ICU 70.1  70.1
bytestrie.h
Go to the documentation of this file.
1 // © 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
3 /*
4 *******************************************************************************
5 * Copyright (C) 2010-2012, International Business Machines
6 * Corporation and others. All Rights Reserved.
7 *******************************************************************************
8 * file name: bytestrie.h
9 * encoding: UTF-8
10 * tab size: 8 (not used)
11 * indentation:4
12 *
13 * created on: 2010sep25
14 * created by: Markus W. Scherer
15 */
16 
17 #ifndef __BYTESTRIE_H__
18 #define __BYTESTRIE_H__
19 
25 #include "unicode/utypes.h"
26 
27 #if U_SHOW_CPLUSPLUS_API
28 
29 #include "unicode/stringpiece.h"
30 #include "unicode/uobject.h"
31 #include "unicode/ustringtrie.h"
32 
33 class BytesTrieTest;
34 
35 U_NAMESPACE_BEGIN
36 
37 class ByteSink;
38 class BytesTrieBuilder;
39 class CharString;
40 class UVector32;
41 
55 class U_COMMON_API BytesTrie : public UMemory {
56 public:
71  BytesTrie(const void *trieBytes)
72  : ownedArray_(NULL), bytes_(static_cast<const uint8_t *>(trieBytes)),
73  pos_(bytes_), remainingMatchLength_(-1) {}
74 
80 
87  BytesTrie(const BytesTrie &other)
88  : ownedArray_(NULL), bytes_(other.bytes_),
89  pos_(other.pos_), remainingMatchLength_(other.remainingMatchLength_) {}
90 
97  pos_=bytes_;
98  remainingMatchLength_=-1;
99  return *this;
100  }
101 
110  uint64_t getState64() const {
111  return (static_cast<uint64_t>(remainingMatchLength_ + 2) << kState64RemainingShift) |
112  (uint64_t)(pos_ - bytes_);
113  }
114 
129  BytesTrie &resetToState64(uint64_t state) {
130  remainingMatchLength_ = static_cast<int32_t>(state >> kState64RemainingShift) - 2;
131  pos_ = bytes_ + (state & kState64PosMask);
132  return *this;
133  }
134 
140  class State : public UMemory {
141  public:
146  State() { bytes=NULL; }
147  private:
148  friend class BytesTrie;
149 
150  const uint8_t *bytes;
151  const uint8_t *pos;
152  int32_t remainingMatchLength;
153  };
154 
162  const BytesTrie &saveState(State &state) const {
163  state.bytes=bytes_;
164  state.pos=pos_;
165  state.remainingMatchLength=remainingMatchLength_;
166  return *this;
167  }
168 
179  BytesTrie &resetToState(const State &state) {
180  if(bytes_==state.bytes && bytes_!=NULL) {
181  pos_=state.pos;
182  remainingMatchLength_=state.remainingMatchLength;
183  }
184  return *this;
185  }
186 
194 
203  inline UStringTrieResult first(int32_t inByte) {
204  remainingMatchLength_=-1;
205  if(inByte<0) {
206  inByte+=0x100;
207  }
208  return nextImpl(bytes_, inByte);
209  }
210 
218  UStringTrieResult next(int32_t inByte);
219 
235  UStringTrieResult next(const char *s, int32_t length);
236 
246  inline int32_t getValue() const {
247  const uint8_t *pos=pos_;
248  int32_t leadByte=*pos++;
249  // U_ASSERT(leadByte>=kMinValueLead);
250  return readValue(pos, leadByte>>1);
251  }
252 
262  inline UBool hasUniqueValue(int32_t &uniqueValue) const {
263  const uint8_t *pos=pos_;
264  // Skip the rest of a pending linear-match node.
265  return pos!=NULL && findUniqueValue(pos+remainingMatchLength_+1, false, uniqueValue);
266  }
267 
276  int32_t getNextBytes(ByteSink &out) const;
277 
282  class U_COMMON_API Iterator : public UMemory {
283  public:
295  Iterator(const void *trieBytes, int32_t maxStringLength, UErrorCode &errorCode);
296 
308  Iterator(const BytesTrie &trie, int32_t maxStringLength, UErrorCode &errorCode);
309 
315 
322 
327  UBool hasNext() const;
328 
343  UBool next(UErrorCode &errorCode);
344 
354  int32_t getValue() const { return value_; }
355 
356  private:
357  UBool truncateAndStop();
358 
359  const uint8_t *branchNext(const uint8_t *pos, int32_t length, UErrorCode &errorCode);
360 
361  const uint8_t *bytes_;
362  const uint8_t *pos_;
363  const uint8_t *initialPos_;
364  int32_t remainingMatchLength_;
365  int32_t initialRemainingMatchLength_;
366 
367  CharString *str_;
368  int32_t maxLength_;
369  int32_t value_;
370 
371  // The stack stores pairs of integers for backtracking to another
372  // outbound edge of a branch node.
373  // The first integer is an offset from bytes_.
374  // The second integer has the str_->length() from before the node in bits 15..0,
375  // and the remaining branch length in bits 24..16. (Bits 31..25 are unused.)
376  // (We could store the remaining branch length minus 1 in bits 23..16 and not use bits 31..24,
377  // but the code looks more confusing that way.)
378  UVector32 *stack_;
379  };
380 
381 private:
382  friend class BytesTrieBuilder;
383  friend class ::BytesTrieTest;
384 
391  BytesTrie(void *adoptBytes, const void *trieBytes)
392  : ownedArray_(static_cast<uint8_t *>(adoptBytes)),
393  bytes_(static_cast<const uint8_t *>(trieBytes)),
394  pos_(bytes_), remainingMatchLength_(-1) {}
395 
396  // No assignment operator.
397  BytesTrie &operator=(const BytesTrie &other);
398 
399  inline void stop() {
400  pos_=NULL;
401  }
402 
403  // Reads a compact 32-bit integer.
404  // pos is already after the leadByte, and the lead byte is already shifted right by 1.
405  static int32_t readValue(const uint8_t *pos, int32_t leadByte);
406  static inline const uint8_t *skipValue(const uint8_t *pos, int32_t leadByte) {
407  // U_ASSERT(leadByte>=kMinValueLead);
408  if(leadByte>=(kMinTwoByteValueLead<<1)) {
409  if(leadByte<(kMinThreeByteValueLead<<1)) {
410  ++pos;
411  } else if(leadByte<(kFourByteValueLead<<1)) {
412  pos+=2;
413  } else {
414  pos+=3+((leadByte>>1)&1);
415  }
416  }
417  return pos;
418  }
419  static inline const uint8_t *skipValue(const uint8_t *pos) {
420  int32_t leadByte=*pos++;
421  return skipValue(pos, leadByte);
422  }
423 
424  // Reads a jump delta and jumps.
425  static const uint8_t *jumpByDelta(const uint8_t *pos);
426 
427  static inline const uint8_t *skipDelta(const uint8_t *pos) {
428  int32_t delta=*pos++;
429  if(delta>=kMinTwoByteDeltaLead) {
430  if(delta<kMinThreeByteDeltaLead) {
431  ++pos;
432  } else if(delta<kFourByteDeltaLead) {
433  pos+=2;
434  } else {
435  pos+=3+(delta&1);
436  }
437  }
438  return pos;
439  }
440 
441  static inline UStringTrieResult valueResult(int32_t node) {
442  return (UStringTrieResult)(USTRINGTRIE_INTERMEDIATE_VALUE-(node&kValueIsFinal));
443  }
444 
445  // Handles a branch node for both next(byte) and next(string).
446  UStringTrieResult branchNext(const uint8_t *pos, int32_t length, int32_t inByte);
447 
448  // Requires remainingLength_<0.
449  UStringTrieResult nextImpl(const uint8_t *pos, int32_t inByte);
450 
451  // Helper functions for hasUniqueValue().
452  // Recursively finds a unique value (or whether there is not a unique one)
453  // from a branch.
454  static const uint8_t *findUniqueValueFromBranch(const uint8_t *pos, int32_t length,
455  UBool haveUniqueValue, int32_t &uniqueValue);
456  // Recursively finds a unique value (or whether there is not a unique one)
457  // starting from a position on a node lead byte.
458  static UBool findUniqueValue(const uint8_t *pos, UBool haveUniqueValue, int32_t &uniqueValue);
459 
460  // Helper functions for getNextBytes().
461  // getNextBytes() when pos is on a branch node.
462  static void getNextBranchBytes(const uint8_t *pos, int32_t length, ByteSink &out);
463  static void append(ByteSink &out, int c);
464 
465  // BytesTrie data structure
466  //
467  // The trie consists of a series of byte-serialized nodes for incremental
468  // string/byte sequence matching. The root node is at the beginning of the trie data.
469  //
470  // Types of nodes are distinguished by their node lead byte ranges.
471  // After each node, except a final-value node, another node follows to
472  // encode match values or continue matching further bytes.
473  //
474  // Node types:
475  // - Value node: Stores a 32-bit integer in a compact, variable-length format.
476  // The value is for the string/byte sequence so far.
477  // One node bit indicates whether the value is final or whether
478  // matching continues with the next node.
479  // - Linear-match node: Matches a number of bytes.
480  // - Branch node: Branches to other nodes according to the current input byte.
481  // The node byte is the length of the branch (number of bytes to select from)
482  // minus 1. It is followed by a sub-node:
483  // - If the length is at most kMaxBranchLinearSubNodeLength, then
484  // there are length-1 (key, value) pairs and then one more comparison byte.
485  // If one of the key bytes matches, then the value is either a final value for
486  // the string/byte sequence so far, or a "jump" delta to the next node.
487  // If the last byte matches, then matching continues with the next node.
488  // (Values have the same encoding as value nodes.)
489  // - If the length is greater than kMaxBranchLinearSubNodeLength, then
490  // there is one byte and one "jump" delta.
491  // If the input byte is less than the sub-node byte, then "jump" by delta to
492  // the next sub-node which will have a length of length/2.
493  // (The delta has its own compact encoding.)
494  // Otherwise, skip the "jump" delta to the next sub-node
495  // which will have a length of length-length/2.
496 
497  // Node lead byte values.
498 
499  // 00..0f: Branch node. If node!=0 then the length is node+1, otherwise
500  // the length is one more than the next byte.
501 
502  // For a branch sub-node with at most this many entries, we drop down
503  // to a linear search.
504  static const int32_t kMaxBranchLinearSubNodeLength=5;
505 
506  // 10..1f: Linear-match node, match 1..16 bytes and continue reading the next node.
507  static const int32_t kMinLinearMatch=0x10;
508  static const int32_t kMaxLinearMatchLength=0x10;
509 
510  // 20..ff: Variable-length value node.
511  // If odd, the value is final. (Otherwise, intermediate value or jump delta.)
512  // Then shift-right by 1 bit.
513  // The remaining lead byte value indicates the number of following bytes (0..4)
514  // and contains the value's top bits.
515  static const int32_t kMinValueLead=kMinLinearMatch+kMaxLinearMatchLength; // 0x20
516  // It is a final value if bit 0 is set.
517  static const int32_t kValueIsFinal=1;
518 
519  // Compact value: After testing bit 0, shift right by 1 and then use the following thresholds.
520  static const int32_t kMinOneByteValueLead=kMinValueLead/2; // 0x10
521  static const int32_t kMaxOneByteValue=0x40; // At least 6 bits in the first byte.
522 
523  static const int32_t kMinTwoByteValueLead=kMinOneByteValueLead+kMaxOneByteValue+1; // 0x51
524  static const int32_t kMaxTwoByteValue=0x1aff;
525 
526  static const int32_t kMinThreeByteValueLead=kMinTwoByteValueLead+(kMaxTwoByteValue>>8)+1; // 0x6c
527  static const int32_t kFourByteValueLead=0x7e;
528 
529  // A little more than Unicode code points. (0x11ffff)
530  static const int32_t kMaxThreeByteValue=((kFourByteValueLead-kMinThreeByteValueLead)<<16)-1;
531 
532  static const int32_t kFiveByteValueLead=0x7f;
533 
534  // Compact delta integers.
535  static const int32_t kMaxOneByteDelta=0xbf;
536  static const int32_t kMinTwoByteDeltaLead=kMaxOneByteDelta+1; // 0xc0
537  static const int32_t kMinThreeByteDeltaLead=0xf0;
538  static const int32_t kFourByteDeltaLead=0xfe;
539  static const int32_t kFiveByteDeltaLead=0xff;
540 
541  static const int32_t kMaxTwoByteDelta=((kMinThreeByteDeltaLead-kMinTwoByteDeltaLead)<<8)-1; // 0x2fff
542  static const int32_t kMaxThreeByteDelta=((kFourByteDeltaLead-kMinThreeByteDeltaLead)<<16)-1; // 0xdffff
543 
544  // For getState64():
545  // The remainingMatchLength_ is -1..14=(kMaxLinearMatchLength=0x10)-2
546  // so we need at least 5 bits for that.
547  // We add 2 to store it as a positive value 1..16=kMaxLinearMatchLength.
548  static constexpr int32_t kState64RemainingShift = 59;
549  static constexpr uint64_t kState64PosMask = (UINT64_C(1) << kState64RemainingShift) - 1;
550 
551  uint8_t *ownedArray_;
552 
553  // Fixed value referencing the BytesTrie bytes.
554  const uint8_t *bytes_;
555 
556  // Iterator variables.
557 
558  // Pointer to next trie byte to read. NULL if no more matches.
559  const uint8_t *pos_;
560  // Remaining length of a linear-match node, minus 1. Negative if not in such a node.
561  int32_t remainingMatchLength_;
562 };
563 
564 U_NAMESPACE_END
565 
566 #endif /* U_SHOW_CPLUSPLUS_API */
567 
568 #endif // __BYTESTRIE_H__
A ByteSink can be filled with bytes.
Definition: bytestream.h:53
Builder class for BytesTrie.
Iterator for all of the (byte sequence, value) pairs in a BytesTrie.
Definition: bytestrie.h:282
StringPiece getString() const
Iterator & reset()
Resets this iterator to its initial state.
int32_t getValue() const
Definition: bytestrie.h:354
Iterator(const void *trieBytes, int32_t maxStringLength, UErrorCode &errorCode)
Iterates from the root of a byte-serialized BytesTrie.
UBool next(UErrorCode &errorCode)
Finds the next (byte sequence, value) pair if there is one.
Iterator(const BytesTrie &trie, int32_t maxStringLength, UErrorCode &errorCode)
Iterates from the current state of the specified BytesTrie.
BytesTrie state object, for saving a trie's current state and resetting the trie back to this state l...
Definition: bytestrie.h:140
State()
Constructs an empty State.
Definition: bytestrie.h:146
Light-weight, non-const reader class for a BytesTrie.
Definition: bytestrie.h:55
BytesTrie & resetToState64(uint64_t state)
Resets this trie to the saved state.
Definition: bytestrie.h:129
int32_t getValue() const
Returns a matching byte sequence's value if called immediately after current()/first()/next() returne...
Definition: bytestrie.h:246
UStringTrieResult current() const
Determines whether the byte sequence so far matches, whether it has a value, and whether another inpu...
UStringTrieResult next(const char *s, int32_t length)
Traverses the trie from the current state for this byte sequence.
UStringTrieResult first(int32_t inByte)
Traverses the trie from the initial state for this input byte.
Definition: bytestrie.h:203
UBool hasUniqueValue(int32_t &uniqueValue) const
Determines whether all byte sequences reachable from the current state map to the same value.
Definition: bytestrie.h:262
BytesTrie & resetToState(const State &state)
Resets this trie to the saved state.
Definition: bytestrie.h:179
BytesTrie & reset()
Resets this trie to its initial state.
Definition: bytestrie.h:96
~BytesTrie()
Destructor.
BytesTrie(const BytesTrie &other)
Copy constructor, copies the other trie reader object and its state, but not the byte array which wil...
Definition: bytestrie.h:87
int32_t getNextBytes(ByteSink &out) const
Finds each byte which continues the byte sequence from the current state.
const BytesTrie & saveState(State &state) const
Saves the state of this trie.
Definition: bytestrie.h:162
uint64_t getState64() const
Returns the state of this trie as a 64-bit integer.
Definition: bytestrie.h:110
BytesTrie(const void *trieBytes)
Constructs a BytesTrie reader instance.
Definition: bytestrie.h:71
UStringTrieResult next(int32_t inByte)
Traverses the trie from the current state for this input byte.
A string-like object that points to a sized piece of memory.
Definition: stringpiece.h:60
UMemory is the common ICU base class.
Definition: uobject.h:115
C++ API: StringPiece: Read-only byte string wrapper class.
#define UINT64_C(c)
Provides a platform independent way to specify an unsigned 64-bit integer constant.
Definition: umachine.h:241
int8_t UBool
The ICU boolean type, a signed-byte integer.
Definition: umachine.h:269
C++ API: Common ICU base class UObject.
C API: Helper definitions for dictionary trie APIs.
UStringTrieResult
Return values for BytesTrie::next(), UCharsTrie::next() and similar methods.
Definition: ustringtrie.h:35
@ USTRINGTRIE_INTERMEDIATE_VALUE
The input unit(s) continued a matching string and there is a value for the string so far.
Definition: ustringtrie.h:66
Basic definitions for ICU, for both C and C++ APIs.
#define NULL
Define NULL if necessary, to nullptr for C++ and to ((void *)0) for C.
Definition: utypes.h:188
UErrorCode
Standard ICU4C error code type, a substitute for exceptions.
Definition: utypes.h:415
#define U_COMMON_API
Set to export library symbols from inside the common library, and to import them from outside.
Definition: utypes.h:300