wctype/wchar-lookup.h

   1 /* Copyright (C) 2000-2024 Free Software Foundation, Inc.
   2    This file is part of the GNU C Library.
   3
   4    The GNU C Library is free software; you can redistribute it and/or
   5    modify it under the terms of the GNU Lesser General Public
   6    License as published by the Free Software Foundation; either
   7    version 2.1 of the License, or (at your option) any later version.
   8
   9    The GNU C Library is distributed in the hope that it will be useful,
  10    but WITHOUT ANY WARRANTY; without even the implied warranty of
  11    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12    Lesser General Public License for more details.
  13
  14    You should have received a copy of the GNU Lesser General Public
  15    License along with the GNU C Library; if not, see
  16    <https://www.gnu.org/licenses/>.  */
  17
  18 #include <stdint.h>
  19
  20 /* Tables indexed by a wide character are compressed through the use
  21    of a multi-level lookup.  The compression effect comes from blocks
  22    that don't need particular data and from blocks that can share their
  23    data.  */
  24
  25 /* Bit tables are accessed by cutting wc in four blocks of bits:
  26    - the high 32-q-p bits,
  27    - the next q bits,
  28    - the next p bits,
  29    - the next 5 bits.
  30
  31             +------------------+-----+-----+-----+
  32      wc  =  +     32-q-p-5     |  q  |  p  |  5  |
  33             +------------------+-----+-----+-----+
  34
  35    p and q are variable.  For 16-bit Unicode it is sufficient to
  36    choose p and q such that q+p+5 <= 16.
  37
  38    The table contains the following uint32_t words:
  39    - q+p+5,
  40    - s = upper exclusive bound for wc >> (q+p+5),
  41    - p+5,
  42    - 2^q-1,
  43    - 2^p-1,
  44    - 1st-level table: s offsets, pointing into the 2nd-level table,
  45    - 2nd-level table: k*2^q offsets, pointing into the 3rd-level table,
  46    - 3rd-level table: j*2^p words, each containing 32 bits of data.
  47 */
  48
  49 static __inline int
  50 __attribute ((always_inline))
  51 wctype_table_lookup (const char *table, uint32_t wc)
  52 {
  53   uint32_t shift1 = ((const uint32_t *) table)[0];
  54   uint32_t index1 = wc >> shift1;
  55   uint32_t bound = ((const uint32_t *) table)[1];
  56   if (index1 < bound)
  57     {
  58       uint32_t lookup1 = ((const uint32_t *) table)[5 + index1];
  59       if (lookup1 != 0)
  60         {
  61           uint32_t shift2 = ((const uint32_t *) table)[2];
  62           uint32_t mask2 = ((const uint32_t *) table)[3];
  63           uint32_t index2 = (wc >> shift2) & mask2;
  64           uint32_t lookup2 = ((const uint32_t *)(table + lookup1))[index2];
  65           if (lookup2 != 0)
  66             {
  67               uint32_t mask3 = ((const uint32_t *) table)[4];
  68               uint32_t index3 = (wc >> 5) & mask3;
  69               uint32_t lookup3 = ((const uint32_t *)(table + lookup2))[index3];
  70
  71               return (lookup3 >> (wc & 0x1f)) & 1;
  72             }
  73         }
  74     }
  75   return 0;
  76 }
  77
  78 /* Byte tables are similar to bit tables, except that the addressing
  79    unit is a single byte, and no 5 bits are used as a word index.  */
  80
  81 static __inline int
  82 __attribute ((always_inline))
  83 wcwidth_table_lookup (const char *table, uint32_t wc)
  84 {
  85   uint32_t shift1 = ((const uint32_t *) table)[0];
  86   uint32_t index1 = wc >> shift1;
  87   uint32_t bound = ((const uint32_t *) table)[1];
  88   if (index1 < bound)
  89     {
  90       uint32_t lookup1 = ((const uint32_t *) table)[5 + index1];
  91       if (lookup1 != 0)
  92         {
  93           uint32_t shift2 = ((const uint32_t *) table)[2];
  94           uint32_t mask2 = ((const uint32_t *) table)[3];
  95           uint32_t index2 = (wc >> shift2) & mask2;
  96           uint32_t lookup2 = ((const uint32_t *)(table + lookup1))[index2];
  97           if (lookup2 != 0)
  98             {
  99               uint32_t mask3 = ((const uint32_t *) table)[4];
 100               uint32_t index3 = wc & mask3;
 101               uint8_t lookup3 = ((const uint8_t *)(table + lookup2))[index3];
 102
 103               return lookup3;
 104             }
 105         }
 106     }
 107   return 0xff;
 108 }
 109
 110 /* Mapping tables are similar to bit tables, except that the
 111    addressing unit is a single signed 32-bit word, containing the
 112    difference between the desired result and the argument, and no 5
 113    bits are used as a word index.  */
 114
 115 static __inline uint32_t
 116 __attribute ((always_inline))
 117 wctrans_table_lookup (const char *table, uint32_t wc)
 118 {
 119   uint32_t shift1 = ((const uint32_t *) table)[0];
 120   uint32_t index1 = wc >> shift1;
 121   uint32_t bound = ((const uint32_t *) table)[1];
 122   if (index1 < bound)
 123     {
 124       uint32_t lookup1 = ((const uint32_t *) table)[5 + index1];
 125       if (lookup1 != 0)
 126         {
 127           uint32_t shift2 = ((const uint32_t *) table)[2];
 128           uint32_t mask2 = ((const uint32_t *) table)[3];
 129           uint32_t index2 = (wc >> shift2) & mask2;
 130           uint32_t lookup2 = ((const uint32_t *)(table + lookup1))[index2];
 131           if (lookup2 != 0)
 132             {
 133               uint32_t mask3 = ((const uint32_t *) table)[4];
 134               uint32_t index3 = wc & mask3;
 135               int32_t lookup3 = ((const int32_t *)(table + lookup2))[index3];
 136
 137               return wc + lookup3;
 138             }
 139         }
 140     }
 141   return wc;
 142 }