Coverage Report

Created: 2018-09-25 17:16

/Users/buildslave/jenkins/workspace/clang-stage2-coverage-R/llvm/include/llvm/ADT/edit_distance.h
Line
Count
Source
1
//===-- llvm/ADT/edit_distance.h - Array edit distance function --- C++ -*-===//
2
//
3
//                     The LLVM Compiler Infrastructure
4
//
5
// This file is distributed under the University of Illinois Open Source
6
// License. See LICENSE.TXT for details.
7
//
8
//===----------------------------------------------------------------------===//
9
//
10
// This file defines a Levenshtein distance function that works for any two
11
// sequences, with each element of each sequence being analogous to a character
12
// in a string.
13
//
14
//===----------------------------------------------------------------------===//
15
16
#ifndef LLVM_ADT_EDIT_DISTANCE_H
17
#define LLVM_ADT_EDIT_DISTANCE_H
18
19
#include "llvm/ADT/ArrayRef.h"
20
#include <algorithm>
21
#include <memory>
22
23
namespace llvm {
24
25
/// Determine the edit distance between two sequences.
26
///
27
/// \param FromArray the first sequence to compare.
28
///
29
/// \param ToArray the second sequence to compare.
30
///
31
/// \param AllowReplacements whether to allow element replacements (change one
32
/// element into another) as a single operation, rather than as two operations
33
/// (an insertion and a removal).
34
///
35
/// \param MaxEditDistance If non-zero, the maximum edit distance that this
36
/// routine is allowed to compute. If the edit distance will exceed that
37
/// maximum, returns \c MaxEditDistance+1.
38
///
39
/// \returns the minimum number of element insertions, removals, or (if
40
/// \p AllowReplacements is \c true) replacements needed to transform one of
41
/// the given sequences into the other. If zero, the sequences are identical.
42
template<typename T>
43
unsigned ComputeEditDistance(ArrayRef<T> FromArray, ArrayRef<T> ToArray,
44
                             bool AllowReplacements = true,
45
1.79M
                             unsigned MaxEditDistance = 0) {
46
1.79M
  // The algorithm implemented below is the "classic"
47
1.79M
  // dynamic-programming algorithm for computing the Levenshtein
48
1.79M
  // distance, which is described here:
49
1.79M
  //
50
1.79M
  //   http://en.wikipedia.org/wiki/Levenshtein_distance
51
1.79M
  //
52
1.79M
  // Although the algorithm is typically described using an m x n
53
1.79M
  // array, only one row plus one element are used at a time, so this
54
1.79M
  // implementation just keeps one vector for the row.  To update one entry,
55
1.79M
  // only the entries to the left, top, and top-left are needed.  The left
56
1.79M
  // entry is in Row[x-1], the top entry is what's in Row[x] from the last
57
1.79M
  // iteration, and the top-left entry is stored in Previous.
58
1.79M
  typename ArrayRef<T>::size_type m = FromArray.size();
59
1.79M
  typename ArrayRef<T>::size_type n = ToArray.size();
60
1.79M
61
1.79M
  const unsigned SmallBufferSize = 64;
62
1.79M
  unsigned SmallBuffer[SmallBufferSize];
63
1.79M
  std::unique_ptr<unsigned[]> Allocated;
64
1.79M
  unsigned *Row = SmallBuffer;
65
1.79M
  if (n + 1 > SmallBufferSize) {
66
10
    Row = new unsigned[n + 1];
67
10
    Allocated.reset(Row);
68
10
  }
69
1.79M
70
20.4M
  for (unsigned i = 1; i <= n; 
++i18.6M
)
71
18.6M
    Row[i] = i;
72
1.79M
73
11.4M
  for (typename ArrayRef<T>::size_type y = 1; y <= m; 
++y9.68M
) {
74
11.3M
    Row[0] = y;
75
11.3M
    unsigned BestThisRow = Row[0];
76
11.3M
77
11.3M
    unsigned Previous = y - 1;
78
174M
    for (typename ArrayRef<T>::size_type x = 1; x <= n; 
++x162M
) {
79
162M
      int OldRow = Row[x];
80
162M
      if (AllowReplacements) {
81
157M
        Row[x] = std::min(
82
157M
            Previous + (FromArray[y-1] == ToArray[x-1] ? 
0u9.21M
:
1u148M
),
83
157M
            std::min(Row[x-1], Row[x])+1);
84
157M
      }
85
5.56M
      else {
86
5.56M
        if (FromArray[y-1] == ToArray[x-1]) 
Row[x] = Previous353k
;
87
5.20M
        else Row[x] = std::min(Row[x-1], Row[x]) + 1;
88
5.56M
      }
89
162M
      Previous = OldRow;
90
162M
      BestThisRow = std::min(BestThisRow, Row[x]);
91
162M
    }
92
11.3M
93
11.3M
    if (MaxEditDistance && 
BestThisRow > MaxEditDistance11.3M
)
94
1.65M
      return MaxEditDistance + 1;
95
11.3M
  }
96
1.79M
97
1.79M
  unsigned Result = Row[n];
98
142k
  return Result;
99
1.79M
}
unsigned int llvm::ComputeEditDistance<char>(llvm::ArrayRef<char>, llvm::ArrayRef<char>, bool, unsigned int)
Line
Count
Source
45
1.78M
                             unsigned MaxEditDistance = 0) {
46
1.78M
  // The algorithm implemented below is the "classic"
47
1.78M
  // dynamic-programming algorithm for computing the Levenshtein
48
1.78M
  // distance, which is described here:
49
1.78M
  //
50
1.78M
  //   http://en.wikipedia.org/wiki/Levenshtein_distance
51
1.78M
  //
52
1.78M
  // Although the algorithm is typically described using an m x n
53
1.78M
  // array, only one row plus one element are used at a time, so this
54
1.78M
  // implementation just keeps one vector for the row.  To update one entry,
55
1.78M
  // only the entries to the left, top, and top-left are needed.  The left
56
1.78M
  // entry is in Row[x-1], the top entry is what's in Row[x] from the last
57
1.78M
  // iteration, and the top-left entry is stored in Previous.
58
1.78M
  typename ArrayRef<T>::size_type m = FromArray.size();
59
1.78M
  typename ArrayRef<T>::size_type n = ToArray.size();
60
1.78M
61
1.78M
  const unsigned SmallBufferSize = 64;
62
1.78M
  unsigned SmallBuffer[SmallBufferSize];
63
1.78M
  std::unique_ptr<unsigned[]> Allocated;
64
1.78M
  unsigned *Row = SmallBuffer;
65
1.78M
  if (n + 1 > SmallBufferSize) {
66
10
    Row = new unsigned[n + 1];
67
10
    Allocated.reset(Row);
68
10
  }
69
1.78M
70
20.4M
  for (unsigned i = 1; i <= n; 
++i18.6M
)
71
18.6M
    Row[i] = i;
72
1.78M
73
11.4M
  for (typename ArrayRef<T>::size_type y = 1; y <= m; 
++y9.67M
) {
74
11.3M
    Row[0] = y;
75
11.3M
    unsigned BestThisRow = Row[0];
76
11.3M
77
11.3M
    unsigned Previous = y - 1;
78
174M
    for (typename ArrayRef<T>::size_type x = 1; x <= n; 
++x162M
) {
79
162M
      int OldRow = Row[x];
80
162M
      if (AllowReplacements) {
81
157M
        Row[x] = std::min(
82
157M
            Previous + (FromArray[y-1] == ToArray[x-1] ? 
0u9.21M
:
1u148M
),
83
157M
            std::min(Row[x-1], Row[x])+1);
84
157M
      }
85
5.56M
      else {
86
5.56M
        if (FromArray[y-1] == ToArray[x-1]) 
Row[x] = Previous353k
;
87
5.20M
        else Row[x] = std::min(Row[x-1], Row[x]) + 1;
88
5.56M
      }
89
162M
      Previous = OldRow;
90
162M
      BestThisRow = std::min(BestThisRow, Row[x]);
91
162M
    }
92
11.3M
93
11.3M
    if (MaxEditDistance && 
BestThisRow > MaxEditDistance11.3M
)
94
1.65M
      return MaxEditDistance + 1;
95
11.3M
  }
96
1.78M
97
1.78M
  unsigned Result = Row[n];
98
133k
  return Result;
99
1.78M
}
unsigned int llvm::ComputeEditDistance<clang::IdentifierInfo const*>(llvm::ArrayRef<clang::IdentifierInfo const*>, llvm::ArrayRef<clang::IdentifierInfo const*>, bool, unsigned int)
Line
Count
Source
45
8.81k
                             unsigned MaxEditDistance = 0) {
46
8.81k
  // The algorithm implemented below is the "classic"
47
8.81k
  // dynamic-programming algorithm for computing the Levenshtein
48
8.81k
  // distance, which is described here:
49
8.81k
  //
50
8.81k
  //   http://en.wikipedia.org/wiki/Levenshtein_distance
51
8.81k
  //
52
8.81k
  // Although the algorithm is typically described using an m x n
53
8.81k
  // array, only one row plus one element are used at a time, so this
54
8.81k
  // implementation just keeps one vector for the row.  To update one entry,
55
8.81k
  // only the entries to the left, top, and top-left are needed.  The left
56
8.81k
  // entry is in Row[x-1], the top entry is what's in Row[x] from the last
57
8.81k
  // iteration, and the top-left entry is stored in Previous.
58
8.81k
  typename ArrayRef<T>::size_type m = FromArray.size();
59
8.81k
  typename ArrayRef<T>::size_type n = ToArray.size();
60
8.81k
61
8.81k
  const unsigned SmallBufferSize = 64;
62
8.81k
  unsigned SmallBuffer[SmallBufferSize];
63
8.81k
  std::unique_ptr<unsigned[]> Allocated;
64
8.81k
  unsigned *Row = SmallBuffer;
65
8.81k
  if (n + 1 > SmallBufferSize) {
66
0
    Row = new unsigned[n + 1];
67
0
    Allocated.reset(Row);
68
0
  }
69
8.81k
70
22.7k
  for (unsigned i = 1; i <= n; 
++i13.9k
)
71
13.9k
    Row[i] = i;
72
8.81k
73
19.0k
  for (typename ArrayRef<T>::size_type y = 1; y <= m; 
++y10.2k
) {
74
10.2k
    Row[0] = y;
75
10.2k
    unsigned BestThisRow = Row[0];
76
10.2k
77
10.2k
    unsigned Previous = y - 1;
78
26.8k
    for (typename ArrayRef<T>::size_type x = 1; x <= n; 
++x16.5k
) {
79
16.5k
      int OldRow = Row[x];
80
16.5k
      if (AllowReplacements) {
81
16.5k
        Row[x] = std::min(
82
16.5k
            Previous + (FromArray[y-1] == ToArray[x-1] ? 
0u1.13k
:
1u15.4k
),
83
16.5k
            std::min(Row[x-1], Row[x])+1);
84
16.5k
      }
85
0
      else {
86
0
        if (FromArray[y-1] == ToArray[x-1]) Row[x] = Previous;
87
0
        else Row[x] = std::min(Row[x-1], Row[x]) + 1;
88
0
      }
89
16.5k
      Previous = OldRow;
90
16.5k
      BestThisRow = std::min(BestThisRow, Row[x]);
91
16.5k
    }
92
10.2k
93
10.2k
    if (MaxEditDistance && 
BestThisRow > MaxEditDistance0
)
94
0
      return MaxEditDistance + 1;
95
10.2k
  }
96
8.81k
97
8.81k
  unsigned Result = Row[n];
98
8.81k
  return Result;
99
8.81k
}
100
101
} // End llvm namespace
102
103
#endif