1 /* 2 ******************************************************************************* 3 * 4 * Copyright (C) 1999-2014, International Business Machines 5 * Corporation and others. All Rights Reserved. 6 * 7 ******************************************************************************* 8 * file name: collationweights.h 9 * encoding: US-ASCII 10 * tab size: 8 (not used) 11 * indentation:4 12 * 13 * created on: 2001mar08 as ucol_wgt.h 14 * created by: Markus W. Scherer 15 */ 16 17 #ifndef __COLLATIONWEIGHTS_H__ 18 #define __COLLATIONWEIGHTS_H__ 19 20 #include "unicode/utypes.h" 21 22 #if !UCONFIG_NO_COLLATION 23 24 #include "unicode/uobject.h" 25 26 U_NAMESPACE_BEGIN 27 28 /** 29 * Allocates n collation element weights between two exclusive limits. 30 * Used only internally by the collation tailoring builder. 31 */ 32 class U_I18N_API CollationWeights : public UMemory { 33 public: 34 CollationWeights(); 35 lengthOfWeight(uint32_t weight)36 static inline int32_t lengthOfWeight(uint32_t weight) { 37 if((weight&0xffffff)==0) { 38 return 1; 39 } else if((weight&0xffff)==0) { 40 return 2; 41 } else if((weight&0xff)==0) { 42 return 3; 43 } else { 44 return 4; 45 } 46 } 47 48 void initForPrimary(UBool compressible); 49 void initForSecondary(); 50 void initForTertiary(); 51 52 /** 53 * Determine heuristically 54 * what ranges to use for a given number of weights between (excluding) 55 * two limits. 56 * 57 * @param lowerLimit A collation element weight; the ranges will be filled to cover 58 * weights greater than this one. 59 * @param upperLimit A collation element weight; the ranges will be filled to cover 60 * weights less than this one. 61 * @param n The number of collation element weights w necessary such that 62 * lowerLimit<w<upperLimit in lexical order. 63 * @return TRUE if it is possible to fit n elements between the limits 64 */ 65 UBool allocWeights(uint32_t lowerLimit, uint32_t upperLimit, int32_t n); 66 67 /** 68 * Given a set of ranges calculated by allocWeights(), 69 * iterate through the weights. 70 * The ranges are modified to keep the current iteration state. 71 * 72 * @return The next weight in the ranges, or 0xffffffff if there is none left. 73 */ 74 uint32_t nextWeight(); 75 76 /** @internal */ 77 struct WeightRange { 78 uint32_t start, end; 79 int32_t length, count; 80 }; 81 82 private: 83 /** @return number of usable byte values for byte idx */ countBytes(int32_t idx)84 inline int32_t countBytes(int32_t idx) const { 85 return (int32_t)(maxBytes[idx] - minBytes[idx] + 1); 86 } 87 88 uint32_t incWeight(uint32_t weight, int32_t length) const; 89 uint32_t incWeightByOffset(uint32_t weight, int32_t length, int32_t offset) const; 90 void lengthenRange(WeightRange &range) const; 91 /** 92 * Takes two CE weights and calculates the 93 * possible ranges of weights between the two limits, excluding them. 94 * For weights with up to 4 bytes there are up to 2*4-1=7 ranges. 95 */ 96 UBool getWeightRanges(uint32_t lowerLimit, uint32_t upperLimit); 97 UBool allocWeightsInShortRanges(int32_t n, int32_t minLength); 98 UBool allocWeightsInMinLengthRanges(int32_t n, int32_t minLength); 99 100 int32_t middleLength; 101 uint32_t minBytes[5]; // for byte 1, 2, 3, 4 102 uint32_t maxBytes[5]; 103 WeightRange ranges[7]; 104 int32_t rangeIndex; 105 int32_t rangeCount; 106 }; 107 108 U_NAMESPACE_END 109 110 #endif // !UCONFIG_NO_COLLATION 111 #endif // __COLLATIONWEIGHTS_H__ 112