1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
|
// Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include <vespa/vespalib/testkit/testapp.h>
#include <vespa/vespalib/util/size_literals.h>
#include <vespa/searchlib/bitcompression/compression.h>
#include <vespa/searchlib/bitcompression/countcompression.h>
#include <vespa/searchlib/bitcompression/pagedict4.h>
#include <vespa/searchlib/test/diskindex/threelevelcountbuffers.h>
#include <vespa/searchlib/test/diskindex/pagedict4_mem_writer.h>
#include <vespa/searchlib/test/diskindex/pagedict4_mem_seq_reader.h>
#include <vespa/searchlib/index/postinglistcounts.h>
#include <vespa/log/log.h>
LOG_SETUP("pagedict4_hugeword_cornercase_test");
using search::index::PostingListCounts;
using search::ComprFileWriteContext;
constexpr uint32_t minChunkDocs = 256_Ki;
constexpr uint32_t numWordIds = 64_Ki;
struct BitBuffer
{
using EncodeContext = search::bitcompression::PostingListCountFileEncodeContext;
EncodeContext encodeCtx;
ComprFileWriteContext writeCtx;
BitBuffer()
: encodeCtx(),
writeCtx(encodeCtx)
{
encodeCtx._minChunkDocs = minChunkDocs;
encodeCtx._numWordIds = numWordIds;
writeCtx.allocComprBuf();
encodeCtx.setWriteContext(&writeCtx);
encodeCtx.setupWrite(writeCtx);
assert(encodeCtx.getWriteOffset() == 0);
}
void clear() { encodeCtx.setupWrite(writeCtx); }
uint64_t getSize(const PostingListCounts &counts)
{
clear();
encodeCtx.writeCounts(counts);
return encodeCtx.getWriteOffset();
}
~BitBuffer() = default;
};
void addSegment(PostingListCounts &counts)
{
PostingListCounts::Segment lastseg = counts._segments.back();
PostingListCounts::Segment fillseg;
fillseg._bitLength = 4000000;
fillseg._numDocs = minChunkDocs;
fillseg._lastDoc = minChunkDocs;
counts._bitLength += fillseg._bitLength;
counts._numDocs += fillseg._numDocs;
counts._segments.back() = fillseg;
uint32_t lastDoc = counts._segments.size() * fillseg._numDocs;
counts._segments.back()._lastDoc = lastDoc;
counts._segments.push_back(lastseg);
counts._segments.back()._lastDoc = lastDoc + lastseg._numDocs;
}
PostingListCounts makeBaseCounts()
{
PostingListCounts counts;
PostingListCounts::Segment lastseg;
lastseg._bitLength = 100;
lastseg._numDocs = 10;
lastseg._lastDoc = 10;
counts._bitLength = lastseg._bitLength;
counts._numDocs = lastseg._numDocs;
counts._segments.push_back(lastseg);
addSegment(counts);
return counts;
}
PostingListCounts makeSegmentedCounts(uint32_t segments)
{
PostingListCounts counts = makeBaseCounts();
while (counts._segments.size() < segments) {
addSegment(counts);
}
return counts;
}
uint32_t
calcSegments(uint32_t maxLen)
{
BitBuffer bb;
PostingListCounts counts = makeBaseCounts();
uint32_t len = bb.getSize(counts);
while (len <= maxLen) {
addSegment(counts);
len = bb.getSize(counts);
}
return counts._segments.size() - 1;
}
/*
* Calculate posting list counts that compresses to wantLen bits.
*/
PostingListCounts makeCounts(uint32_t wantLen)
{
BitBuffer bb;
uint32_t segments = calcSegments(wantLen);
PostingListCounts counts = makeSegmentedCounts(segments);
PostingListCounts counts2 = makeSegmentedCounts(segments - 1);
uint32_t len = bb.getSize(counts);
uint32_t len2 = bb.getSize(counts2);
for (uint32_t i = 1; i + 2 < counts._segments.size(); ++i) {
counts._bitLength += counts._segments[0]._bitLength;
counts._segments[i]._bitLength += counts._segments[0]._bitLength;
counts2._bitLength += counts2._segments[0]._bitLength;
counts2._segments[i]._bitLength += counts2._segments[0]._bitLength;
len = bb.getSize(counts);
len2 = bb.getSize(counts2);
if (len == wantLen) {
return counts;
}
if (len2 == wantLen) {
return counts2;
}
}
LOG_ABORT("Could not calculate counts with wanted compressed length");
}
using StartOffset = search::bitcompression::PageDict4StartOffset;
using Writer = search::diskindex::test::PageDict4MemWriter;
using SeqReader = search::diskindex::test::PageDict4MemSeqReader;
/*
* Test corner case where a dictionary page has a single word, and the
* page header and compressed counts completely fills the page.
*/
void testPageSizedCounts()
{
uint32_t pageBitSize = 32_Ki;
uint32_t startBits = 15 * 3 + 12;
uint32_t ssPad = 64;
uint32_t spPad = 64;
uint32_t pPad = 64;
Writer w(minChunkDocs, numWordIds, ssPad, spPad, pPad);
PostingListCounts baseCounts = makeBaseCounts();
PostingListCounts largeCounts = makeCounts(pageBitSize - startBits);
w.addCounts("a", baseCounts);
w.addCounts("b", baseCounts);
w.addCounts("c", largeCounts);
w.addCounts("d", baseCounts);
w.addCounts("e", baseCounts);
w.flush();
SeqReader r(minChunkDocs, numWordIds, w._buffers);
uint64_t checkWordNum = 0;
PostingListCounts counts;
for (uint64_t wordNum = 1; wordNum < 7; ++wordNum) {
vespalib::string word;
counts.clear();
r.readCounts(word, checkWordNum, counts);
if (wordNum < 6) {
EXPECT_EQUAL(checkWordNum, wordNum);
if (wordNum == 3) {
EXPECT_TRUE(counts == largeCounts);
} else {
EXPECT_TRUE(counts == baseCounts);
}
} else {
EXPECT_GREATER(checkWordNum, 100u);
}
}
}
TEST("require that counts exactly filling dictionary page works")
{
testPageSizedCounts();
}
TEST_MAIN() { TEST_RUN_ALL(); }
|
