a maximum of <count> total bytes in <dst> after |
| 333 | * the copy. <dst> is assumed to be <count> or less before the call. The new |
| 334 | * string's length is returned, or -1 if a truncation happened. In all cases, |
| 335 | * the contents are copied prior to reporting an error, so that the destination |
| 336 | * at least contains a valid but truncated string. |
| 337 | */ |
| 338 | static inline ssize_t istcat(struct ist *dst, const struct ist src, size_t count) |
| 339 | { |
| 340 | const char *s = src.ptr; |
| 341 | |
| 342 | while (dst->len < count && s != src.ptr + src.len) |
| 343 | dst->ptr[dst->len++] = *s++; |
| 344 | |
| 345 | if (s == src.ptr + src.len) |
| 346 | return dst->len; |
| 347 | |
| 348 | return -1; |
| 349 | } |
| 350 | |
| 351 | /* appends <src> after <dst> for a maximum of <count> total bytes in <dst> after |
| 352 | * the copy. <dst> is assumed to be <count> or less before the call. The new |
| 353 | * string's length is returned, or -1 if a truncation happened. In all cases, |
| 354 | * the contents are copied prior to reporting an error, so that the destination |
| 355 | * at least contains a valid but truncated string. |
| 356 | */ |
| 357 | static inline ssize_t istscat(struct ist *dst, const struct ist src, size_t count) |
| 358 | { |
| 359 | const char *s = src.ptr; |
| 360 | |
| 361 | if (!count) |
| 362 | goto fail; |
| 363 | |
| 364 | while (dst->len < count - 1 && s != src.ptr + src.len) { |
| 365 | dst->ptr[dst->len++] = *s++; |
| 366 | } |
| 367 | |
| 368 | dst->ptr[dst->len] = 0; |
| 369 | if (s == src.ptr + src.len) |
| 370 | return dst->len; |
| 371 | fail: |
| 372 | return -1; |
| 373 | } |
| 374 | |
| 375 | /* looks for first occurrence of character <chr> in string <ist>. Returns the |
| 376 | * pointer if found, or NULL if not found. |
| 377 | */ |
| 378 | static inline char *istchr(const struct ist ist, char chr) |
| 379 | { |
| 380 | char *s = ist.ptr; |
| 381 | |
| 382 | do { |
| 383 | if (s >= ist.ptr + ist.len) |
| 384 | return NULL; |
| 385 | } while (*s++ != chr); |
| 386 | return s - 1; |
| 387 | } |
| 388 | |
| 389 | /* looks for first occurrence of character <chr> in string <ist> and returns |
| 390 | * the tail of the string starting with this character, or (ist.end,0) if not |
| 391 | * found. |
| 392 | */ |
| 393 | static inline struct ist istfind(const struct ist ist, char chr) |
| 394 | { |
| 395 | struct ist ret = ist; |
| 396 | |
| 397 | while (ret.len--) { |
| 398 | if (*ret.ptr++ == chr) |
| 399 | return ist2(ret.ptr - 1, ret.len + 1); |
| 400 | } |
| 401 | return ist2(ret.ptr, 0); |
| 402 | } |
| 403 | |
| 404 | /* looks for first occurrence of character different from <chr> in string <ist> |
| 405 | * and returns the tail of the string starting at this character, or (ist_end,0) |
| 406 | * if not found. |
| 407 | */ |
| 408 | static inline struct ist istskip(const struct ist ist, char chr) |
| 409 | { |
| 410 | struct ist ret = ist; |
| 411 | |
| 412 | while (ret.len--) { |
| 413 | if (*ret.ptr++ != chr) |
| 414 | return ist2(ret.ptr - 1, ret.len + 1); |
| 415 | } |
| 416 | return ist2(ret.ptr, 0); |
| 417 | } |
| 418 | |
| 419 | /* looks for first occurrence of string <pat> in string <ist> and returns the |
| 420 | * tail of the string starting at this position, or (NULL,0) if not found. The |
| 421 | * empty pattern is found everywhere. |
| 422 | */ |
| 423 | static inline struct ist istist(const struct ist ist, const struct ist pat) |
| 424 | { |
| 425 | struct ist ret = ist; |
| 426 | size_t pos; |
| 427 | |
| 428 | if (!pat.len) |
| 429 | return ret; |
| 430 | |
| 431 | while (1) { |
| 432 | loop: |
| 433 | ret = istfind(ret, *pat.ptr); |
| 434 | if (ret.len < pat.len) |
| 435 | break; |
| 436 | |
| 437 | /* ret.len >= 1, pat.len >= 1 and *ret.ptr == *pat.ptr */ |
| 438 | |
| 439 | ret = istnext(ret); |
| 440 | for (pos = 0; pos < pat.len - 1; ) { |
| 441 | ++pos; |
| 442 | if (ret.ptr[pos - 1] != pat.ptr[pos]) |
| 443 | goto loop; |
| 444 | } |
| 445 | return ist2(ret.ptr - 1, ret.len + 1); |
| 446 | } |
| 447 | return ist2(NULL, 0); |
| 448 | } |
| 449 | |
| 450 | #endif |