• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2 *************************************************************************
3 *   © 2016 and later: Unicode, Inc. and others.
4 *   License & terms of use: http://www.unicode.org/copyright.html#License
5 *************************************************************************
6 *************************************************************************
7 *   Copyright (C) 2014, International Business Machines
8 *   Corporation and others.  All Rights Reserved.
9 *************************************************************************
10 *   file name:  bitset.cpp
11 *   encoding:   UTF-8
12 *   tab size:   8 (not used)
13 *   indentation:4
14 *
15 *   created on: 2007jan15
16 *   created by: Markus Scherer
17 *
18 *   Idea for a "compiled", fast, read-only (immutable) version of a UnicodeSet
19 *   using a folded bit set consisting of a 1k-entry index table and a
20 *   compacted array of 64-bit words.
21 *   Uses a simple hash table for compaction.
22 *   Uses the original set for supplementary code points.
23 */
24 
25 #include "unicode/utypes.h"
26 #include "unicont.h"
27 #include "cmemory.h" // for UPRV_LENGTHOF
28 
29 /*
30  * Hash table for up to 1k 64-bit words, for 1 bit per BMP code point.
31  * Hashes 64-bit words and maps them to 16-bit integers which are
32  * assigned in order of new incoming words for subsequent storage
33  * in a contiguous array.
34  */
35 struct BMPBitHash : public UObject {
36     int64_t keys[0x800];  // 2k
37     uint16_t values[0x800];
38     uint16_t reverse[0x400];
39     uint16_t count;
40     const int32_t prime=1301;  // Less than 2k.
41 
BMPBitHashBMPBitHash42     BMPBitHash() : count(0) {
43         // Fill values[] with 0xffff.
44         uprv_memset(values, 0xff, sizeof(values));
45     }
46 
47     /*
48      * Map a key to an integer count.
49      * Map at most 1k=0x400 different keys with this data structure.
50      */
mapBMPBitHash51     uint16_t map(int64_t key) {
52         int32_t hash=(int32_t)(key>>55)&0x1ff;
53         hash^=(int32_t)(key>>44)&0x7ff;
54         hash^=(int32_t)(key>>33)&0x7ff;
55         hash^=(int32_t)(key>>22)&0x7ff;
56         hash^=(int32_t)(key>>11)&0x7ff;
57         hash^=(int32_t)key&0x7ff;
58         for(;;) {
59             if(values[hash]==0xffff) {
60                 // Unused slot.
61                 keys[hash]=key;
62                 reverse[count]=hash;
63                 return values[hash]=count++;
64             } else if(keys[hash]==key) {
65                 // Found a slot with this key.
66                 return values[hash];
67             } else {
68                 // Used slot with a different key, move to another slot.
69                 hash=(hash+prime)&0x7ff;
70             }
71         }
72     }
73 
countKeysBMPBitHash74     uint16_t countKeys() const { return count; }
75 
76     /*
77      * Invert the hash map: Fill an array of length countKeys() with the keys
78      * indexed by their mapped values.
79      */
invertBMPBitHash80     void invert(int64_t *k) const {
81         uint16_t i;
82 
83         for(i=0; i<count; ++i) {
84             k[i]=keys[reverse[i]];
85         }
86     }
87 };
88 
89 class BitSet : public UObject, public UnicodeContainable {
90 public:
BitSet(const UnicodeSet & set,UErrorCode & errorCode)91     BitSet(const UnicodeSet &set, UErrorCode &errorCode) : bits(shortBits), restSet(set.clone()) {
92         if(U_FAILURE(errorCode)) {
93             return;
94         }
95         BMPBitHash *bitHash=new BMPBitHash;
96         if(bitHash==NULL || restSet==NULL) {
97             errorCode=U_MEMORY_ALLOCATION_ERROR;
98             return;
99         }
100 
101         UnicodeSetIterator iter(set);
102         int64_t b;
103         UChar32 start, end;
104         int32_t prevIndex, i, j;
105 
106         b=0;  // Not necessary but makes compilers happy.
107         prevIndex=-1;
108         for(;;) {
109             if(iter.nextRange() && !iter.isString()) {
110                 start=iter.getCodepoint();
111                 end=iter.getCodepointEnd();
112             } else {
113                 start=0x10000;
114             }
115             i=start>>6;
116             if(prevIndex!=i) {
117                 // Finish the end of the previous range.
118                 if(prevIndex<0) {
119                     prevIndex=0;
120                 } else {
121                     index[prevIndex++]=bitHash->map(b);
122                 }
123                 // Fill all-zero entries between ranges.
124                 if(prevIndex<i) {
125                     uint16_t zero=bitHash->map(0);
126                     do {
127                         index[prevIndex++]=zero;
128                     } while(prevIndex<i);
129                 }
130                 b=0;
131             }
132             if(start>0xffff) {
133                 break;
134             }
135             b|=~((INT64_C(1)<<(start&0x3f))-1);
136             j=end>>6;
137             if(i<j) {
138                 // Set bits for the start of the range.
139                 index[i++]=bitHash->map(b);
140                 // Fill all-one entries inside the range.
141                 if(i<j) {
142                     uint16_t all=bitHash->map(INT64_C(0xffffffffffffffff));
143                     do {
144                         index[i++]=all;
145                     } while(i<j);
146                 }
147                 b=INT64_C(0xffffffffffffffff);
148             }
149             /* i==j */
150             b&=(INT64_C(1)<<(end&0x3f))-1;
151             prevIndex=j;
152         }
153 
154         if(bitHash->countKeys()>UPRV_LENGTHOF(shortBits)) {
155             bits=(int64_t *)uprv_malloc(bitHash->countKeys()*8);
156         }
157         if(bits!=NULL) {
158             bitHash->invert(bits);
159         } else {
160             bits=shortBits;
161             errorCode=U_MEMORY_ALLOCATION_ERROR;
162             return;
163         }
164 
165         latin1Set[0]=(uint32_t)bits[0];
166         latin1Set[1]=(uint32_t)(bits[0]>>32);
167         latin1Set[2]=(uint32_t)bits[1];
168         latin1Set[3]=(uint32_t)(bits[1]>>32);
169         latin1Set[4]=(uint32_t)bits[2];
170         latin1Set[5]=(uint32_t)(bits[2]>>32);
171         latin1Set[6]=(uint32_t)bits[3];
172         latin1Set[7]=(uint32_t)(bits[3]>>32);
173 
174         restSet.remove(0, 0xffff);
175     }
176 
~BitSet()177     ~BitSet() {
178         if(bits!=shortBits) {
179             uprv_free(bits);
180         }
181         delete restSet;
182     }
183 
contains(UChar32 c) const184     UBool contains(UChar32 c) const {
185         if((uint32_t)c<=0xff) {
186             return (UBool)((latin1Set[c>>5]&((uint32_t)1<<(c&0x1f)))!=0);
187         } else if((uint32_t)c<0xffff) {
188             return (UBool)((bits[c>>6]&(INT64_C(1)<<(c&0x3f)))!=0);
189         } else {
190             return restSet->contains(c);
191         }
192     }
193 
194 private:
195     uint16_t index[0x400];
196     int64_t shortBits[32];
197     int64_t *bits;
198 
199     uint32_t latin1Bits[8];
200 
201     UnicodeSet *restSet;
202 };
203