1 /* Copyright (C) 1995,96,97,2002, 2004, 2005 Free Software Foundation, Inc.
2 This file is part of the GNU C Library.
3 Written by Ulrich Drepper <drepper@gnu.org>, 1995.
5 The GNU C Library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 2.1 of the License, or (at your option) any later version.
10 The GNU C Library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
15 You should have received a copy of the GNU Lesser General Public
16 License along with the GNU C Library; if not, write to the Free
17 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
27 #include <sys/param.h>
30 # define STRING_TYPE char
31 # define USTRING_TYPE unsigned char
32 # define STRXFRM __strxfrm_l
33 # define STRCMP strcmp
34 # define STRLEN strlen
35 # define STPNCPY __stpncpy
36 # define WEIGHT_H "../locale/weight.h"
41 #define CONCAT(a,b) CONCAT1(a,b)
42 #define CONCAT1(a,b) a##b
44 #include "../locale/localeinfo.h"
47 #ifndef WIDE_CHAR_VERSION
49 /* We need UTF-8 encoding of numbers. */
51 utf8_encode (char *buf
, int val
)
64 for (step
= 2; step
< 6; ++step
)
65 if ((val
& (~(uint32_t)0 << (5 * step
+ 1))) == 0)
69 *buf
= (unsigned char) (~0xff >> step
);
73 buf
[step
] = 0x80 | (val
& 0x3f);
86 STRXFRM (STRING_TYPE
*dest
, const STRING_TYPE
*src
, size_t n
, __locale_t l
)
88 struct locale_data
*current
= l
->__locales
[LC_COLLATE
];
89 uint_fast32_t nrules
= current
->values
[_NL_ITEM_INDEX (_NL_COLLATE_NRULES
)].word
;
90 /* We don't assign the following values right away since it might be
91 unnecessary in case there are no rules. */
92 const unsigned char *rulesets
;
94 const USTRING_TYPE
*weights
;
95 const USTRING_TYPE
*extra
;
96 const int32_t *indirect
;
99 const USTRING_TYPE
*usrc
;
100 size_t srclen
= STRLEN (src
);
102 unsigned char *rulearr
;
112 STPNCPY (dest
, src
, MIN (srclen
+ 1, n
));
117 rulesets
= (const unsigned char *)
118 current
->values
[_NL_ITEM_INDEX (_NL_COLLATE_RULESETS
)].string
;
119 table
= (const int32_t *)
120 current
->values
[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_TABLE
,SUFFIX
))].string
;
121 weights
= (const USTRING_TYPE
*)
122 current
->values
[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_WEIGHT
,SUFFIX
))].string
;
123 extra
= (const USTRING_TYPE
*)
124 current
->values
[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_EXTRA
,SUFFIX
))].string
;
125 indirect
= (const int32_t *)
126 current
->values
[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_INDIRECT
,SUFFIX
))].string
;
129 assert (((uintptr_t) table
) % __alignof__ (table
[0]) == 0);
130 assert (((uintptr_t) weights
) % __alignof__ (weights
[0]) == 0);
131 assert (((uintptr_t) extra
) % __alignof__ (extra
[0]) == 0);
132 assert (((uintptr_t) indirect
) % __alignof__ (indirect
[0]) == 0);
134 /* Handle an empty string as a special case. */
142 /* We need the elements of the string as unsigned values since they
143 are used as indeces. */
144 usrc
= (const USTRING_TYPE
*) src
;
146 /* Perform the first pass over the string and while doing this find
147 and store the weights for each character. Since we want this to
148 be as fast as possible we are using `alloca' to store the temporary
149 values. But since there is no limit on the length of the string
150 we have to use `malloc' if the string is too long. We should be
151 very conservative here. */
152 if (! __libc_use_alloca (srclen
))
154 idxarr
= (int32_t *) malloc ((srclen
+ 1) * (sizeof (int32_t) + 1));
155 rulearr
= (unsigned char *) &idxarr
[srclen
];
158 /* No memory. Well, go with the stack then.
160 XXX Once this implementation is stable we will handle this
161 differently. Instead of precomputing the indeces we will
162 do this in time. This means, though, that this happens for
170 idxarr
= (int32_t *) alloca (srclen
* sizeof (int32_t));
171 rulearr
= (unsigned char *) alloca (srclen
+ 1);
177 int32_t tmp
= findidx (&usrc
);
178 rulearr
[idxmax
] = tmp
>> 24;
179 idxarr
[idxmax
] = tmp
& 0xffffff;
183 while (*usrc
!= L('\0'));
185 /* This element is only read, the value never used but to determine
186 another value which then is ignored. */
187 rulearr
[idxmax
] = '\0';
189 /* Now the passes over the weights. We now use the indeces we found
192 for (pass
= 0; pass
< nrules
; ++pass
)
194 size_t backw_stop
= ~0ul;
195 int rule
= rulesets
[rulearr
[0] * nrules
+ pass
];
196 /* We assume that if a rule has defined `position' in one section
197 this is true for all of them. */
198 int position
= rule
& sort_position
;
202 for (idxcnt
= 0; idxcnt
< idxmax
; ++idxcnt
)
204 if ((rule
& sort_forward
) != 0)
208 if (backw_stop
!= ~0ul)
210 /* Handle the pushed elements now. */
213 for (backw
= idxcnt
; backw
> backw_stop
; )
216 len
= weights
[idxarr
[backw
]++];
218 if (needed
+ len
< n
)
220 dest
[needed
++] = weights
[idxarr
[backw
]++];
223 /* No more characters fit into the buffer. */
225 idxarr
[backw
] += len
;
232 /* Now handle the forward element. */
233 len
= weights
[idxarr
[idxcnt
]++];
234 if (needed
+ len
< n
)
236 dest
[needed
++] = weights
[idxarr
[idxcnt
]++];
239 /* No more characters fit into the buffer. */
241 idxarr
[idxcnt
] += len
;
246 /* Remember where the backwards series started. */
247 if (backw_stop
== ~0ul)
251 rule
= rulesets
[rulearr
[idxcnt
+ 1] * nrules
+ pass
];
255 if (backw_stop
!= ~0ul)
257 /* Handle the pushed elements now. */
261 while (backw
> backw_stop
)
263 size_t len
= weights
[idxarr
[--backw
]++];
265 if (needed
+ len
< n
)
267 dest
[needed
++] = weights
[idxarr
[backw
]++];
270 /* No more characters fit into the buffer. */
272 idxarr
[backw
] += len
;
280 #ifndef WIDE_CHAR_VERSION
286 for (idxcnt
= 0; idxcnt
< idxmax
; ++idxcnt
)
288 if ((rule
& sort_forward
) != 0)
292 if (backw_stop
!= ~0ul)
294 /* Handle the pushed elements now. */
297 for (backw
= idxcnt
; backw
> backw_stop
; )
300 len
= weights
[idxarr
[backw
]++];
303 #ifdef WIDE_CHAR_VERSION
304 if (needed
+ 1 + len
< n
)
307 for (i
= 0; i
< len
; ++i
)
308 dest
[needed
+ 1 + i
] =
309 weights
[idxarr
[backw
] + i
];
313 buflen
= utf8_encode (buf
, val
);
314 if (needed
+ buflen
+ len
< n
)
316 for (i
= 0; i
< buflen
; ++i
)
317 dest
[needed
+ i
] = buf
[i
];
318 for (i
= 0; i
< len
; ++i
)
319 dest
[needed
+ buflen
+ i
] =
320 weights
[idxarr
[backw
] + i
];
322 needed
+= buflen
+ len
;
324 idxarr
[backw
] += len
;
334 /* Now handle the forward element. */
335 len
= weights
[idxarr
[idxcnt
]++];
338 #ifdef WIDE_CHAR_VERSION
339 if (needed
+ 1+ len
< n
)
342 for (i
= 0; i
< len
; ++i
)
343 dest
[needed
+ 1 + i
] =
344 weights
[idxarr
[idxcnt
] + i
];
348 buflen
= utf8_encode (buf
, val
);
349 if (needed
+ buflen
+ len
< n
)
351 for (i
= 0; i
< buflen
; ++i
)
352 dest
[needed
+ i
] = buf
[i
];
353 for (i
= 0; i
< len
; ++i
)
354 dest
[needed
+ buflen
+ i
] =
355 weights
[idxarr
[idxcnt
] + i
];
357 needed
+= buflen
+ len
;
359 idxarr
[idxcnt
] += len
;
363 /* Note that we don't have to increment `idxarr[idxcnt]'
364 since the length is zero. */
369 /* Remember where the backwards series started. */
370 if (backw_stop
== ~0ul)
374 rule
= rulesets
[rulearr
[idxcnt
+ 1] * nrules
+ pass
];
377 if (backw_stop
!= ~0ul)
379 /* Handle the pushed elements now. */
383 while (backw
> backw_stop
)
385 size_t len
= weights
[idxarr
[--backw
]++];
388 #ifdef WIDE_CHAR_VERSION
389 if (needed
+ 1 + len
< n
)
392 for (i
= 0; i
< len
; ++i
)
393 dest
[needed
+ 1 + i
] =
394 weights
[idxarr
[backw
] + i
];
398 buflen
= utf8_encode (buf
, val
);
399 if (needed
+ buflen
+ len
< n
)
401 for (i
= 0; i
< buflen
; ++i
)
402 dest
[needed
+ i
] = buf
[i
];
403 for (i
= 0; i
< len
; ++i
)
404 dest
[needed
+ buflen
+ i
] =
405 weights
[idxarr
[backw
] + i
];
407 needed
+= buflen
+ len
;
409 idxarr
[backw
] += len
;
418 /* Finally store the byte to separate the passes or terminate
421 dest
[needed
] = pass
+ 1 < nrules
? L('\1') : L('\0');
425 /* This is a little optimization: many collation specifications have
426 a `position' rule at the end and if no non-ignored character
427 is found the last \1 byte is immediately followed by a \0 byte
428 signalling this. We can avoid the \1 byte(s). */
429 if (needed
<= n
&& needed
> 2 && dest
[needed
- 2] == L('\1'))
431 /* Remove the \1 byte. */
433 dest
[needed
- 1] = L('\0');
436 /* Free the memory if needed. */
440 /* Return the number of bytes/words we need, but don't count the NUL
441 byte/word at the end. */
444 libc_hidden_def (STRXFRM
)
446 #ifndef WIDE_CHAR_VERSION
447 weak_alias (__strxfrm_l
, strxfrm_l
)