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
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
|
// Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include <vespa/vespalib/testkit/test_kit.h>
#include <vespa/searchlib/attribute/attributefactory.h>
#include <vespa/searchlib/attribute/attributevector.h>
#include <vespa/searchlib/attribute/integerbase.h>
#include <vespa/searchlib/attribute/stringbase.h>
#include <vespa/searchlib/features/setup.h>
#include <vespa/searchlib/features/tensor_from_labels_feature.h>
#include <vespa/searchlib/fef/fef.h>
#include <vespa/searchlib/fef/test/ftlib.h>
#include <vespa/searchlib/fef/test/indexenvironment.h>
#include <vespa/searchcommon/attribute/config.h>
#include <vespa/eval/eval/function.h>
#include <vespa/eval/eval/simple_value.h>
#include <vespa/eval/eval/tensor_spec.h>
#include <vespa/eval/eval/value.h>
#include <vespa/eval/eval/test/value_compare.h>
using search::feature_t;
using namespace search::fef;
using namespace search::fef::test;
using namespace search::features;
using search::AttributeFactory;
using search::IntegerAttribute;
using search::StringAttribute;
using vespalib::eval::Value;
using vespalib::eval::Function;
using vespalib::eval::TensorSpec;
using vespalib::eval::SimpleValue;
using AVC = search::attribute::Config;
using AVBT = search::attribute::BasicType;
using AVCT = search::attribute::CollectionType;
using AttributePtr = search::AttributeVector::SP;
using FTA = FtTestApp;
Value::UP make_tensor(const TensorSpec &spec) {
return SimpleValue::from_spec(spec);
}
Value::UP make_empty(const vespalib::string &type) {
return make_tensor(TensorSpec(type));
}
struct SetupFixture
{
TensorFromLabelsBlueprint blueprint;
IndexEnvironment indexEnv;
SetupFixture()
: blueprint(),
indexEnv()
{
}
};
TEST_F("require that blueprint can be created from factory", SetupFixture)
{
EXPECT_TRUE(FTA::assertCreateInstance(f.blueprint, "tensorFromLabels"));
}
TEST_F("require that setup fails if source spec is invalid", SetupFixture)
{
FTA::FT_SETUP_FAIL(f.blueprint, f.indexEnv, StringList().add("source(foo)"));
}
TEST_F("require that setup succeeds with attribute source", SetupFixture)
{
FTA::FT_SETUP_OK(f.blueprint, f.indexEnv, StringList().add("attribute(foo)"),
StringList(), StringList().add("tensor"));
}
TEST_F("require that setup succeeds with query source", SetupFixture)
{
FTA::FT_SETUP_OK(f.blueprint, f.indexEnv, StringList().add("query(foo)"),
StringList(), StringList().add("tensor"));
}
struct ExecFixture
{
BlueprintFactory factory;
FtFeatureTest test;
ExecFixture(const vespalib::string &feature)
: factory(),
test(factory, feature)
{
setup_search_features(factory);
setupAttributeVectors();
setupQueryEnvironment();
ASSERT_TRUE(test.setup());
}
void setupAttributeVectors() {
std::vector<AttributePtr> attrs;
attrs.push_back(AttributeFactory::createAttribute("astr", AVC(AVBT::STRING, AVCT::ARRAY)));
attrs.push_back(AttributeFactory::createAttribute("aint", AVC(AVBT::INT32, AVCT::ARRAY)));
attrs.push_back(AttributeFactory::createAttribute("wsstr", AVC(AVBT::STRING, AVCT::WSET)));
attrs.push_back(AttributeFactory::createAttribute("sint", AVC(AVBT::INT32, AVCT::SINGLE)));
for (const auto &attr : attrs) {
attr->addReservedDoc();
attr->addDocs(1);
test.getIndexEnv().getAttributeMap().add(attr);
}
StringAttribute *astr = static_cast<StringAttribute *>(attrs[0].get());
// Note that the weight parameter is not used
astr->append(1, "a", 0);
astr->append(1, "b", 0);
astr->append(1, "c", 0);
IntegerAttribute *aint = static_cast<IntegerAttribute *>(attrs[1].get());
aint->append(1, 3, 0);
aint->append(1, 5, 0);
aint->append(1, 7, 0);
IntegerAttribute *sint = static_cast<IntegerAttribute *>(attrs[3].get());
sint->update(1, 5);
for (const auto &attr : attrs) {
attr->commit();
}
}
void setupQueryEnvironment() {
test.getQueryEnv().getProperties().add("astr_query", "[d e f e]");
test.getQueryEnv().getProperties().add("aint_query", "[11 13 17]");
}
const Value &extractTensor(uint32_t docid) {
return test.resolveObjectFeature(docid);
}
const Value &execute() {
return extractTensor(1);
}
};
// Tests for attribute source:
TEST_F("require that array string attribute can be converted to tensor (default dimension)",
ExecFixture("tensorFromLabels(attribute(astr))"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(astr{})")
.add({{"astr", "a"}}, 1)
.add({{"astr", "b"}}, 1)
.add({{"astr", "c"}}, 1)), f.execute());
}
TEST_F("require that array string attribute can be converted to tensor (explicit dimension)",
ExecFixture("tensorFromLabels(attribute(astr),dim)"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(dim{})")
.add({{"dim", "a"}}, 1)
.add({{"dim", "b"}}, 1)
.add({{"dim", "c"}}, 1)), f.execute());
}
TEST_F("require that array integer attribute can be converted to tensor (default dimension)",
ExecFixture("tensorFromLabels(attribute(aint))"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(aint{})")
.add({{"aint", "7"}}, 1)
.add({{"aint", "3"}}, 1)
.add({{"aint", "5"}}, 1)), f.execute());
}
TEST_F("require that array attribute can be converted to tensor (explicit dimension)",
ExecFixture("tensorFromLabels(attribute(aint),dim)"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(dim{})")
.add({{"dim", "7"}}, 1)
.add({{"dim", "3"}}, 1)
.add({{"dim", "5"}}, 1)), f.execute());
}
TEST_F("require that single-value integer attribute can be converted to tensor (default dimension)",
ExecFixture("tensorFromLabels(attribute(sint))"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(sint{})")
.add({{"sint", "5"}}, 1)), f.execute());
}
TEST_F("require that single-value integer attribute can be converted to tensor (explicit dimension)",
ExecFixture("tensorFromLabels(attribute(sint),foobar)"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(foobar{})")
.add({{"foobar", "5"}}, 1)), f.execute());
}
TEST_F("require that empty tensor is created if attribute does not exists",
ExecFixture("tensorFromLabels(attribute(null))"))
{
EXPECT_EQUAL(*make_empty("tensor(null{})"), f.execute());
}
TEST_F("require that empty tensor is created if attribute type is not supported",
ExecFixture("tensorFromLabels(attribute(wsstr))"))
{
EXPECT_EQUAL(*make_empty("tensor(wsstr{})"), f.execute());
}
// Tests for query source:
TEST_F("require that string array from query can be converted to tensor (default dimension)",
ExecFixture("tensorFromLabels(query(astr_query))"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(astr_query{})")
.add({{"astr_query", "d"}}, 1)
.add({{"astr_query", "e"}}, 1)
.add({{"astr_query", "f"}}, 1)), f.execute());
}
TEST_F("require that integer array from query can be converted to tensor (default dimension)",
ExecFixture("tensorFromLabels(query(aint_query))"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(aint_query{})")
.add({{"aint_query", "13"}}, 1)
.add({{"aint_query", "17"}}, 1)
.add({{"aint_query", "11"}}, 1)), f.execute());
}
TEST_F("require that string array from query can be converted to tensor (explicit dimension)",
ExecFixture("tensorFromLabels(query(astr_query),dim)"))
{
EXPECT_EQUAL(*make_tensor(TensorSpec("tensor(dim{})")
.add({{"dim", "d"}}, 1)
.add({{"dim", "e"}}, 1)
.add({{"dim", "f"}}, 1)), f.execute());
}
TEST_F("require that empty tensor is created if query parameter is not found",
ExecFixture("tensorFromLabels(query(null))"))
{
EXPECT_EQUAL(*make_empty("tensor(null{})"), f.execute());
}
TEST_MAIN() { TEST_RUN_ALL(); }
|
