]>
git.saurik.com Git - apple/icu.git/blob - icuSources/i18n/ucol_wgt.cpp
2 *******************************************************************************
4 * Copyright (C) 1999-2010, International Business Machines
5 * Corporation and others. All Rights Reserved.
7 *******************************************************************************
8 * file name: ucol_wgt.c
10 * tab size: 8 (not used)
13 * created on: 2001mar08
14 * created by: Markus W. Scherer
16 * This file contains code for allocating n collation element weights
17 * between two exclusive limits.
18 * It is used only internally by ucol_bld.
21 #include "unicode/utypes.h"
23 #if !UCONFIG_NO_COLLATION
34 /* collation element weight allocation -------------------------------------- */
36 /* helper functions for CE weights */
38 static U_INLINE
int32_t
39 lengthOfWeight(uint32_t weight
) {
40 if((weight
&0xffffff)==0) {
42 } else if((weight
&0xffff)==0) {
44 } else if((weight
&0xff)==0) {
51 static U_INLINE
uint32_t
52 getWeightTrail(uint32_t weight
, int32_t length
) {
53 return (uint32_t)(weight
>>(8*(4-length
)))&0xff;
56 static U_INLINE
uint32_t
57 setWeightTrail(uint32_t weight
, int32_t length
, uint32_t trail
) {
59 return (uint32_t)((weight
&(0xffffff00<<length
))|(trail
<<length
));
62 static U_INLINE
uint32_t
63 getWeightByte(uint32_t weight
, int32_t idx
) {
64 return getWeightTrail(weight
, idx
); /* same calculation */
67 static U_INLINE
uint32_t
68 setWeightByte(uint32_t weight
, int32_t idx
, uint32_t byte
) {
69 uint32_t mask
; /* 0xffffffff except a 00 "hole" for the index-th byte */
72 mask
=((uint32_t)0xffffffff)>>idx
;
74 mask
|=0xffffff00<<idx
;
75 return (uint32_t)((weight
&mask
)|(byte
<<idx
));
78 static U_INLINE
uint32_t
79 truncateWeight(uint32_t weight
, int32_t length
) {
80 return (uint32_t)(weight
&(0xffffffff<<(8*(4-length
))));
83 static U_INLINE
uint32_t
84 incWeightTrail(uint32_t weight
, int32_t length
) {
85 return (uint32_t)(weight
+(1UL<<(8*(4-length
))));
88 static U_INLINE
uint32_t
89 decWeightTrail(uint32_t weight
, int32_t length
) {
90 return (uint32_t)(weight
-(1UL<<(8*(4-length
))));
93 static U_INLINE
uint32_t
94 incWeight(uint32_t weight
, int32_t length
, uint32_t maxByte
) {
98 byte
=getWeightByte(weight
, length
);
100 return setWeightByte(weight
, length
, byte
+1);
102 /* roll over, set this byte to UCOL_BYTE_FIRST_TAILORED and increment the previous one */
103 weight
=setWeightByte(weight
, length
, UCOL_BYTE_FIRST_TAILORED
);
109 static U_INLINE
int32_t
110 lengthenRange(WeightRange
*range
, uint32_t maxByte
, uint32_t countBytes
) {
113 length
=range
->length2
+1;
114 range
->start
=setWeightTrail(range
->start
, length
, UCOL_BYTE_FIRST_TAILORED
);
115 range
->end
=setWeightTrail(range
->end
, length
, maxByte
);
116 range
->count2
*=countBytes
;
117 range
->length2
=length
;
121 /* for uprv_sortArray: sort ranges in weight order */
122 static int32_t U_CALLCONV
123 compareRanges(const void * /*context*/, const void *left
, const void *right
) {
126 l
=((const WeightRange
*)left
)->start
;
127 r
=((const WeightRange
*)right
)->start
;
138 * take two CE weights and calculate the
139 * possible ranges of weights between the two limits, excluding them
140 * for weights with up to 4 bytes there are up to 2*4-1=7 ranges
142 static U_INLINE
int32_t
143 getWeightRanges(uint32_t lowerLimit
, uint32_t upperLimit
,
144 uint32_t maxByte
, uint32_t countBytes
,
145 WeightRange ranges
[7]) {
146 WeightRange lower
[5], middle
, upper
[5]; /* [0] and [1] are not used - this simplifies indexing */
147 uint32_t weight
, trail
;
148 int32_t length
, lowerLength
, upperLength
, rangeCount
;
150 /* assume that both lowerLimit & upperLimit are not 0 */
152 /* get the lengths of the limits */
153 lowerLength
=lengthOfWeight(lowerLimit
);
154 upperLength
=lengthOfWeight(upperLimit
);
157 printf("length of lower limit 0x%08lx is %ld\n", lowerLimit
, lowerLength
);
158 printf("length of upper limit 0x%08lx is %ld\n", upperLimit
, upperLength
);
161 if(lowerLimit
>=upperLimit
) {
163 printf("error: no space between lower & upper limits\n");
168 /* check that neither is a prefix of the other */
169 if(lowerLength
<upperLength
) {
170 if(lowerLimit
==truncateWeight(upperLimit
, lowerLength
)) {
172 printf("error: lower limit 0x%08lx is a prefix of upper limit 0x%08lx\n", lowerLimit
, upperLimit
);
177 /* if the upper limit is a prefix of the lower limit then the earlier test lowerLimit>=upperLimit has caught it */
179 /* reset local variables */
180 uprv_memset(lower
, 0, sizeof(lower
));
181 uprv_memset(&middle
, 0, sizeof(middle
));
182 uprv_memset(upper
, 0, sizeof(upper
));
185 * With the limit lengths of 1..4, there are up to 7 ranges for allocation:
186 * range minimum length
195 * We are now going to calculate up to 7 ranges.
196 * Some of them will typically overlap, so we will then have to merge and eliminate ranges.
199 for(length
=lowerLength
; length
>=2; --length
) {
200 trail
=getWeightTrail(weight
, length
);
202 lower
[length
].start
=incWeightTrail(weight
, length
);
203 lower
[length
].end
=setWeightTrail(weight
, length
, maxByte
);
204 lower
[length
].length
=length
;
205 lower
[length
].count
=maxByte
-trail
;
207 weight
=truncateWeight(weight
, length
-1);
209 middle
.start
=incWeightTrail(weight
, 1);
212 for(length
=upperLength
; length
>=2; --length
) {
213 trail
=getWeightTrail(weight
, length
);
214 if(trail
>UCOL_BYTE_FIRST_TAILORED
) {
215 upper
[length
].start
=setWeightTrail(weight
, length
, UCOL_BYTE_FIRST_TAILORED
);
216 upper
[length
].end
=decWeightTrail(weight
, length
);
217 upper
[length
].length
=length
;
218 upper
[length
].count
=trail
-UCOL_BYTE_FIRST_TAILORED
;
220 weight
=truncateWeight(weight
, length
-1);
222 middle
.end
=decWeightTrail(weight
, 1);
224 /* set the middle range */
226 if(middle
.end
>=middle
.start
) {
227 middle
.count
=(int32_t)((middle
.end
-middle
.start
)>>24)+1;
229 /* eliminate overlaps */
232 /* remove the middle range */
235 /* reduce or remove the lower ranges that go beyond upperLimit */
236 for(length
=4; length
>=2; --length
) {
237 if(lower
[length
].count
>0 && upper
[length
].count
>0) {
238 start
=upper
[length
].start
;
239 end
=lower
[length
].end
;
241 if(end
>=start
|| incWeight(end
, length
, maxByte
)==start
) {
242 /* lower and upper ranges collide or are directly adjacent: merge these two and remove all shorter ranges */
243 start
=lower
[length
].start
;
244 end
=lower
[length
].end
=upper
[length
].end
;
246 * merging directly adjacent ranges needs to subtract the 0/1 gaps in between;
247 * it may result in a range with count>countBytes
250 (int32_t)(getWeightTrail(end
, length
)-getWeightTrail(start
, length
)+1+
251 countBytes
*(getWeightByte(end
, length
-1)-getWeightByte(start
, length
-1)));
252 upper
[length
].count
=0;
254 lower
[length
].count
=upper
[length
].count
=0;
264 for(length
=4; length
>=2; --length
) {
265 if(lower
[length
].count
>0) {
266 printf("lower[%ld] .start=0x%08lx .end=0x%08lx .count=%ld\n", length
, lower
[length
].start
, lower
[length
].end
, lower
[length
].count
);
270 printf("middle .start=0x%08lx .end=0x%08lx .count=%ld\n", middle
.start
, middle
.end
, middle
.count
);
272 for(length
=2; length
<=4; ++length
) {
273 if(upper
[length
].count
>0) {
274 printf("upper[%ld] .start=0x%08lx .end=0x%08lx .count=%ld\n", length
, upper
[length
].start
, upper
[length
].end
, upper
[length
].count
);
279 /* copy the ranges, shortest first, into the result array */
282 uprv_memcpy(ranges
, &middle
, sizeof(WeightRange
));
285 for(length
=2; length
<=4; ++length
) {
286 /* copy upper first so that later the middle range is more likely the first one to use */
287 if(upper
[length
].count
>0) {
288 uprv_memcpy(ranges
+rangeCount
, upper
+length
, sizeof(WeightRange
));
291 if(lower
[length
].count
>0) {
292 uprv_memcpy(ranges
+rangeCount
, lower
+length
, sizeof(WeightRange
));
300 * call getWeightRanges and then determine heuristically
301 * which ranges to use for a given number of weights between (excluding)
305 ucol_allocWeights(uint32_t lowerLimit
, uint32_t upperLimit
,
308 WeightRange ranges
[7]) {
309 /* number of usable byte values 3..maxByte */
310 uint32_t countBytes
=maxByte
-UCOL_BYTE_FIRST_TAILORED
+1;
312 uint32_t lengthCounts
[6]; /* [0] unused, [5] to make index checks unnecessary */
314 int32_t i
, rangeCount
, minLength
/*, maxLength*/;
316 /* countBytes to the power of index */
318 /* gcc requires explicit initialization */
320 powers
[1] = countBytes
;
321 powers
[2] = countBytes
*countBytes
;
322 powers
[3] = countBytes
*countBytes
*countBytes
;
323 powers
[4] = countBytes
*countBytes
*countBytes
*countBytes
;
329 rangeCount
=getWeightRanges(lowerLimit
, upperLimit
, maxByte
, countBytes
, ranges
);
332 printf("error: unable to get Weight ranges\n");
337 /* what is the maximum number of weights with these ranges? */
339 for(i
=0; i
<rangeCount
; ++i
) {
340 maxCount
+=(uint32_t)ranges
[i
].count
*powers
[4-ranges
[i
].length
];
344 printf("the maximum number of %lu weights is sufficient for n=%lu\n", maxCount
, n
);
348 printf("error: the maximum number of %lu weights is insufficient for n=%lu\n", maxCount
, n
);
353 /* set the length2 and count2 fields */
354 for(i
=0; i
<rangeCount
; ++i
) {
355 ranges
[i
].length2
=ranges
[i
].length
;
356 ranges
[i
].count2
=(uint32_t)ranges
[i
].count
;
359 /* try until we find suitably large ranges */
361 /* get the smallest number of bytes in a range */
362 minLength
=ranges
[0].length2
;
364 /* sum up the number of elements that fit into ranges of each byte length */
365 uprv_memset(lengthCounts
, 0, sizeof(lengthCounts
));
366 for(i
=0; i
<rangeCount
; ++i
) {
367 lengthCounts
[ranges
[i
].length2
]+=ranges
[i
].count2
;
370 /* now try to allocate n elements in the available short ranges */
371 if(n
<=(lengthCounts
[minLength
]+lengthCounts
[minLength
+1])) {
372 /* trivial cases, use the first few ranges */
376 maxCount
+=ranges
[rangeCount
].count2
;
380 printf("take first %ld ranges\n", rangeCount
);
383 } else if(n
<=ranges
[0].count2
*countBytes
) {
384 /* easy case, just make this one range large enough by lengthening it once more, possibly split it */
385 uint32_t count1
, count2
, power_1
, power
;
387 /*maxLength=minLength+1;*/
389 /* calculate how to split the range between maxLength-1 (count1) and maxLength (count2) */
390 power_1
=powers
[minLength
-ranges
[0].length
];
391 power
=power_1
*countBytes
;
392 count2
=(n
+power
-1)/power
;
393 count1
=ranges
[0].count
-count2
;
395 /* split the range */
397 printf("split the first range %ld:%ld\n", count1
, count2
);
402 /* lengthen the entire range to maxLength */
403 lengthenRange(ranges
, maxByte
, countBytes
);
405 /* really split the range */
408 /* create a new range with the end and initial and current length of the old one */
410 ranges
[1].end
=ranges
[0].end
;
411 ranges
[1].length
=ranges
[0].length
;
412 ranges
[1].length2
=minLength
;
414 /* set the end of the first range according to count1 */
416 byte
=getWeightByte(ranges
[0].start
, i
)+count1
-1;
419 * ranges[0].count and count1 may be >countBytes
420 * from merging adjacent ranges;
421 * byte>maxByte is possible
424 ranges
[0].end
=setWeightByte(ranges
[0].start
, i
, byte
);
425 } else /* byte>maxByte */ {
426 ranges
[0].end
=setWeightByte(incWeight(ranges
[0].start
, i
-1, maxByte
), i
, byte
-countBytes
);
429 /* set the bytes in the end weight at length+1..length2 to maxByte */
430 byte
=(maxByte
<<24)|(maxByte
<<16)|(maxByte
<<8)|maxByte
; /* this used to be 0xffffffff */
431 ranges
[0].end
=truncateWeight(ranges
[0].end
, i
)|
432 ((byte
>>(8*i
))&(byte
<<(8*(4-minLength
))));
434 /* set the start of the second range to immediately follow the end of the first one */
435 ranges
[1].start
=incWeight(ranges
[0].end
, minLength
, maxByte
);
437 /* set the count values (informational) */
438 ranges
[0].count
=count1
;
439 ranges
[1].count
=count2
;
441 ranges
[0].count2
=count1
*power_1
;
442 ranges
[1].count2
=count2
*power_1
; /* will be *countBytes when lengthened */
444 /* lengthen the second range to maxLength */
445 lengthenRange(ranges
+1, maxByte
, countBytes
);
450 /* no good match, lengthen all minLength ranges and iterate */
452 printf("lengthen the short ranges from %ld bytes to %ld and iterate\n", minLength
, minLength
+1);
454 for(i
=0; ranges
[i
].length2
==minLength
; ++i
) {
455 lengthenRange(ranges
+i
, maxByte
, countBytes
);
460 /* sort the ranges by weight values */
461 UErrorCode errorCode
=U_ZERO_ERROR
;
462 uprv_sortArray(ranges
, rangeCount
, sizeof(WeightRange
), compareRanges
, NULL
, FALSE
, &errorCode
);
463 /* ignore error code: we know that the internal sort function will not fail here */
467 puts("final ranges:");
468 for(i
=0; i
<rangeCount
; ++i
) {
469 printf("ranges[%ld] .start=0x%08lx .end=0x%08lx .length=%ld .length2=%ld .count=%ld .count2=%lu\n",
470 i
, ranges
[i
].start
, ranges
[i
].end
, ranges
[i
].length
, ranges
[i
].length2
, ranges
[i
].count
, ranges
[i
].count2
);
474 /* set maxByte in ranges[0] for ucol_nextWeight() */
475 ranges
[0].count
=maxByte
;
481 * given a set of ranges calculated by ucol_allocWeights(),
482 * iterate through the weights
485 ucol_nextWeight(WeightRange ranges
[], int32_t *pRangeCount
) {
486 if(*pRangeCount
<=0) {
489 uint32_t weight
, maxByte
;
491 /* get maxByte from the .count field */
492 maxByte
=ranges
[0].count
;
494 /* get the next weight */
495 weight
=ranges
[0].start
;
496 if(weight
==ranges
[0].end
) {
497 /* this range is finished, remove it and move the following ones up */
498 if(--*pRangeCount
>0) {
499 uprv_memmove(ranges
, ranges
+1, *pRangeCount
*sizeof(WeightRange
));
500 ranges
[0].count
=maxByte
; /* keep maxByte in ranges[0] */
503 /* increment the weight for the next value */
504 ranges
[0].start
=incWeight(weight
, ranges
[0].length2
, maxByte
);
511 #if 0 // #ifdef UCOL_DEBUG
514 testAlloc(uint32_t lowerLimit
, uint32_t upperLimit
, uint32_t n
, UBool enumerate
) {
515 WeightRange ranges
[8];
518 rangeCount
=ucol_allocWeights(lowerLimit
, upperLimit
, n
, ranges
);
523 weight
=ucol_nextWeight(ranges
, &rangeCount
);
524 if(weight
==0xffffffff) {
525 printf("error: 0xffffffff with %lu more weights to go\n", n
);
528 printf(" 0x%08lx\n", weight
);
535 main(int argc
, const char *argv
[]) {
538 testAlloc(0x364214fc, 0x44b87d23, 5, FALSE
);
539 testAlloc(0x36421500, 0x44b87d23, 5, FALSE
);
540 testAlloc(0x36421500, 0x44b87d23, 20, FALSE
);
541 testAlloc(0x36421500, 0x44b87d23, 13700, FALSE
);
542 testAlloc(0x36421500, 0x38b87d23, 1, FALSE
);
543 testAlloc(0x36421500, 0x38b87d23, 20, FALSE
);
544 testAlloc(0x36421500, 0x38b87d23, 200, TRUE
);
545 testAlloc(0x36421500, 0x38b87d23, 13700, FALSE
);
546 testAlloc(0x36421500, 0x37b87d23, 13700, FALSE
);
547 testAlloc(0x36ef1500, 0x37b87d23, 13700, FALSE
);
548 testAlloc(0x36421500, 0x36b87d23, 13700, FALSE
);
549 testAlloc(0x36b87122, 0x36b87d23, 13700, FALSE
);
550 testAlloc(0x49000000, 0x4a600000, 13700, FALSE
);
551 testAlloc(0x9fffffff, 0xd0000000, 13700, FALSE
);
552 testAlloc(0x9fffffff, 0xd0000000, 67400, FALSE
);
553 testAlloc(0x9fffffff, 0xa0030000, 67400, FALSE
);
554 testAlloc(0x9fffffff, 0xa0030000, 40000, FALSE
);
555 testAlloc(0xa0000000, 0xa0030000, 40000, FALSE
);
556 testAlloc(0xa0031100, 0xa0030000, 40000, FALSE
);
564 #endif /* #if !UCONFIG_NO_COLLATION */