diff options
Diffstat (limited to 'strings/ctype-uca.ic')
| -rw-r--r-- | strings/ctype-uca.ic | 276 |
1 files changed, 174 insertions, 102 deletions
diff --git a/strings/ctype-uca.ic b/strings/ctype-uca.ic index cee12cf4d7b..1fc3480e5b5 100644 --- a/strings/ctype-uca.ic +++ b/strings/ctype-uca.ic @@ -36,108 +36,9 @@ #error MY_UCA_COLL_INIT is not defined #endif - -static inline int -MY_FUNCTION_NAME(scanner_next)(my_uca_scanner *scanner) -{ - /* - Check if the weights for the previous character have been - already fully scanned. If yes, then get the next character and - initialize wbeg and wlength to its weight string. - */ - - if (scanner->wbeg[0]) /* More weights left from the previous step: */ - return *scanner->wbeg++; /* return the next weight from expansion */ - - do - { - const uint16 *wpage; - my_wc_t wc[MY_UCA_MAX_CONTRACTION]; - int mblen; - - /* Get next character */ -#if MY_UCA_ASCII_OPTIMIZE - /* Get next ASCII character */ - if (scanner->sbeg < scanner->send && scanner->sbeg[0] < 0x80) - { - wc[0]= scanner->sbeg[0]; - scanner->sbeg+= 1; - -#if MY_UCA_COMPILE_CONTRACTIONS - if (my_uca_needs_context_handling(scanner->level, wc[0])) - { - uint16 *cweight= my_uca_context_weight_find(scanner, wc); - if (cweight) - return *cweight; - } -#endif - - scanner->page= 0; - scanner->code= (int) wc[0]; - scanner->wbeg= scanner->level->weights[0] + scanner->code * scanner->level->lengths[0]; - if (scanner->wbeg[0]) - return *scanner->wbeg++; - continue; - } - else -#endif - /* Get next MB character */ - if (((mblen= MY_MB_WC(scanner, wc, scanner->sbeg, - scanner->send)) <= 0)) - { - if (scanner->sbeg >= scanner->send) - return -1; /* No more bytes, end of line reached */ - /* - There are some more bytes left. Non-positive mb_len means that - we got an incomplete or a bad byte sequence. Consume mbminlen bytes. - */ - if ((scanner->sbeg+= scanner->cs->mbminlen) > scanner->send) - { - /* For safety purposes don't go beyond the string range. */ - scanner->sbeg= scanner->send; - } - /* - Treat every complete or incomplete mbminlen unit as a weight which is - greater than weight for any possible normal character. - 0xFFFF is greater than any possible weight in the UCA weight table. - */ - return 0xFFFF; - } - - scanner->sbeg+= mblen; - if (wc[0] > scanner->level->maxchar) - { - /* Return 0xFFFD as weight for all characters outside BMP */ - scanner->wbeg= nochar; - return 0xFFFD; - } - -#if MY_UCA_COMPILE_CONTRACTIONS - if (my_uca_needs_context_handling(scanner->level, wc[0])) - { - uint16 *cweight= my_uca_context_weight_find(scanner, wc); - if (cweight) - return *cweight; - } -#endif - - /* Process single character */ - scanner->page= wc[0] >> 8; - scanner->code= wc[0] & 0xFF; - - /* If weight page for w[0] does not exist, then calculate algoritmically */ - if (!(wpage= scanner->level->weights[scanner->page])) - return my_uca_scanner_next_implicit(scanner); - - /* Calculate pointer to w[0]'s weight, using page and offset */ - scanner->wbeg= wpage + - scanner->code * scanner->level->lengths[scanner->page]; - } while (!scanner->wbeg[0]); /* Skip ignorable characters */ - - return *scanner->wbeg++; -} - - +#include "ctype-uca-scanner_next.inl" +#define SCANNER_NEXT_NCHARS +#include "ctype-uca-scanner_next.inl" /* Compares two strings according to the collation @@ -410,6 +311,173 @@ MY_FUNCTION_NAME(strnncollsp_nopad_multilevel)(CHARSET_INFO *cs, } +/* + Scan the next weight and perform space padding + or trimming according to "nchars". +*/ +static inline weight_and_nchars_t +MY_FUNCTION_NAME(scanner_next_pad_trim)(my_uca_scanner *scanner, + size_t nchars, + uint *generated) +{ + weight_and_nchars_t res; + if (nchars > 0 || + scanner->wbeg[0] /* Some weights from a previous expansion left */) + { + if ((res= MY_FUNCTION_NAME(scanner_next_with_nchars)(scanner, + nchars)).weight < 0) + { + /* + We reached the end of the string, but the caller wants more weights. + Perform space padding. + */ + res.weight= my_space_weight(scanner->level); + res.nchars= 1; + (*generated)++; + } + else if (res.nchars > nchars) + { + /* + We scanned the next collation element, but it does not fit into + the "nchars" limit. This is possible in case of: + - A contraction, e.g. Czech 'ch' with nchars=1 + - A sequence of ignorable characters followed by non-ignorable ones, + e.g. CONCAT(x'00','a') with nchars=1. + Perform trimming. + */ + res.weight= scanner->cs->state & MY_CS_NOPAD ? + 0 : my_space_weight(scanner->level); + res.nchars= (uint) nchars; + (*generated)++; + } + } + else + { + /* The caller wants nchars==0. Perform trimming. */ + res.weight= scanner->cs->state & MY_CS_NOPAD ? + 0 : my_space_weight(scanner->level); + res.nchars= 0; + (*generated)++; + } + return res; +} + + +static int +MY_FUNCTION_NAME(strnncollsp_nchars_onelevel)(CHARSET_INFO *cs, + const MY_UCA_WEIGHT_LEVEL *level, + const uchar *s, size_t slen, + const uchar *t, size_t tlen, + size_t nchars) +{ + my_uca_scanner sscanner; + my_uca_scanner tscanner; + size_t s_nchars_left= nchars; + size_t t_nchars_left= nchars; + + my_uca_scanner_init_any(&sscanner, cs, level, s, slen); + my_uca_scanner_init_any(&tscanner, cs, level, t, tlen); + + for ( ; ; ) + { + weight_and_nchars_t s_res; + weight_and_nchars_t t_res; + uint generated= 0; + int diff; + + s_res= MY_FUNCTION_NAME(scanner_next_pad_trim)(&sscanner, s_nchars_left, + &generated); + t_res= MY_FUNCTION_NAME(scanner_next_pad_trim)(&tscanner, t_nchars_left, + &generated); + if ((diff= (s_res.weight - t_res.weight))) + return diff; + + if (generated == 2) + { + if (cs->state & MY_CS_NOPAD) + { + /* + Both values are auto-generated. There's no real data any more. + We need to handle the remaining virtual trailing spaces. + The two strings still have s_nchars_left and t_nchars_left imaginary + trailing spaces at the end. If s_nchars_left != t_nchars_left, + the strings will be not equal in case of a NOPAD collation. + + Example: + "B" is German "U+00DF LATIN SMALL LETTER SHARP S" + When we have these values in a + CHAR(3) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_nopad_ci + column: + 'B ' (one character, two trailing spaces) + 'ss ' (two characters, one trailing space) + The 'B ' is greater than the 'ss '. + They are compared in the following steps: + 1. 'B' == 'ss' + 2. ' ' == ' ' + 3. ' ' > '' + + We need to emulate the same behavior in this function even if + it's called with strings 'B' and 'ss' (with space trimmed). + The side which has more remaining virtual spaces at the end + is greater. + */ + if (s_nchars_left < t_nchars_left) + return -1; + if (s_nchars_left > t_nchars_left) + return +1; + } + return 0; + } + + DBUG_ASSERT(s_nchars_left >= s_res.nchars); + DBUG_ASSERT(t_nchars_left >= t_res.nchars); + s_nchars_left-= s_res.nchars; + t_nchars_left-= t_res.nchars; + } + + return 0; +} + + +/* + One-level collations. +*/ +static int +MY_FUNCTION_NAME(strnncollsp_nchars)(CHARSET_INFO *cs, + const uchar *s, size_t slen, + const uchar *t, size_t tlen, + size_t nchars) +{ + return MY_FUNCTION_NAME(strnncollsp_nchars_onelevel)(cs, &cs->uca->level[0], + s, slen, t, tlen, + nchars); +} + + +/* + Multi-level collations. +*/ +static int +MY_FUNCTION_NAME(strnncollsp_nchars_multilevel)(CHARSET_INFO *cs, + const uchar *s, size_t slen, + const uchar *t, size_t tlen, + size_t nchars) +{ + uint num_level= cs->levels_for_order; + uint i; + for (i= 0; i != num_level; i++) + { + int ret= MY_FUNCTION_NAME(strnncollsp_nchars_onelevel)(cs, + &cs->uca->level[i], + s, slen, + t, tlen, + nchars); + if (ret) + return ret; + } + return 0; +} + /* Calculates hash value for the given string, @@ -753,6 +821,7 @@ MY_COLLATION_HANDLER MY_FUNCTION_NAME(collation_handler)= MY_UCA_COLL_INIT, MY_FUNCTION_NAME(strnncoll), MY_FUNCTION_NAME(strnncollsp), + MY_FUNCTION_NAME(strnncollsp_nchars), MY_FUNCTION_NAME(strnxfrm), my_strnxfrmlen_any_uca, MY_LIKE_RANGE, @@ -776,6 +845,7 @@ MY_COLLATION_HANDLER MY_FUNCTION_NAME(collation_handler_nopad)= MY_UCA_COLL_INIT, MY_FUNCTION_NAME(strnncoll), MY_FUNCTION_NAME(strnncollsp_nopad), + MY_FUNCTION_NAME(strnncollsp_nchars), MY_FUNCTION_NAME(strnxfrm_nopad), my_strnxfrmlen_any_uca, MY_LIKE_RANGE, /* my_like_range_mb or my_like_range_generic */ @@ -797,6 +867,7 @@ MY_COLLATION_HANDLER MY_FUNCTION_NAME(collation_handler_multilevel)= MY_UCA_COLL_INIT, MY_FUNCTION_NAME(strnncoll_multilevel), MY_FUNCTION_NAME(strnncollsp_multilevel), + MY_FUNCTION_NAME(strnncollsp_nchars_multilevel), MY_FUNCTION_NAME(strnxfrm_multilevel), my_strnxfrmlen_any_uca_multilevel, MY_LIKE_RANGE, @@ -818,6 +889,7 @@ MY_COLLATION_HANDLER MY_FUNCTION_NAME(collation_handler_nopad_multilevel)= MY_UCA_COLL_INIT, MY_FUNCTION_NAME(strnncoll_multilevel), MY_FUNCTION_NAME(strnncollsp_nopad_multilevel), + MY_FUNCTION_NAME(strnncollsp_nchars_multilevel), MY_FUNCTION_NAME(strnxfrm_multilevel), my_strnxfrmlen_any_uca_multilevel, MY_LIKE_RANGE, |