chipper 0.4.2
Sign up to get free protection for your applications and to get access to all the features.
- data/README.rdoc +51 -0
- data/ext/extconf.rb +58 -0
- data/ext/libstemmer_c/Makefile +10 -0
- data/ext/libstemmer_c/examples/stemwords.c +209 -0
- data/ext/libstemmer_c/include/libstemmer.h +79 -0
- data/ext/libstemmer_c/libstemmer/libstemmer.c +95 -0
- data/ext/libstemmer_c/libstemmer/libstemmer_utf8.c +95 -0
- data/ext/libstemmer_c/libstemmer/modules.h +190 -0
- data/ext/libstemmer_c/libstemmer/modules_utf8.h +121 -0
- data/ext/libstemmer_c/mkinc.mak +82 -0
- data/ext/libstemmer_c/mkinc_utf8.mak +52 -0
- data/ext/libstemmer_c/runtime/api.c +66 -0
- data/ext/libstemmer_c/runtime/api.h +26 -0
- data/ext/libstemmer_c/runtime/header.h +58 -0
- data/ext/libstemmer_c/runtime/utilities.c +478 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_danish.c +337 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_danish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_dutch.c +624 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_dutch.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_english.c +1117 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_english.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_finnish.c +762 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_finnish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_french.c +1246 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_french.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_german.c +521 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_german.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_hungarian.c +1230 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_hungarian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_italian.c +1065 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_italian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_norwegian.c +297 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_norwegian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_porter.c +749 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_porter.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_portuguese.c +1017 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_portuguese.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_spanish.c +1093 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_spanish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_swedish.c +307 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_1_swedish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_2_romanian.c +998 -0
- data/ext/libstemmer_c/src_c/stem_ISO_8859_2_romanian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_KOI8_R_russian.c +700 -0
- data/ext/libstemmer_c/src_c/stem_KOI8_R_russian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_danish.c +339 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_danish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_dutch.c +634 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_dutch.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_english.c +1125 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_english.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_finnish.c +768 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_finnish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_french.c +1256 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_french.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_german.c +527 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_german.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_hungarian.c +1234 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_hungarian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_italian.c +1073 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_italian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_norwegian.c +299 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_norwegian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_porter.c +755 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_porter.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_portuguese.c +1023 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_portuguese.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_romanian.c +1004 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_romanian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_russian.c +694 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_russian.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_spanish.c +1097 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_spanish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_swedish.c +309 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_swedish.h +16 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_turkish.c +2205 -0
- data/ext/libstemmer_c/src_c/stem_UTF_8_turkish.h +16 -0
- data/ext/re2/bitstate.cc +378 -0
- data/ext/re2/compile.cc +1138 -0
- data/ext/re2/dfa.cc +2086 -0
- data/ext/re2/filtered_re2.cc +100 -0
- data/ext/re2/filtered_re2.h +99 -0
- data/ext/re2/hash.cc +231 -0
- data/ext/re2/mimics_pcre.cc +185 -0
- data/ext/re2/nfa.cc +709 -0
- data/ext/re2/onepass.cc +614 -0
- data/ext/re2/parse.cc +2202 -0
- data/ext/re2/perl_groups.cc +119 -0
- data/ext/re2/prefilter.cc +671 -0
- data/ext/re2/prefilter.h +105 -0
- data/ext/re2/prefilter_tree.cc +398 -0
- data/ext/re2/prefilter_tree.h +130 -0
- data/ext/re2/prog.cc +341 -0
- data/ext/re2/prog.h +376 -0
- data/ext/re2/re2.cc +1180 -0
- data/ext/re2/re2.h +837 -0
- data/ext/re2/regexp.cc +920 -0
- data/ext/re2/regexp.h +632 -0
- data/ext/re2/rune.cc +258 -0
- data/ext/re2/set.cc +113 -0
- data/ext/re2/set.h +55 -0
- data/ext/re2/simplify.cc +393 -0
- data/ext/re2/stringpiece.cc +87 -0
- data/ext/re2/stringpiece.h +182 -0
- data/ext/re2/tostring.cc +341 -0
- data/ext/re2/unicode_casefold.cc +469 -0
- data/ext/re2/unicode_casefold.h +75 -0
- data/ext/re2/unicode_groups.cc +4851 -0
- data/ext/re2/unicode_groups.h +64 -0
- data/ext/re2/valgrind.cc +24 -0
- data/ext/re2/variadic_function.h +346 -0
- data/ext/re2/walker-inl.h +244 -0
- data/ext/src/chipper.cc +626 -0
- data/ext/src/version.h +1 -0
- data/ext/stemmer.rb +40 -0
- data/ext/util/arena.h +103 -0
- data/ext/util/atomicops.h +79 -0
- data/ext/util/benchmark.h +41 -0
- data/ext/util/flags.h +27 -0
- data/ext/util/logging.h +78 -0
- data/ext/util/mutex.h +190 -0
- data/ext/util/pcre.h +679 -0
- data/ext/util/random.h +29 -0
- data/ext/util/sparse_array.h +451 -0
- data/ext/util/sparse_set.h +177 -0
- data/ext/util/test.h +57 -0
- data/ext/util/thread.h +26 -0
- data/ext/util/utf.h +43 -0
- data/ext/util/util.h +127 -0
- data/ext/util/valgrind.h +4517 -0
- data/test/helper.rb +5 -0
- data/test/test_entities.rb +57 -0
- data/test/test_tokens.rb +118 -0
- metadata +199 -0
@@ -0,0 +1,100 @@
|
|
1
|
+
// Copyright 2009 The RE2 Authors. All Rights Reserved.
|
2
|
+
// Use of this source code is governed by a BSD-style
|
3
|
+
// license that can be found in the LICENSE file.
|
4
|
+
|
5
|
+
#include <string>
|
6
|
+
#include "util/util.h"
|
7
|
+
#include "re2/filtered_re2.h"
|
8
|
+
#include "re2/prefilter.h"
|
9
|
+
#include "re2/prefilter_tree.h"
|
10
|
+
|
11
|
+
namespace re2 {
|
12
|
+
|
13
|
+
FilteredRE2::FilteredRE2()
|
14
|
+
: compiled_(false),
|
15
|
+
prefilter_tree_(new PrefilterTree()) {
|
16
|
+
}
|
17
|
+
|
18
|
+
FilteredRE2::~FilteredRE2() {
|
19
|
+
for (int i = 0; i < re2_vec_.size(); i++)
|
20
|
+
delete re2_vec_[i];
|
21
|
+
delete prefilter_tree_;
|
22
|
+
}
|
23
|
+
|
24
|
+
RE2::ErrorCode FilteredRE2::Add(const StringPiece& pattern,
|
25
|
+
const RE2::Options& options, int* id) {
|
26
|
+
RE2* re = new RE2(pattern, options);
|
27
|
+
RE2::ErrorCode code = re->error_code();
|
28
|
+
|
29
|
+
if (!re->ok()) {
|
30
|
+
LOG(ERROR) << "Couldn't compile regular expression, skipping: "
|
31
|
+
<< re << " due to error " << re->error();
|
32
|
+
delete re;
|
33
|
+
} else {
|
34
|
+
*id = re2_vec_.size();
|
35
|
+
re2_vec_.push_back(re);
|
36
|
+
}
|
37
|
+
|
38
|
+
return code;
|
39
|
+
}
|
40
|
+
|
41
|
+
void FilteredRE2::Compile(vector<string>* atoms) {
|
42
|
+
if (compiled_ || re2_vec_.size() == 0) {
|
43
|
+
LOG(INFO) << "C: " << compiled_ << " S:" << re2_vec_.size();
|
44
|
+
return;
|
45
|
+
}
|
46
|
+
|
47
|
+
for (int i = 0; i < re2_vec_.size(); i++) {
|
48
|
+
Prefilter* prefilter = Prefilter::FromRE2(re2_vec_[i]);
|
49
|
+
prefilter_tree_->Add(prefilter);
|
50
|
+
}
|
51
|
+
atoms->clear();
|
52
|
+
prefilter_tree_->Compile(atoms);
|
53
|
+
compiled_ = true;
|
54
|
+
}
|
55
|
+
|
56
|
+
int FilteredRE2::SlowFirstMatch(const StringPiece& text) const {
|
57
|
+
for (int i = 0; i < re2_vec_.size(); i++)
|
58
|
+
if (RE2::PartialMatch(text, *re2_vec_[i]))
|
59
|
+
return i;
|
60
|
+
return -1;
|
61
|
+
}
|
62
|
+
|
63
|
+
int FilteredRE2::FirstMatch(const StringPiece& text,
|
64
|
+
const vector<int>& atoms) const {
|
65
|
+
if (!compiled_) {
|
66
|
+
LOG(DFATAL) << "FirstMatch called before Compile";
|
67
|
+
return -1;
|
68
|
+
}
|
69
|
+
vector<int> regexps;
|
70
|
+
prefilter_tree_->RegexpsGivenStrings(atoms, ®exps);
|
71
|
+
for (int i = 0; i < regexps.size(); i++)
|
72
|
+
if (RE2::PartialMatch(text, *re2_vec_[regexps[i]]))
|
73
|
+
return regexps[i];
|
74
|
+
return -1;
|
75
|
+
}
|
76
|
+
|
77
|
+
bool FilteredRE2::AllMatches(
|
78
|
+
const StringPiece& text,
|
79
|
+
const vector<int>& atoms,
|
80
|
+
vector<int>* matching_regexps) const {
|
81
|
+
matching_regexps->clear();
|
82
|
+
vector<int> regexps;
|
83
|
+
prefilter_tree_->RegexpsGivenStrings(atoms, ®exps);
|
84
|
+
for (int i = 0; i < regexps.size(); i++)
|
85
|
+
if (RE2::PartialMatch(text, *re2_vec_[regexps[i]]))
|
86
|
+
matching_regexps->push_back(regexps[i]);
|
87
|
+
return !matching_regexps->empty();
|
88
|
+
}
|
89
|
+
|
90
|
+
void FilteredRE2::RegexpsGivenStrings(const vector<int>& matched_atoms,
|
91
|
+
vector<int>* passed_regexps) {
|
92
|
+
prefilter_tree_->RegexpsGivenStrings(matched_atoms, passed_regexps);
|
93
|
+
}
|
94
|
+
|
95
|
+
|
96
|
+
void FilteredRE2::PrintPrefilter(int regexpid) {
|
97
|
+
prefilter_tree_->PrintPrefilter(regexpid);
|
98
|
+
}
|
99
|
+
|
100
|
+
} // namespace re2
|
@@ -0,0 +1,99 @@
|
|
1
|
+
// Copyright 2009 The RE2 Authors. All Rights Reserved.
|
2
|
+
// Use of this source code is governed by a BSD-style
|
3
|
+
// license that can be found in the LICENSE file.
|
4
|
+
|
5
|
+
// The class FilteredRE2 is used as a wrapper to multiple RE2 regexps.
|
6
|
+
// It provides a prefilter mechanism that helps in cutting down the
|
7
|
+
// number of regexps that need to be actually searched.
|
8
|
+
//
|
9
|
+
// By design, it does not include a string matching engine. This is to
|
10
|
+
// allow the user of the class to use their favorite string match
|
11
|
+
// engine. The overall flow is: Add all the regexps using Add, then
|
12
|
+
// Compile the FilteredRE2. The compile returns strings that need to
|
13
|
+
// be matched. Note that all returned strings are lowercase. For
|
14
|
+
// applying regexps to a search text, the caller does the string
|
15
|
+
// matching using the strings returned. When doing the string match,
|
16
|
+
// note that the caller has to do that on lower cased version of the
|
17
|
+
// search text. Then call FirstMatch or AllMatches with a vector of
|
18
|
+
// indices of strings that were found in the text to get the actual
|
19
|
+
// regexp matches.
|
20
|
+
|
21
|
+
#ifndef RE2_FILTERED_RE2_H_
|
22
|
+
#define RE2_FILTERED_RE2_H_
|
23
|
+
|
24
|
+
#include <vector>
|
25
|
+
#include "re2/re2.h"
|
26
|
+
|
27
|
+
namespace re2 {
|
28
|
+
using std::vector;
|
29
|
+
|
30
|
+
class PrefilterTree;
|
31
|
+
|
32
|
+
class FilteredRE2 {
|
33
|
+
public:
|
34
|
+
FilteredRE2();
|
35
|
+
~FilteredRE2();
|
36
|
+
|
37
|
+
// Uses RE2 constructor to create a RE2 object (re). Returns
|
38
|
+
// re->error_code(). If error_code is other than NoError, then re is
|
39
|
+
// deleted and not added to re2_vec_.
|
40
|
+
RE2::ErrorCode Add(const StringPiece& pattern,
|
41
|
+
const RE2::Options& options,
|
42
|
+
int *id);
|
43
|
+
|
44
|
+
// Prepares the regexps added by Add for filtering. Returns a set
|
45
|
+
// of strings that the caller should check for in candidate texts.
|
46
|
+
// The returned strings are lowercased. When doing string matching,
|
47
|
+
// the search text should be lowercased first to find matching
|
48
|
+
// strings from the set of strings returned by Compile. Call after
|
49
|
+
// all Add calls are done.
|
50
|
+
void Compile(vector<string>* strings_to_match);
|
51
|
+
|
52
|
+
// Returns the index of the first matching regexp.
|
53
|
+
// Returns -1 on no match. Can be called prior to Compile.
|
54
|
+
// Does not do any filtering: simply tries to Match the
|
55
|
+
// regexps in a loop.
|
56
|
+
int SlowFirstMatch(const StringPiece& text) const;
|
57
|
+
|
58
|
+
// Returns the index of the first matching regexp.
|
59
|
+
// Returns -1 on no match. Compile has to be called before
|
60
|
+
// calling this.
|
61
|
+
int FirstMatch(const StringPiece& text,
|
62
|
+
const vector<int>& atoms) const;
|
63
|
+
|
64
|
+
// Returns the indices of all matching regexps, after first clearing
|
65
|
+
// matched_regexps.
|
66
|
+
bool AllMatches(const StringPiece& text,
|
67
|
+
const vector<int>& atoms,
|
68
|
+
vector<int>* matching_regexps) const;
|
69
|
+
|
70
|
+
// The number of regexps added.
|
71
|
+
int NumRegexps() const { return re2_vec_.size(); }
|
72
|
+
|
73
|
+
private:
|
74
|
+
|
75
|
+
// Get the individual RE2 objects. Useful for testing.
|
76
|
+
RE2* GetRE2(int regexpid) const { return re2_vec_[regexpid]; }
|
77
|
+
|
78
|
+
// Print prefilter.
|
79
|
+
void PrintPrefilter(int regexpid);
|
80
|
+
|
81
|
+
// Useful for testing and debugging.
|
82
|
+
void RegexpsGivenStrings(const vector<int>& matched_atoms,
|
83
|
+
vector<int>* passed_regexps);
|
84
|
+
|
85
|
+
// All the regexps in the FilteredRE2.
|
86
|
+
vector<RE2*> re2_vec_;
|
87
|
+
|
88
|
+
// Has the FilteredRE2 been compiled using Compile()
|
89
|
+
bool compiled_;
|
90
|
+
|
91
|
+
// An AND-OR tree of string atoms used for filtering regexps.
|
92
|
+
PrefilterTree* prefilter_tree_;
|
93
|
+
|
94
|
+
DISALLOW_EVIL_CONSTRUCTORS(FilteredRE2);
|
95
|
+
};
|
96
|
+
|
97
|
+
} // namespace re2
|
98
|
+
|
99
|
+
#endif // RE2_FILTERED_RE2_H_
|
data/ext/re2/hash.cc
ADDED
@@ -0,0 +1,231 @@
|
|
1
|
+
// Modified by Russ Cox to add "namespace re2".
|
2
|
+
// Also threw away all but hashword and hashword2.
|
3
|
+
// http://burtleburtle.net/bob/c/lookup3.c
|
4
|
+
|
5
|
+
/*
|
6
|
+
-------------------------------------------------------------------------------
|
7
|
+
lookup3.c, by Bob Jenkins, May 2006, Public Domain.
|
8
|
+
|
9
|
+
These are functions for producing 32-bit hashes for hash table lookup.
|
10
|
+
hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
|
11
|
+
are externally useful functions. Routines to test the hash are included
|
12
|
+
if SELF_TEST is defined. You can use this free for any purpose. It's in
|
13
|
+
the public domain. It has no warranty.
|
14
|
+
|
15
|
+
You probably want to use hashlittle(). hashlittle() and hashbig()
|
16
|
+
hash byte arrays. hashlittle() is is faster than hashbig() on
|
17
|
+
little-endian machines. Intel and AMD are little-endian machines.
|
18
|
+
On second thought, you probably want hashlittle2(), which is identical to
|
19
|
+
hashlittle() except it returns two 32-bit hashes for the price of one.
|
20
|
+
You could implement hashbig2() if you wanted but I haven't bothered here.
|
21
|
+
|
22
|
+
If you want to find a hash of, say, exactly 7 integers, do
|
23
|
+
a = i1; b = i2; c = i3;
|
24
|
+
mix(a,b,c);
|
25
|
+
a += i4; b += i5; c += i6;
|
26
|
+
mix(a,b,c);
|
27
|
+
a += i7;
|
28
|
+
final(a,b,c);
|
29
|
+
then use c as the hash value. If you have a variable length array of
|
30
|
+
4-byte integers to hash, use hashword(). If you have a byte array (like
|
31
|
+
a character string), use hashlittle(). If you have several byte arrays, or
|
32
|
+
a mix of things, see the comments above hashlittle().
|
33
|
+
|
34
|
+
Why is this so big? I read 12 bytes at a time into 3 4-byte integers,
|
35
|
+
then mix those integers. This is fast (you can do a lot more thorough
|
36
|
+
mixing with 12*3 instructions on 3 integers than you can with 3 instructions
|
37
|
+
on 1 byte), but shoehorning those bytes into integers efficiently is messy.
|
38
|
+
-------------------------------------------------------------------------------
|
39
|
+
*/
|
40
|
+
|
41
|
+
#include "util/util.h"
|
42
|
+
|
43
|
+
#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
|
44
|
+
|
45
|
+
/*
|
46
|
+
-------------------------------------------------------------------------------
|
47
|
+
mix -- mix 3 32-bit values reversibly.
|
48
|
+
|
49
|
+
This is reversible, so any information in (a,b,c) before mix() is
|
50
|
+
still in (a,b,c) after mix().
|
51
|
+
|
52
|
+
If four pairs of (a,b,c) inputs are run through mix(), or through
|
53
|
+
mix() in reverse, there are at least 32 bits of the output that
|
54
|
+
are sometimes the same for one pair and different for another pair.
|
55
|
+
This was tested for:
|
56
|
+
* pairs that differed by one bit, by two bits, in any combination
|
57
|
+
of top bits of (a,b,c), or in any combination of bottom bits of
|
58
|
+
(a,b,c).
|
59
|
+
* "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
|
60
|
+
the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
|
61
|
+
is commonly produced by subtraction) look like a single 1-bit
|
62
|
+
difference.
|
63
|
+
* the base values were pseudorandom, all zero but one bit set, or
|
64
|
+
all zero plus a counter that starts at zero.
|
65
|
+
|
66
|
+
Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
|
67
|
+
satisfy this are
|
68
|
+
4 6 8 16 19 4
|
69
|
+
9 15 3 18 27 15
|
70
|
+
14 9 3 7 17 3
|
71
|
+
Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
|
72
|
+
for "differ" defined as + with a one-bit base and a two-bit delta. I
|
73
|
+
used http://burtleburtle.net/bob/hash/avalanche.html to choose
|
74
|
+
the operations, constants, and arrangements of the variables.
|
75
|
+
|
76
|
+
This does not achieve avalanche. There are input bits of (a,b,c)
|
77
|
+
that fail to affect some output bits of (a,b,c), especially of a. The
|
78
|
+
most thoroughly mixed value is c, but it doesn't really even achieve
|
79
|
+
avalanche in c.
|
80
|
+
|
81
|
+
This allows some parallelism. Read-after-writes are good at doubling
|
82
|
+
the number of bits affected, so the goal of mixing pulls in the opposite
|
83
|
+
direction as the goal of parallelism. I did what I could. Rotates
|
84
|
+
seem to cost as much as shifts on every machine I could lay my hands
|
85
|
+
on, and rotates are much kinder to the top and bottom bits, so I used
|
86
|
+
rotates.
|
87
|
+
-------------------------------------------------------------------------------
|
88
|
+
*/
|
89
|
+
#define mix(a,b,c) \
|
90
|
+
{ \
|
91
|
+
a -= c; a ^= rot(c, 4); c += b; \
|
92
|
+
b -= a; b ^= rot(a, 6); a += c; \
|
93
|
+
c -= b; c ^= rot(b, 8); b += a; \
|
94
|
+
a -= c; a ^= rot(c,16); c += b; \
|
95
|
+
b -= a; b ^= rot(a,19); a += c; \
|
96
|
+
c -= b; c ^= rot(b, 4); b += a; \
|
97
|
+
}
|
98
|
+
|
99
|
+
/*
|
100
|
+
-------------------------------------------------------------------------------
|
101
|
+
final -- final mixing of 3 32-bit values (a,b,c) into c
|
102
|
+
|
103
|
+
Pairs of (a,b,c) values differing in only a few bits will usually
|
104
|
+
produce values of c that look totally different. This was tested for
|
105
|
+
* pairs that differed by one bit, by two bits, in any combination
|
106
|
+
of top bits of (a,b,c), or in any combination of bottom bits of
|
107
|
+
(a,b,c).
|
108
|
+
* "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
|
109
|
+
the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
|
110
|
+
is commonly produced by subtraction) look like a single 1-bit
|
111
|
+
difference.
|
112
|
+
* the base values were pseudorandom, all zero but one bit set, or
|
113
|
+
all zero plus a counter that starts at zero.
|
114
|
+
|
115
|
+
These constants passed:
|
116
|
+
14 11 25 16 4 14 24
|
117
|
+
12 14 25 16 4 14 24
|
118
|
+
and these came close:
|
119
|
+
4 8 15 26 3 22 24
|
120
|
+
10 8 15 26 3 22 24
|
121
|
+
11 8 15 26 3 22 24
|
122
|
+
-------------------------------------------------------------------------------
|
123
|
+
*/
|
124
|
+
#define final(a,b,c) \
|
125
|
+
{ \
|
126
|
+
c ^= b; c -= rot(b,14); \
|
127
|
+
a ^= c; a -= rot(c,11); \
|
128
|
+
b ^= a; b -= rot(a,25); \
|
129
|
+
c ^= b; c -= rot(b,16); \
|
130
|
+
a ^= c; a -= rot(c,4); \
|
131
|
+
b ^= a; b -= rot(a,14); \
|
132
|
+
c ^= b; c -= rot(b,24); \
|
133
|
+
}
|
134
|
+
|
135
|
+
namespace re2 {
|
136
|
+
|
137
|
+
/*
|
138
|
+
--------------------------------------------------------------------
|
139
|
+
This works on all machines. To be useful, it requires
|
140
|
+
-- that the key be an array of uint32_t's, and
|
141
|
+
-- that the length be the number of uint32_t's in the key
|
142
|
+
|
143
|
+
The function hashword() is identical to hashlittle() on little-endian
|
144
|
+
machines, and identical to hashbig() on big-endian machines,
|
145
|
+
except that the length has to be measured in uint32_ts rather than in
|
146
|
+
bytes. hashlittle() is more complicated than hashword() only because
|
147
|
+
hashlittle() has to dance around fitting the key bytes into registers.
|
148
|
+
--------------------------------------------------------------------
|
149
|
+
*/
|
150
|
+
uint32 hashword(
|
151
|
+
const uint32 *k, /* the key, an array of uint32_t values */
|
152
|
+
size_t length, /* the length of the key, in uint32_ts */
|
153
|
+
uint32 initval) /* the previous hash, or an arbitrary value */
|
154
|
+
{
|
155
|
+
uint32_t a,b,c;
|
156
|
+
|
157
|
+
/* Set up the internal state */
|
158
|
+
a = b = c = 0xdeadbeef + (((uint32_t)length)<<2) + initval;
|
159
|
+
|
160
|
+
/*------------------------------------------------- handle most of the key */
|
161
|
+
while (length > 3)
|
162
|
+
{
|
163
|
+
a += k[0];
|
164
|
+
b += k[1];
|
165
|
+
c += k[2];
|
166
|
+
mix(a,b,c);
|
167
|
+
length -= 3;
|
168
|
+
k += 3;
|
169
|
+
}
|
170
|
+
|
171
|
+
/*------------------------------------------- handle the last 3 uint32_t's */
|
172
|
+
switch(length) /* all the case statements fall through */
|
173
|
+
{
|
174
|
+
case 3 : c+=k[2];
|
175
|
+
case 2 : b+=k[1];
|
176
|
+
case 1 : a+=k[0];
|
177
|
+
final(a,b,c);
|
178
|
+
case 0: /* case 0: nothing left to add */
|
179
|
+
break;
|
180
|
+
}
|
181
|
+
/*------------------------------------------------------ report the result */
|
182
|
+
return c;
|
183
|
+
}
|
184
|
+
|
185
|
+
|
186
|
+
/*
|
187
|
+
--------------------------------------------------------------------
|
188
|
+
hashword2() -- same as hashword(), but take two seeds and return two
|
189
|
+
32-bit values. pc and pb must both be nonnull, and *pc and *pb must
|
190
|
+
both be initialized with seeds. If you pass in (*pb)==0, the output
|
191
|
+
(*pc) will be the same as the return value from hashword().
|
192
|
+
--------------------------------------------------------------------
|
193
|
+
*/
|
194
|
+
void hashword2 (
|
195
|
+
const uint32 *k, /* the key, an array of uint32_t values */
|
196
|
+
size_t length, /* the length of the key, in uint32_ts */
|
197
|
+
uint32 *pc, /* IN: seed OUT: primary hash value */
|
198
|
+
uint32 *pb) /* IN: more seed OUT: secondary hash value */
|
199
|
+
{
|
200
|
+
uint32_t a,b,c;
|
201
|
+
|
202
|
+
/* Set up the internal state */
|
203
|
+
a = b = c = 0xdeadbeef + ((uint32_t)(length<<2)) + *pc;
|
204
|
+
c += *pb;
|
205
|
+
|
206
|
+
/*------------------------------------------------- handle most of the key */
|
207
|
+
while (length > 3)
|
208
|
+
{
|
209
|
+
a += k[0];
|
210
|
+
b += k[1];
|
211
|
+
c += k[2];
|
212
|
+
mix(a,b,c);
|
213
|
+
length -= 3;
|
214
|
+
k += 3;
|
215
|
+
}
|
216
|
+
|
217
|
+
/*------------------------------------------- handle the last 3 uint32_t's */
|
218
|
+
switch(length) /* all the case statements fall through */
|
219
|
+
{
|
220
|
+
case 3 : c+=k[2];
|
221
|
+
case 2 : b+=k[1];
|
222
|
+
case 1 : a+=k[0];
|
223
|
+
final(a,b,c);
|
224
|
+
case 0: /* case 0: nothing left to add */
|
225
|
+
break;
|
226
|
+
}
|
227
|
+
/*------------------------------------------------------ report the result */
|
228
|
+
*pc=c; *pb=b;
|
229
|
+
}
|
230
|
+
|
231
|
+
} // namespace re2
|
@@ -0,0 +1,185 @@
|
|
1
|
+
// Copyright 2008 The RE2 Authors. All Rights Reserved.
|
2
|
+
// Use of this source code is governed by a BSD-style
|
3
|
+
// license that can be found in the LICENSE file.
|
4
|
+
|
5
|
+
// Determine whether this library should match PCRE exactly
|
6
|
+
// for a particular Regexp. (If so, the testing framework can
|
7
|
+
// check that it does.)
|
8
|
+
//
|
9
|
+
// This library matches PCRE except in these cases:
|
10
|
+
// * the regexp contains a repetition of an empty string,
|
11
|
+
// like (a*)* or (a*)+. In this case, PCRE will treat
|
12
|
+
// the repetition sequence as ending with an empty string,
|
13
|
+
// while this library does not.
|
14
|
+
// * Perl and PCRE differ on whether \v matches \n.
|
15
|
+
// For historical reasons, this library implements the Perl behavior.
|
16
|
+
// * Perl and PCRE allow $ in one-line mode to match either the very
|
17
|
+
// end of the text or just before a \n at the end of the text.
|
18
|
+
// This library requires it to match only the end of the text.
|
19
|
+
// * Similarly, Perl and PCRE do not allow ^ in multi-line mode to
|
20
|
+
// match the end of the text if the last character is a \n.
|
21
|
+
// This library does allow it.
|
22
|
+
//
|
23
|
+
// Regexp::MimicsPCRE checks for any of these conditions.
|
24
|
+
|
25
|
+
#include "util/util.h"
|
26
|
+
#include "re2/regexp.h"
|
27
|
+
#include "re2/walker-inl.h"
|
28
|
+
|
29
|
+
namespace re2 {
|
30
|
+
|
31
|
+
// Returns whether re might match an empty string.
|
32
|
+
static bool CanBeEmptyString(Regexp *re);
|
33
|
+
|
34
|
+
// Walker class to compute whether library handles a regexp
|
35
|
+
// exactly as PCRE would. See comment at top for conditions.
|
36
|
+
|
37
|
+
class PCREWalker : public Regexp::Walker<bool> {
|
38
|
+
public:
|
39
|
+
PCREWalker() {}
|
40
|
+
bool PostVisit(Regexp* re, bool parent_arg, bool pre_arg, bool* child_args,
|
41
|
+
int nchild_args);
|
42
|
+
|
43
|
+
bool ShortVisit(Regexp* re, bool a) {
|
44
|
+
// Should never be called: we use Walk not WalkExponential.
|
45
|
+
LOG(DFATAL) << "EmptyStringWalker::ShortVisit called";
|
46
|
+
return a;
|
47
|
+
}
|
48
|
+
};
|
49
|
+
|
50
|
+
// Called after visiting each of re's children and accumulating
|
51
|
+
// the return values in child_args. So child_args contains whether
|
52
|
+
// this library mimics PCRE for those subexpressions.
|
53
|
+
bool PCREWalker::PostVisit(Regexp* re, bool parent_arg, bool pre_arg,
|
54
|
+
bool* child_args, int nchild_args) {
|
55
|
+
// If children failed, so do we.
|
56
|
+
for (int i = 0; i < nchild_args; i++)
|
57
|
+
if (!child_args[i])
|
58
|
+
return false;
|
59
|
+
|
60
|
+
// Otherwise look for other reasons to fail.
|
61
|
+
switch (re->op()) {
|
62
|
+
// Look for repeated empty string.
|
63
|
+
case kRegexpStar:
|
64
|
+
case kRegexpPlus:
|
65
|
+
case kRegexpQuest:
|
66
|
+
if (CanBeEmptyString(re->sub()[0]))
|
67
|
+
return false;
|
68
|
+
break;
|
69
|
+
case kRegexpRepeat:
|
70
|
+
if (re->max() == -1 && CanBeEmptyString(re->sub()[0]))
|
71
|
+
return false;
|
72
|
+
break;
|
73
|
+
|
74
|
+
// Look for \v
|
75
|
+
case kRegexpLiteral:
|
76
|
+
if (re->rune() == '\v')
|
77
|
+
return false;
|
78
|
+
break;
|
79
|
+
|
80
|
+
// Look for $ in single-line mode.
|
81
|
+
case kRegexpEndText:
|
82
|
+
case kRegexpEmptyMatch:
|
83
|
+
if (re->parse_flags() & Regexp::WasDollar)
|
84
|
+
return false;
|
85
|
+
break;
|
86
|
+
|
87
|
+
// Look for ^ in multi-line mode.
|
88
|
+
case kRegexpBeginLine:
|
89
|
+
// No condition: in single-line mode ^ becomes kRegexpBeginText.
|
90
|
+
return false;
|
91
|
+
|
92
|
+
default:
|
93
|
+
break;
|
94
|
+
}
|
95
|
+
|
96
|
+
// Not proven guilty.
|
97
|
+
return true;
|
98
|
+
}
|
99
|
+
|
100
|
+
// Returns whether this regexp's behavior will mimic PCRE's exactly.
|
101
|
+
bool Regexp::MimicsPCRE() {
|
102
|
+
PCREWalker w;
|
103
|
+
return w.Walk(this, true);
|
104
|
+
}
|
105
|
+
|
106
|
+
|
107
|
+
// Walker class to compute whether a Regexp can match an empty string.
|
108
|
+
// It is okay to overestimate. For example, \b\B cannot match an empty
|
109
|
+
// string, because \b and \B are mutually exclusive, but this isn't
|
110
|
+
// that smart and will say it can. Spurious empty strings
|
111
|
+
// will reduce the number of regexps we sanity check against PCRE,
|
112
|
+
// but they won't break anything.
|
113
|
+
|
114
|
+
class EmptyStringWalker : public Regexp::Walker<bool> {
|
115
|
+
public:
|
116
|
+
EmptyStringWalker() { }
|
117
|
+
bool PostVisit(Regexp* re, bool parent_arg, bool pre_arg,
|
118
|
+
bool* child_args, int nchild_args);
|
119
|
+
|
120
|
+
bool ShortVisit(Regexp* re, bool a) {
|
121
|
+
// Should never be called: we use Walk not WalkExponential.
|
122
|
+
LOG(DFATAL) << "EmptyStringWalker::ShortVisit called";
|
123
|
+
return a;
|
124
|
+
}
|
125
|
+
|
126
|
+
private:
|
127
|
+
DISALLOW_EVIL_CONSTRUCTORS(EmptyStringWalker);
|
128
|
+
};
|
129
|
+
|
130
|
+
// Called after visiting re's children. child_args contains the return
|
131
|
+
// value from each of the children's PostVisits (i.e., whether each child
|
132
|
+
// can match an empty string). Returns whether this clause can match an
|
133
|
+
// empty string.
|
134
|
+
bool EmptyStringWalker::PostVisit(Regexp* re, bool parent_arg, bool pre_arg,
|
135
|
+
bool* child_args, int nchild_args) {
|
136
|
+
switch (re->op()) {
|
137
|
+
case kRegexpNoMatch: // never empty
|
138
|
+
case kRegexpLiteral:
|
139
|
+
case kRegexpAnyChar:
|
140
|
+
case kRegexpAnyByte:
|
141
|
+
case kRegexpCharClass:
|
142
|
+
case kRegexpLiteralString:
|
143
|
+
return false;
|
144
|
+
|
145
|
+
case kRegexpEmptyMatch: // always empty
|
146
|
+
case kRegexpBeginLine: // always empty, when they match
|
147
|
+
case kRegexpEndLine:
|
148
|
+
case kRegexpNoWordBoundary:
|
149
|
+
case kRegexpWordBoundary:
|
150
|
+
case kRegexpBeginText:
|
151
|
+
case kRegexpEndText:
|
152
|
+
case kRegexpStar: // can always be empty
|
153
|
+
case kRegexpQuest:
|
154
|
+
case kRegexpHaveMatch:
|
155
|
+
return true;
|
156
|
+
|
157
|
+
case kRegexpConcat: // can be empty if all children can
|
158
|
+
for (int i = 0; i < nchild_args; i++)
|
159
|
+
if (!child_args[i])
|
160
|
+
return false;
|
161
|
+
return true;
|
162
|
+
|
163
|
+
case kRegexpAlternate: // can be empty if any child can
|
164
|
+
for (int i = 0; i < nchild_args; i++)
|
165
|
+
if (child_args[i])
|
166
|
+
return true;
|
167
|
+
return false;
|
168
|
+
|
169
|
+
case kRegexpPlus: // can be empty if the child can
|
170
|
+
case kRegexpCapture:
|
171
|
+
return child_args[0];
|
172
|
+
|
173
|
+
case kRegexpRepeat: // can be empty if child can or is x{0}
|
174
|
+
return child_args[0] || re->min() == 0;
|
175
|
+
}
|
176
|
+
return false;
|
177
|
+
}
|
178
|
+
|
179
|
+
// Returns whether re can match an empty string.
|
180
|
+
static bool CanBeEmptyString(Regexp* re) {
|
181
|
+
EmptyStringWalker w;
|
182
|
+
return w.Walk(re, true);
|
183
|
+
}
|
184
|
+
|
185
|
+
} // namespace re2
|