1 // © 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
4 ******************************************************************************
6 * Copyright (C) 2001-2012, International Business Machines
7 * Corporation and others. All Rights Reserved.
9 ******************************************************************************
10 * file name: utrie.cpp
12 * tab size: 8 (not used)
15 * created on: 2001oct20
16 * created by: Markus W. Scherer
18 * This is a common implementation of a "folded" trie.
19 * It is a kind of compressed, serializable table of 16- or 32-bit values associated with
20 * Unicode code points (0..0x10ffff).
27 #include "unicode/utypes.h"
31 /* miscellaneous ------------------------------------------------------------ */
34 #define ABS(x) ((x)>=0 ? (x) : -(x))
37 equal_uint32(const uint32_t *s
, const uint32_t *t
, int32_t length
) {
38 while(length
>0 && *s
==*t
) {
43 return (UBool
)(length
==0);
46 /* Building a trie ----------------------------------------------------------*/
48 U_CAPI UNewTrie
* U_EXPORT2
49 utrie_open(UNewTrie
*fillIn
,
50 uint32_t *aliasData
, int32_t maxDataLength
,
51 uint32_t initialValue
, uint32_t leadUnitValue
,
56 if( maxDataLength
<UTRIE_DATA_BLOCK_LENGTH
||
57 (latin1Linear
&& maxDataLength
<1024)
65 trie
=(UNewTrie
*)uprv_malloc(sizeof(UNewTrie
));
70 uprv_memset(trie
, 0, sizeof(UNewTrie
));
71 trie
->isAllocated
= (UBool
)(fillIn
==NULL
);
75 trie
->isDataAllocated
=FALSE
;
77 trie
->data
=(uint32_t *)uprv_malloc(maxDataLength
*4);
78 if(trie
->data
==NULL
) {
82 trie
->isDataAllocated
=TRUE
;
85 /* preallocate and reset the first data block (block index 0) */
86 j
=UTRIE_DATA_BLOCK_LENGTH
;
89 /* preallocate and reset the first block (number 0) and Latin-1 (U+0000..U+00ff) after that */
90 /* made sure above that maxDataLength>=1024 */
92 /* set indexes to point to consecutive data blocks */
95 /* do this at least for trie->index[0] even if that block is only partly used for Latin-1 */
97 j
+=UTRIE_DATA_BLOCK_LENGTH
;
98 } while(i
<(256>>UTRIE_SHIFT
));
101 /* reset the initially allocated blocks to the initial value */
104 trie
->data
[--j
]=initialValue
;
107 trie
->leadUnitValue
=leadUnitValue
;
108 trie
->indexLength
=UTRIE_MAX_INDEX_LENGTH
;
109 trie
->dataCapacity
=maxDataLength
;
110 trie
->isLatin1Linear
=latin1Linear
;
111 trie
->isCompacted
=FALSE
;
115 U_CAPI UNewTrie
* U_EXPORT2
116 utrie_clone(UNewTrie
*fillIn
, const UNewTrie
*other
, uint32_t *aliasData
, int32_t aliasDataCapacity
) {
118 UBool isDataAllocated
;
120 /* do not clone if other is not valid or already compacted */
121 if(other
==NULL
|| other
->data
==NULL
|| other
->isCompacted
) {
126 if(aliasData
!=NULL
&& aliasDataCapacity
>=other
->dataCapacity
) {
127 isDataAllocated
=FALSE
;
129 aliasDataCapacity
=other
->dataCapacity
;
130 aliasData
=(uint32_t *)uprv_malloc(other
->dataCapacity
*4);
131 if(aliasData
==NULL
) {
134 isDataAllocated
=TRUE
;
137 trie
=utrie_open(fillIn
, aliasData
, aliasDataCapacity
,
138 other
->data
[0], other
->leadUnitValue
,
139 other
->isLatin1Linear
);
141 uprv_free(aliasData
);
143 uprv_memcpy(trie
->index
, other
->index
, sizeof(trie
->index
));
144 uprv_memcpy(trie
->data
, other
->data
, (size_t)other
->dataLength
*4);
145 trie
->dataLength
=other
->dataLength
;
146 trie
->isDataAllocated
=isDataAllocated
;
152 U_CAPI
void U_EXPORT2
153 utrie_close(UNewTrie
*trie
) {
155 if(trie
->isDataAllocated
) {
156 uprv_free(trie
->data
);
159 if(trie
->isAllocated
) {
165 U_CAPI
uint32_t * U_EXPORT2
166 utrie_getData(UNewTrie
*trie
, int32_t *pLength
) {
167 if(trie
==NULL
|| pLength
==NULL
) {
171 *pLength
=trie
->dataLength
;
176 utrie_allocDataBlock(UNewTrie
*trie
) {
177 int32_t newBlock
, newTop
;
179 newBlock
=trie
->dataLength
;
180 newTop
=newBlock
+UTRIE_DATA_BLOCK_LENGTH
;
181 if(newTop
>trie
->dataCapacity
) {
182 /* out of memory in the data array */
185 trie
->dataLength
=newTop
;
190 * No error checking for illegal arguments.
192 * @return -1 if no new data block available (out of memory in data array)
196 utrie_getDataBlock(UNewTrie
*trie
, UChar32 c
) {
197 int32_t indexValue
, newBlock
;
200 indexValue
=trie
->index
[c
];
205 /* allocate a new data block */
206 newBlock
=utrie_allocDataBlock(trie
);
208 /* out of memory in the data array */
211 trie
->index
[c
]=newBlock
;
213 /* copy-on-write for a block from a setRange() */
214 uprv_memcpy(trie
->data
+newBlock
, trie
->data
-indexValue
, 4*UTRIE_DATA_BLOCK_LENGTH
);
219 * @return TRUE if the value was successfully set
221 U_CAPI UBool U_EXPORT2
222 utrie_set32(UNewTrie
*trie
, UChar32 c
, uint32_t value
) {
225 /* valid, uncompacted trie and valid c? */
226 if(trie
==NULL
|| trie
->isCompacted
|| (uint32_t)c
>0x10ffff) {
230 block
=utrie_getDataBlock(trie
, c
);
235 trie
->data
[block
+(c
&UTRIE_MASK
)]=value
;
239 U_CAPI
uint32_t U_EXPORT2
240 utrie_get32(UNewTrie
*trie
, UChar32 c
, UBool
*pInBlockZero
) {
243 /* valid, uncompacted trie and valid c? */
244 if(trie
==NULL
|| trie
->isCompacted
|| (uint32_t)c
>0x10ffff) {
245 if(pInBlockZero
!=NULL
) {
251 block
=trie
->index
[c
>>UTRIE_SHIFT
];
252 if(pInBlockZero
!=NULL
) {
253 *pInBlockZero
= (UBool
)(block
==0);
256 return trie
->data
[ABS(block
)+(c
&UTRIE_MASK
)];
263 utrie_fillBlock(uint32_t *block
, UChar32 start
, UChar32 limit
,
264 uint32_t value
, uint32_t initialValue
, UBool overwrite
) {
270 while(block
<pLimit
) {
274 while(block
<pLimit
) {
275 if(*block
==initialValue
) {
283 U_CAPI UBool U_EXPORT2
284 utrie_setRange32(UNewTrie
*trie
, UChar32 start
, UChar32 limit
, uint32_t value
, UBool overwrite
) {
286 * repeat value in [start..limit[
287 * mark index values for repeat-data blocks by setting bit 31 of the index values
288 * fill around existing values if any, if(overwrite)
290 uint32_t initialValue
;
291 int32_t block
, rest
, repeatBlock
;
293 /* valid, uncompacted trie and valid indexes? */
294 if( trie
==NULL
|| trie
->isCompacted
||
295 (uint32_t)start
>0x10ffff || (uint32_t)limit
>0x110000 || start
>limit
300 return TRUE
; /* nothing to do */
303 initialValue
=trie
->data
[0];
304 if(start
&UTRIE_MASK
) {
307 /* set partial block at [start..following block boundary[ */
308 block
=utrie_getDataBlock(trie
, start
);
313 nextStart
=(start
+UTRIE_DATA_BLOCK_LENGTH
)&~UTRIE_MASK
;
314 if(nextStart
<=limit
) {
315 utrie_fillBlock(trie
->data
+block
, start
&UTRIE_MASK
, UTRIE_DATA_BLOCK_LENGTH
,
316 value
, initialValue
, overwrite
);
319 utrie_fillBlock(trie
->data
+block
, start
&UTRIE_MASK
, limit
&UTRIE_MASK
,
320 value
, initialValue
, overwrite
);
325 /* number of positions in the last, partial block */
326 rest
=limit
&UTRIE_MASK
;
328 /* round down limit to a block boundary */
331 /* iterate over all-value blocks */
332 if(value
==initialValue
) {
338 /* get index value */
339 block
=trie
->index
[start
>>UTRIE_SHIFT
];
341 /* already allocated, fill in value */
342 utrie_fillBlock(trie
->data
+block
, 0, UTRIE_DATA_BLOCK_LENGTH
, value
, initialValue
, overwrite
);
343 } else if(trie
->data
[-block
]!=value
&& (block
==0 || overwrite
)) {
344 /* set the repeatBlock instead of the current block 0 or range block */
346 trie
->index
[start
>>UTRIE_SHIFT
]=-repeatBlock
;
348 /* create and set and fill the repeatBlock */
349 repeatBlock
=utrie_getDataBlock(trie
, start
);
354 /* set the negative block number to indicate that it is a repeat block */
355 trie
->index
[start
>>UTRIE_SHIFT
]=-repeatBlock
;
356 utrie_fillBlock(trie
->data
+repeatBlock
, 0, UTRIE_DATA_BLOCK_LENGTH
, value
, initialValue
, TRUE
);
360 start
+=UTRIE_DATA_BLOCK_LENGTH
;
364 /* set partial block at [last block boundary..limit[ */
365 block
=utrie_getDataBlock(trie
, start
);
370 utrie_fillBlock(trie
->data
+block
, 0, rest
, value
, initialValue
, overwrite
);
377 _findSameIndexBlock(const int32_t *idx
, int32_t indexLength
,
378 int32_t otherBlock
) {
381 for(block
=UTRIE_BMP_INDEX_LENGTH
; block
<indexLength
; block
+=UTRIE_SURROGATE_BLOCK_COUNT
) {
382 for(i
=0; i
<UTRIE_SURROGATE_BLOCK_COUNT
; ++i
) {
383 if(idx
[block
+i
]!=idx
[otherBlock
+i
]) {
387 if(i
==UTRIE_SURROGATE_BLOCK_COUNT
) {
395 * Fold the normalization data for supplementary code points into
396 * a compact area on top of the BMP-part of the trie index,
397 * with the lead surrogates indexing this compact area.
399 * Duplicate the index values for lead surrogates:
400 * From inside the BMP area, where some may be overridden with folded values,
401 * to just after the BMP area, where they can be retrieved for
402 * code point lookups.
405 utrie_fold(UNewTrie
*trie
, UNewTrieGetFoldedValue
*getFoldedValue
, UErrorCode
*pErrorCode
) {
406 int32_t leadIndexes
[UTRIE_SURROGATE_BLOCK_COUNT
];
410 int32_t indexLength
, block
;
412 int countLeadCUWithData
=0;
417 /* copy the lead surrogate indexes into a temporary array */
418 uprv_memcpy(leadIndexes
, idx
+(0xd800>>UTRIE_SHIFT
), 4*UTRIE_SURROGATE_BLOCK_COUNT
);
421 * set all values for lead surrogate code *units* to leadUnitValue
422 * so that, by default, runtime lookups will find no data for associated
423 * supplementary code points, unless there is data for such code points
424 * which will result in a non-zero folding value below that is set for
425 * the respective lead units
427 * the above saved the indexes for surrogate code *points*
428 * fill the indexes with simplified code from utrie_setRange32()
430 if(trie
->leadUnitValue
==trie
->data
[0]) {
431 block
=0; /* leadUnitValue==initialValue, use all-initial-value block */
433 /* create and fill the repeatBlock */
434 block
=utrie_allocDataBlock(trie
);
436 /* data table overflow */
437 *pErrorCode
=U_MEMORY_ALLOCATION_ERROR
;
440 utrie_fillBlock(trie
->data
+block
, 0, UTRIE_DATA_BLOCK_LENGTH
, trie
->leadUnitValue
, trie
->data
[0], TRUE
);
441 block
=-block
; /* negative block number to indicate that it is a repeat block */
443 for(c
=(0xd800>>UTRIE_SHIFT
); c
<(0xdc00>>UTRIE_SHIFT
); ++c
) {
444 trie
->index
[c
]=block
;
448 * Fold significant index values into the area just after the BMP indexes.
449 * In case the first lead surrogate has significant data,
450 * its index block must be used first (in which case the folding is a no-op).
451 * Later all folded index blocks are moved up one to insert the copied
452 * lead surrogate indexes.
454 indexLength
=UTRIE_BMP_INDEX_LENGTH
;
456 /* search for any index (stage 1) entries for supplementary code points */
457 for(c
=0x10000; c
<0x110000;) {
458 if(idx
[c
>>UTRIE_SHIFT
]!=0) {
459 /* there is data, treat the full block for a lead surrogate */
463 ++countLeadCUWithData
;
464 /* printf("supplementary data for lead surrogate U+%04lx\n", (long)(0xd7c0+(c>>10))); */
467 /* is there an identical index block? */
468 block
=_findSameIndexBlock(idx
, indexLength
, c
>>UTRIE_SHIFT
);
471 * get a folded value for [c..c+0x400[ and,
472 * if different from the value for the lead surrogate code point,
473 * set it for the lead surrogate code unit
475 value
=getFoldedValue(trie
, c
, block
+UTRIE_SURROGATE_BLOCK_COUNT
);
476 if(value
!=utrie_get32(trie
, U16_LEAD(c
), NULL
)) {
477 if(!utrie_set32(trie
, U16_LEAD(c
), value
)) {
478 /* data table overflow */
479 *pErrorCode
=U_MEMORY_ALLOCATION_ERROR
;
483 /* if we did not find an identical index block... */
484 if(block
==indexLength
) {
485 /* move the actual index (stage 1) entries from the supplementary position to the new one */
486 uprv_memmove(idx
+indexLength
,
487 idx
+(c
>>UTRIE_SHIFT
),
488 4*UTRIE_SURROGATE_BLOCK_COUNT
);
489 indexLength
+=UTRIE_SURROGATE_BLOCK_COUNT
;
494 c
+=UTRIE_DATA_BLOCK_LENGTH
;
498 if(countLeadCUWithData
>0) {
499 printf("supplementary data for %d lead surrogates\n", countLeadCUWithData
);
504 * index array overflow?
505 * This is to guarantee that a folding offset is of the form
506 * UTRIE_BMP_INDEX_LENGTH+n*UTRIE_SURROGATE_BLOCK_COUNT with n=0..1023.
507 * If the index is too large, then n>=1024 and more than 10 bits are necessary.
509 * In fact, it can only ever become n==1024 with completely unfoldable data and
510 * the additional block of duplicated values for lead surrogates.
512 if(indexLength
>=UTRIE_MAX_INDEX_LENGTH
) {
513 *pErrorCode
=U_INDEX_OUTOFBOUNDS_ERROR
;
518 * make space for the lead surrogate index block and
519 * insert it between the BMP indexes and the folded ones
521 uprv_memmove(idx
+UTRIE_BMP_INDEX_LENGTH
+UTRIE_SURROGATE_BLOCK_COUNT
,
522 idx
+UTRIE_BMP_INDEX_LENGTH
,
523 4*(indexLength
-UTRIE_BMP_INDEX_LENGTH
));
524 uprv_memcpy(idx
+UTRIE_BMP_INDEX_LENGTH
,
526 4*UTRIE_SURROGATE_BLOCK_COUNT
);
527 indexLength
+=UTRIE_SURROGATE_BLOCK_COUNT
;
530 printf("trie index count: BMP %ld all Unicode %ld folded %ld\n",
531 UTRIE_BMP_INDEX_LENGTH
, (long)UTRIE_MAX_INDEX_LENGTH
, indexLength
);
534 trie
->indexLength
=indexLength
;
538 * Set a value in the trie index map to indicate which data block
539 * is referenced and which one is not.
540 * utrie_compact() will remove data blocks that are not used at all.
543 * - -1 if it is not used
546 _findUnusedBlocks(UNewTrie
*trie
) {
549 /* fill the entire map with "not used" */
550 uprv_memset(trie
->map
, 0xff, (UTRIE_MAX_BUILD_TIME_DATA_LENGTH
>>UTRIE_SHIFT
)*4);
552 /* mark each block that _is_ used with 0 */
553 for(i
=0; i
<trie
->indexLength
; ++i
) {
554 trie
->map
[ABS(trie
->index
[i
])>>UTRIE_SHIFT
]=0;
557 /* never move the all-initial-value block 0 */
562 _findSameDataBlock(const uint32_t *data
, int32_t dataLength
,
563 int32_t otherBlock
, int32_t step
) {
566 /* ensure that we do not even partially get past dataLength */
567 dataLength
-=UTRIE_DATA_BLOCK_LENGTH
;
569 for(block
=0; block
<=dataLength
; block
+=step
) {
570 if(equal_uint32(data
+block
, data
+otherBlock
, UTRIE_DATA_BLOCK_LENGTH
)) {
578 * Compact a folded build-time trie.
581 * - removes blocks that are identical with earlier ones
582 * - overlaps adjacent blocks as much as possible (if overlap==TRUE)
583 * - moves blocks in steps of the data granularity
584 * - moves and overlaps blocks that overlap with multiple values in the overlap region
587 * - try to move and overlap blocks that are not already adjacent
590 utrie_compact(UNewTrie
*trie
, UBool overlap
, UErrorCode
*pErrorCode
) {
591 int32_t i
, start
, newStart
, overlapStart
;
593 if(pErrorCode
==NULL
|| U_FAILURE(*pErrorCode
)) {
597 /* valid, uncompacted trie? */
599 *pErrorCode
=U_ILLEGAL_ARGUMENT_ERROR
;
602 if(trie
->isCompacted
) {
603 return; /* nothing left to do */
608 /* initialize the index map with "block is used/unused" flags */
609 _findUnusedBlocks(trie
);
611 /* if Latin-1 is preallocated and linear, then do not compact Latin-1 data */
612 if(trie
->isLatin1Linear
&& UTRIE_SHIFT
<=8) {
613 overlapStart
=UTRIE_DATA_BLOCK_LENGTH
+256;
615 overlapStart
=UTRIE_DATA_BLOCK_LENGTH
;
618 newStart
=UTRIE_DATA_BLOCK_LENGTH
;
619 for(start
=newStart
; start
<trie
->dataLength
;) {
621 * start: index of first entry of current block
622 * newStart: index where the current block is to be moved
623 * (right after current end of already-compacted data)
626 /* skip blocks that are not used */
627 if(trie
->map
[start
>>UTRIE_SHIFT
]<0) {
628 /* advance start to the next block */
629 start
+=UTRIE_DATA_BLOCK_LENGTH
;
631 /* leave newStart with the previous block! */
635 /* search for an identical block */
636 if( start
>=overlapStart
&&
637 (i
=_findSameDataBlock(trie
->data
, newStart
, start
,
638 overlap
? UTRIE_DATA_GRANULARITY
: UTRIE_DATA_BLOCK_LENGTH
))
641 /* found an identical block, set the other block's index value for the current block */
642 trie
->map
[start
>>UTRIE_SHIFT
]=i
;
644 /* advance start to the next block */
645 start
+=UTRIE_DATA_BLOCK_LENGTH
;
647 /* leave newStart with the previous block! */
651 /* see if the beginning of this block can be overlapped with the end of the previous block */
652 if(overlap
&& start
>=overlapStart
) {
653 /* look for maximum overlap (modulo granularity) with the previous, adjacent block */
654 for(i
=UTRIE_DATA_BLOCK_LENGTH
-UTRIE_DATA_GRANULARITY
;
655 i
>0 && !equal_uint32(trie
->data
+(newStart
-i
), trie
->data
+start
, i
);
656 i
-=UTRIE_DATA_GRANULARITY
) {}
663 trie
->map
[start
>>UTRIE_SHIFT
]=newStart
-i
;
665 /* move the non-overlapping indexes to their new positions */
667 for(i
=UTRIE_DATA_BLOCK_LENGTH
-i
; i
>0; --i
) {
668 trie
->data
[newStart
++]=trie
->data
[start
++];
670 } else if(newStart
<start
) {
671 /* no overlap, just move the indexes to their new positions */
672 trie
->map
[start
>>UTRIE_SHIFT
]=newStart
;
673 for(i
=UTRIE_DATA_BLOCK_LENGTH
; i
>0; --i
) {
674 trie
->data
[newStart
++]=trie
->data
[start
++];
676 } else /* no overlap && newStart==start */ {
677 trie
->map
[start
>>UTRIE_SHIFT
]=start
;
678 newStart
+=UTRIE_DATA_BLOCK_LENGTH
;
683 /* now adjust the index (stage 1) table */
684 for(i
=0; i
<trie
->indexLength
; ++i
) {
685 trie
->index
[i
]=trie
->map
[ABS(trie
->index
[i
])>>UTRIE_SHIFT
];
689 /* we saved some space */
690 printf("compacting trie: count of 32-bit words %lu->%lu\n",
691 (long)trie
->dataLength
, (long)newStart
);
694 trie
->dataLength
=newStart
;
697 /* serialization ------------------------------------------------------------ */
700 * Default function for the folding value:
701 * Just store the offset (16 bits) if there is any non-initial-value entry.
703 * The offset parameter is never 0.
704 * Returning the offset itself is safe for UTRIE_SHIFT>=5 because
705 * for UTRIE_SHIFT==5 the maximum index length is UTRIE_MAX_INDEX_LENGTH==0x8800
706 * which fits into 16-bit trie values;
707 * for higher UTRIE_SHIFT, UTRIE_MAX_INDEX_LENGTH decreases.
709 * Theoretically, it would be safer for all possible UTRIE_SHIFT including
710 * those of 4 and lower to return offset>>UTRIE_SURROGATE_BLOCK_BITS
711 * which would always result in a value of 0x40..0x43f
712 * (start/end 1k blocks of supplementary Unicode code points).
713 * However, this would be uglier, and would not work for some existing
714 * binary data file formats.
716 * Also, we do not plan to change UTRIE_SHIFT because it would change binary
717 * data file formats, and we would probably not make it smaller because of
718 * the then even larger BMP index length even for empty tries.
720 static uint32_t U_CALLCONV
721 defaultGetFoldedValue(UNewTrie
*trie
, UChar32 start
, int32_t offset
) {
722 uint32_t value
, initialValue
;
726 initialValue
=trie
->data
[0];
729 value
=utrie_get32(trie
, start
, &inBlockZero
);
731 start
+=UTRIE_DATA_BLOCK_LENGTH
;
732 } else if(value
!=initialValue
) {
733 return (uint32_t)offset
;
741 U_CAPI
int32_t U_EXPORT2
742 utrie_serialize(UNewTrie
*trie
, void *dt
, int32_t capacity
,
743 UNewTrieGetFoldedValue
*getFoldedValue
,
744 UBool reduceTo16Bits
,
745 UErrorCode
*pErrorCode
) {
750 uint8_t* data
= NULL
;
753 if(pErrorCode
==NULL
|| U_FAILURE(*pErrorCode
)) {
757 if(trie
==NULL
|| capacity
<0 || (capacity
>0 && dt
==NULL
)) {
758 *pErrorCode
=U_ILLEGAL_ARGUMENT_ERROR
;
761 if(getFoldedValue
==NULL
) {
762 getFoldedValue
=defaultGetFoldedValue
;
766 /* fold and compact if necessary, also checks that indexLength is within limits */
767 if(!trie
->isCompacted
) {
768 /* compact once without overlap to improve folding */
769 utrie_compact(trie
, FALSE
, pErrorCode
);
771 /* fold the supplementary part of the index array */
772 utrie_fold(trie
, getFoldedValue
, pErrorCode
);
774 /* compact again with overlap for minimum data array length */
775 utrie_compact(trie
, TRUE
, pErrorCode
);
777 trie
->isCompacted
=TRUE
;
778 if(U_FAILURE(*pErrorCode
)) {
783 /* is dataLength within limits? */
784 if( (reduceTo16Bits
? (trie
->dataLength
+trie
->indexLength
) : trie
->dataLength
) >= UTRIE_MAX_DATA_LENGTH
) {
785 *pErrorCode
=U_INDEX_OUTOFBOUNDS_ERROR
;
788 length
=sizeof(UTrieHeader
)+2*trie
->indexLength
;
790 length
+=2*trie
->dataLength
;
792 length
+=4*trie
->dataLength
;
795 if(length
>capacity
) {
796 return length
; /* preflighting */
800 printf("**UTrieLengths(serialize)** index:%6ld data:%6ld serialized:%6ld\n",
801 (long)trie
->indexLength
, (long)trie
->dataLength
, (long)length
);
804 /* set the header fields */
805 header
=(UTrieHeader
*)data
;
806 data
+=sizeof(UTrieHeader
);
808 header
->signature
=0x54726965; /* "Trie" */
809 header
->options
=UTRIE_SHIFT
| (UTRIE_INDEX_SHIFT
<<UTRIE_OPTIONS_INDEX_SHIFT
);
811 if(!reduceTo16Bits
) {
812 header
->options
|=UTRIE_OPTIONS_DATA_IS_32_BIT
;
814 if(trie
->isLatin1Linear
) {
815 header
->options
|=UTRIE_OPTIONS_LATIN1_IS_LINEAR
;
818 header
->indexLength
=trie
->indexLength
;
819 header
->dataLength
=trie
->dataLength
;
821 /* write the index (stage 1) array and the 16/32-bit data (stage 2) array */
823 /* write 16-bit index values shifted right by UTRIE_INDEX_SHIFT, after adding indexLength */
824 p
=(uint32_t *)trie
->index
;
825 dest16
=(uint16_t *)data
;
826 for(i
=trie
->indexLength
; i
>0; --i
) {
827 *dest16
++=(uint16_t)((*p
++ + trie
->indexLength
)>>UTRIE_INDEX_SHIFT
);
830 /* write 16-bit data values */
832 for(i
=trie
->dataLength
; i
>0; --i
) {
833 *dest16
++=(uint16_t)*p
++;
836 /* write 16-bit index values shifted right by UTRIE_INDEX_SHIFT */
837 p
=(uint32_t *)trie
->index
;
838 dest16
=(uint16_t *)data
;
839 for(i
=trie
->indexLength
; i
>0; --i
) {
840 *dest16
++=(uint16_t)(*p
++ >> UTRIE_INDEX_SHIFT
);
843 /* write 32-bit data values */
844 uprv_memcpy(dest16
, trie
->data
, 4*(size_t)trie
->dataLength
);
850 /* inverse to defaultGetFoldedValue() */
851 U_CAPI
int32_t U_EXPORT2
852 utrie_defaultGetFoldingOffset(uint32_t data
) {
853 return (int32_t)data
;
856 U_CAPI
int32_t U_EXPORT2
857 utrie_unserialize(UTrie
*trie
, const void *data
, int32_t length
, UErrorCode
*pErrorCode
) {
858 const UTrieHeader
*header
;
862 if(pErrorCode
==NULL
|| U_FAILURE(*pErrorCode
)) {
866 /* enough data for a trie header? */
867 if(length
<(int32_t)sizeof(UTrieHeader
)) {
868 *pErrorCode
=U_INVALID_FORMAT_ERROR
;
872 /* check the signature */
873 header
=(const UTrieHeader
*)data
;
874 if(header
->signature
!=0x54726965) {
875 *pErrorCode
=U_INVALID_FORMAT_ERROR
;
879 /* get the options and check the shift values */
880 options
=header
->options
;
881 if( (options
&UTRIE_OPTIONS_SHIFT_MASK
)!=UTRIE_SHIFT
||
882 ((options
>>UTRIE_OPTIONS_INDEX_SHIFT
)&UTRIE_OPTIONS_SHIFT_MASK
)!=UTRIE_INDEX_SHIFT
884 *pErrorCode
=U_INVALID_FORMAT_ERROR
;
887 trie
->isLatin1Linear
= (UBool
)((options
&UTRIE_OPTIONS_LATIN1_IS_LINEAR
)!=0);
889 /* get the length values */
890 trie
->indexLength
=header
->indexLength
;
891 trie
->dataLength
=header
->dataLength
;
893 length
-=(int32_t)sizeof(UTrieHeader
);
895 /* enough data for the index? */
896 if(length
<2*trie
->indexLength
) {
897 *pErrorCode
=U_INVALID_FORMAT_ERROR
;
900 p16
=(const uint16_t *)(header
+1);
902 p16
+=trie
->indexLength
;
903 length
-=2*trie
->indexLength
;
906 if(options
&UTRIE_OPTIONS_DATA_IS_32_BIT
) {
907 if(length
<4*trie
->dataLength
) {
908 *pErrorCode
=U_INVALID_FORMAT_ERROR
;
911 trie
->data32
=(const uint32_t *)p16
;
912 trie
->initialValue
=trie
->data32
[0];
913 length
=(int32_t)sizeof(UTrieHeader
)+2*trie
->indexLength
+4*trie
->dataLength
;
915 if(length
<2*trie
->dataLength
) {
916 *pErrorCode
=U_INVALID_FORMAT_ERROR
;
920 /* the "data16" data is used via the index pointer */
922 trie
->initialValue
=trie
->index
[trie
->indexLength
];
923 length
=(int32_t)sizeof(UTrieHeader
)+2*trie
->indexLength
+2*trie
->dataLength
;
926 trie
->getFoldingOffset
=utrie_defaultGetFoldingOffset
;
931 U_CAPI
int32_t U_EXPORT2
932 utrie_unserializeDummy(UTrie
*trie
,
933 void *data
, int32_t length
,
934 uint32_t initialValue
, uint32_t leadUnitValue
,
936 UErrorCode
*pErrorCode
) {
938 int32_t actualLength
, latin1Length
, i
, limit
;
941 if(pErrorCode
==NULL
|| U_FAILURE(*pErrorCode
)) {
945 /* calculate the actual size of the dummy trie data */
947 /* max(Latin-1, block 0) */
948 latin1Length
= 256; /*UTRIE_SHIFT<=8 ? 256 : UTRIE_DATA_BLOCK_LENGTH;*/
950 trie
->indexLength
=UTRIE_BMP_INDEX_LENGTH
+UTRIE_SURROGATE_BLOCK_COUNT
;
951 trie
->dataLength
=latin1Length
;
952 if(leadUnitValue
!=initialValue
) {
953 trie
->dataLength
+=UTRIE_DATA_BLOCK_LENGTH
;
956 actualLength
=trie
->indexLength
*2;
958 actualLength
+=trie
->dataLength
*2;
960 actualLength
+=trie
->dataLength
*4;
963 /* enough space for the dummy trie? */
964 if(length
<actualLength
) {
965 *pErrorCode
=U_BUFFER_OVERFLOW_ERROR
;
969 trie
->isLatin1Linear
=TRUE
;
970 trie
->initialValue
=initialValue
;
972 /* fill the index and data arrays */
973 p16
=(uint16_t *)data
;
977 /* indexes to block 0 */
978 block
=(uint16_t)(trie
->indexLength
>>UTRIE_INDEX_SHIFT
);
979 limit
=trie
->indexLength
;
980 for(i
=0; i
<limit
; ++i
) {
984 if(leadUnitValue
!=initialValue
) {
985 /* indexes for lead surrogate code units to the block after Latin-1 */
986 block
+=(uint16_t)(latin1Length
>>UTRIE_INDEX_SHIFT
);
987 i
=0xd800>>UTRIE_SHIFT
;
988 limit
=0xdc00>>UTRIE_SHIFT
;
989 for(; i
<limit
; ++i
) {
997 p16
+=trie
->indexLength
;
998 for(i
=0; i
<latin1Length
; ++i
) {
999 p16
[i
]=(uint16_t)initialValue
;
1002 /* data for lead surrogate code units */
1003 if(leadUnitValue
!=initialValue
) {
1004 limit
=latin1Length
+UTRIE_DATA_BLOCK_LENGTH
;
1005 for(/* i=latin1Length */; i
<limit
; ++i
) {
1006 p16
[i
]=(uint16_t)leadUnitValue
;
1012 /* indexes to block 0 */
1013 uprv_memset(p16
, 0, trie
->indexLength
*2);
1015 if(leadUnitValue
!=initialValue
) {
1016 /* indexes for lead surrogate code units to the block after Latin-1 */
1017 block
=(uint16_t)(latin1Length
>>UTRIE_INDEX_SHIFT
);
1018 i
=0xd800>>UTRIE_SHIFT
;
1019 limit
=0xdc00>>UTRIE_SHIFT
;
1020 for(; i
<limit
; ++i
) {
1025 trie
->data32
=p32
=(uint32_t *)(p16
+trie
->indexLength
);
1028 for(i
=0; i
<latin1Length
; ++i
) {
1029 p32
[i
]=initialValue
;
1032 /* data for lead surrogate code units */
1033 if(leadUnitValue
!=initialValue
) {
1034 limit
=latin1Length
+UTRIE_DATA_BLOCK_LENGTH
;
1035 for(/* i=latin1Length */; i
<limit
; ++i
) {
1036 p32
[i
]=leadUnitValue
;
1041 trie
->getFoldingOffset
=utrie_defaultGetFoldingOffset
;
1043 return actualLength
;
1046 /* enumeration -------------------------------------------------------------- */
1048 /* default UTrieEnumValue() returns the input value itself */
1049 static uint32_t U_CALLCONV
1050 enumSameValue(const void * /*context*/, uint32_t value
) {
1055 * Enumerate all ranges of code points with the same relevant values.
1056 * The values are transformed from the raw trie entries by the enumValue function.
1058 U_CAPI
void U_EXPORT2
1059 utrie_enum(const UTrie
*trie
,
1060 UTrieEnumValue
*enumValue
, UTrieEnumRange
*enumRange
, const void *context
) {
1061 const uint32_t *data32
;
1062 const uint16_t *idx
;
1064 uint32_t value
, prevValue
, initialValue
;
1066 int32_t l
, i
, j
, block
, prevBlock
, nullBlock
, offset
;
1068 /* check arguments */
1069 if(trie
==NULL
|| trie
->index
==NULL
|| enumRange
==NULL
) {
1072 if(enumValue
==NULL
) {
1073 enumValue
=enumSameValue
;
1077 data32
=trie
->data32
;
1079 /* get the enumeration value that corresponds to an initial-value trie data entry */
1080 initialValue
=enumValue(context
, trie
->initialValue
);
1083 nullBlock
=trie
->indexLength
;
1088 /* set variables for previous range */
1089 prevBlock
=nullBlock
;
1091 prevValue
=initialValue
;
1093 /* enumerate BMP - the main loop enumerates data blocks */
1094 for(i
=0, c
=0; c
<=0xffff; ++i
) {
1096 /* skip lead surrogate code _units_, go to lead surr. code _points_ */
1097 i
=UTRIE_BMP_INDEX_LENGTH
;
1098 } else if(c
==0xdc00) {
1099 /* go back to regular BMP code points */
1103 block
=idx
[i
]<<UTRIE_INDEX_SHIFT
;
1104 if(block
==prevBlock
) {
1105 /* the block is the same as the previous one, and filled with value */
1106 c
+=UTRIE_DATA_BLOCK_LENGTH
;
1107 } else if(block
==nullBlock
) {
1108 /* this is the all-initial-value block */
1109 if(prevValue
!=initialValue
) {
1111 if(!enumRange(context
, prev
, c
, prevValue
)) {
1115 prevBlock
=nullBlock
;
1117 prevValue
=initialValue
;
1119 c
+=UTRIE_DATA_BLOCK_LENGTH
;
1122 for(j
=0; j
<UTRIE_DATA_BLOCK_LENGTH
; ++j
) {
1123 value
=enumValue(context
, data32
!=NULL
? data32
[block
+j
] : idx
[block
+j
]);
1124 if(value
!=prevValue
) {
1126 if(!enumRange(context
, prev
, c
, prevValue
)) {
1131 /* the block is not filled with all the same value */
1142 /* enumerate supplementary code points */
1143 for(l
=0xd800; l
<0xdc00;) {
1144 /* lead surrogate access */
1145 offset
=idx
[l
>>UTRIE_SHIFT
]<<UTRIE_INDEX_SHIFT
;
1146 if(offset
==nullBlock
) {
1147 /* no entries for a whole block of lead surrogates */
1148 if(prevValue
!=initialValue
) {
1150 if(!enumRange(context
, prev
, c
, prevValue
)) {
1154 prevBlock
=nullBlock
;
1156 prevValue
=initialValue
;
1159 l
+=UTRIE_DATA_BLOCK_LENGTH
;
1160 c
+=UTRIE_DATA_BLOCK_LENGTH
<<10;
1164 value
= data32
!=NULL
? data32
[offset
+(l
&UTRIE_MASK
)] : idx
[offset
+(l
&UTRIE_MASK
)];
1166 /* enumerate trail surrogates for this lead surrogate */
1167 offset
=trie
->getFoldingOffset(value
);
1169 /* no data for this lead surrogate */
1170 if(prevValue
!=initialValue
) {
1172 if(!enumRange(context
, prev
, c
, prevValue
)) {
1176 prevBlock
=nullBlock
;
1178 prevValue
=initialValue
;
1181 /* nothing else to do for the supplementary code points for this lead surrogate */
1184 /* enumerate code points for this lead surrogate */
1186 offset
+=UTRIE_SURROGATE_BLOCK_COUNT
;
1188 /* copy of most of the body of the BMP loop */
1189 block
=idx
[i
]<<UTRIE_INDEX_SHIFT
;
1190 if(block
==prevBlock
) {
1191 /* the block is the same as the previous one, and filled with value */
1192 c
+=UTRIE_DATA_BLOCK_LENGTH
;
1193 } else if(block
==nullBlock
) {
1194 /* this is the all-initial-value block */
1195 if(prevValue
!=initialValue
) {
1197 if(!enumRange(context
, prev
, c
, prevValue
)) {
1201 prevBlock
=nullBlock
;
1203 prevValue
=initialValue
;
1205 c
+=UTRIE_DATA_BLOCK_LENGTH
;
1208 for(j
=0; j
<UTRIE_DATA_BLOCK_LENGTH
; ++j
) {
1209 value
=enumValue(context
, data32
!=NULL
? data32
[block
+j
] : idx
[block
+j
]);
1210 if(value
!=prevValue
) {
1212 if(!enumRange(context
, prev
, c
, prevValue
)) {
1217 /* the block is not filled with all the same value */
1226 } while(++i
<offset
);
1232 /* deliver last range */
1233 enumRange(context
, prev
, c
, prevValue
);