unstructured

Форк
0
298 строк · 9.5 Кб
1
import os
2
import pathlib
3
from unittest.mock import patch
4

5
import pytest
6
import requests
7

8
from test_unstructured.unit_utils import assert_round_trips_through_JSON, example_doc_path
9
from unstructured.chunking.title import chunk_by_title
10
from unstructured.documents.elements import ElementType, Title
11
from unstructured.partition.md import partition_md
12
from unstructured.partition.utils.constants import UNSTRUCTURED_INCLUDE_DEBUG_METADATA
13

14
DIRECTORY = pathlib.Path(__file__).parent.resolve()
15

16

17
def test_partition_md_from_filename():
18
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
19
    elements = partition_md(filename=filename)
20
    assert "PageBreak" not in [elem.category for elem in elements]
21
    assert len(elements) > 0
22
    for element in elements:
23
        assert element.metadata.filename == "README.md"
24
    if UNSTRUCTURED_INCLUDE_DEBUG_METADATA:
25
        assert {element.metadata.detection_origin for element in elements} == {"md"}
26

27

28
def test_partition_md_from_filename_returns_uns_elements():
29
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
30
    elements = partition_md(filename=filename)
31
    assert len(elements) > 0
32
    assert isinstance(elements[0], Title)
33

34

35
def test_partition_md_from_filename_with_metadata_filename():
36
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
37
    elements = partition_md(filename=filename, metadata_filename="test")
38
    assert "PageBreak" not in [elem.category for elem in elements]
39
    assert len(elements) > 0
40
    for element in elements:
41
        assert element.metadata.filename == "test"
42

43

44
def test_partition_md_from_file():
45
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
46
    with open(filename) as f:
47
        elements = partition_md(file=f)
48
    assert len(elements) > 0
49
    for element in elements:
50
        assert element.metadata.filename is None
51

52

53
def test_partition_md_from_file_with_metadata_filename():
54
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
55
    with open(filename) as f:
56
        elements = partition_md(file=f, metadata_filename="test")
57
    assert len(elements) > 0
58
    assert all(element.metadata.filename == "test" for element in elements)
59

60

61
def test_partition_md_from_text():
62
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
63
    with open(filename) as f:
64
        text = f.read()
65
    elements = partition_md(text=text)
66
    assert len(elements) > 0
67
    for element in elements:
68
        assert element.metadata.filename is None
69

70

71
class MockResponse:
72
    def __init__(self, text, status_code, headers={}):
73
        self.text = text
74
        self.status_code = status_code
75
        self.ok = status_code < 300
76
        self.headers = headers
77

78

79
def test_partition_md_from_url():
80
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
81
    with open(filename) as f:
82
        text = f.read()
83

84
    response = MockResponse(
85
        text=text,
86
        status_code=200,
87
        headers={"Content-Type": "text/markdown"},
88
    )
89
    with patch.object(requests, "get", return_value=response) as _:
90
        elements = partition_md(url="https://fake.url")
91

92
    assert len(elements) > 0
93
    for element in elements:
94
        assert element.metadata.filename is None
95

96

97
def test_partition_md_from_url_raises_with_bad_status_code():
98
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
99
    with open(filename) as f:
100
        text = f.read()
101

102
    response = MockResponse(
103
        text=text,
104
        status_code=500,
105
        headers={"Content-Type": "text/html"},
106
    )
107
    with patch.object(requests, "get", return_value=response) as _, pytest.raises(ValueError):
108
        partition_md(url="https://fake.url")
109

110

111
def test_partition_md_from_url_raises_with_bad_content_type():
112
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
113
    with open(filename) as f:
114
        text = f.read()
115

116
    response = MockResponse(
117
        text=text,
118
        status_code=200,
119
        headers={"Content-Type": "application/json"},
120
    )
121
    with patch.object(requests, "get", return_value=response) as _, pytest.raises(ValueError):
122
        partition_md(url="https://fake.url")
123

124

125
def test_partition_md_raises_with_none_specified():
126
    with pytest.raises(ValueError):
127
        partition_md()
128

129

130
def test_partition_md_raises_with_too_many_specified():
131
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
132
    with open(filename) as f:
133
        text = f.read()
134

135
    with pytest.raises(ValueError):
136
        partition_md(filename=filename, text=text)
137

138

139
def test_partition_md_from_filename_exclude_metadata():
140
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
141
    elements = partition_md(filename=filename, include_metadata=False)
142
    for i in range(len(elements)):
143
        assert elements[i].metadata.to_dict() == {}
144

145

146
def test_partition_md_from_file_exclude_metadata():
147
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
148
    with open(filename) as f:
149
        elements = partition_md(file=f, include_metadata=False)
150
    for i in range(len(elements)):
151
        assert elements[i].metadata.to_dict() == {}
152

153

154
def test_partition_md_from_text_exclude_metadata():
155
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "README.md")
156
    with open(filename) as f:
157
        text = f.read()
158
    elements = partition_md(text=text, include_metadata=False)
159
    for i in range(len(elements)):
160
        assert elements[i].metadata.to_dict() == {}
161

162

163
def test_partition_md_metadata_date(
164
    mocker,
165
    filename="example-docs/README.md",
166
):
167
    mocked_last_modification_date = "2029-07-05T09:24:28"
168

169
    mocker.patch(
170
        "unstructured.partition.md.get_last_modified_date",
171
        return_value=mocked_last_modification_date,
172
    )
173

174
    elements = partition_md(
175
        filename=filename,
176
    )
177

178
    assert elements[0].metadata.last_modified == mocked_last_modification_date
179

180

181
def test_partition_md_with_custom_metadata_date(
182
    mocker,
183
    filename="example-docs/README.md",
184
):
185
    mocked_last_modification_date = "2029-07-05T09:24:28"
186
    expected_last_modification_date = "2020-07-05T09:24:28"
187

188
    mocker.patch(
189
        "unstructured.partition.md.get_last_modified_date",
190
        return_value=mocked_last_modification_date,
191
    )
192

193
    elements = partition_md(
194
        filename=filename,
195
        metadata_last_modified=expected_last_modification_date,
196
    )
197

198
    assert elements[0].metadata.last_modified == expected_last_modification_date
199

200

201
def test_partition_md_from_file_metadata_date(
202
    mocker,
203
    filename="example-docs/README.md",
204
):
205
    mocked_last_modification_date = "2029-07-05T09:24:28"
206

207
    mocker.patch(
208
        "unstructured.partition.md.get_last_modified_date_from_file",
209
        return_value=mocked_last_modification_date,
210
    )
211

212
    with open(filename, "rb") as f:
213
        elements = partition_md(
214
            file=f,
215
        )
216

217
    assert elements[0].metadata.last_modified == mocked_last_modification_date
218

219

220
def test_partition_md_from_file_with_custom_metadata_date(
221
    mocker,
222
    filename="example-docs/README.md",
223
):
224
    mocked_last_modification_date = "2029-07-05T09:24:28"
225
    expected_last_modification_date = "2020-07-05T09:24:28"
226

227
    mocker.patch(
228
        "unstructured.partition.md.get_last_modified_date_from_file",
229
        return_value=mocked_last_modification_date,
230
    )
231

232
    with open(filename, "rb") as f:
233
        elements = partition_md(file=f, metadata_last_modified=expected_last_modification_date)
234

235
    assert elements[0].metadata.last_modified == expected_last_modification_date
236

237

238
def test_partition_md_from_text_metadata_date(
239
    filename="example-docs/README.md",
240
):
241
    with open(filename) as f:
242
        text = f.read()
243

244
    elements = partition_md(
245
        text=text,
246
    )
247

248
    assert elements[0].metadata.last_modified is None
249

250

251
def test_partition_md_from_text_with_custom_metadata_date(
252
    filename="example-docs/README.md",
253
):
254
    expected_last_modification_date = "2020-07-05T09:24:28"
255

256
    with open(filename) as f:
257
        text = f.read()
258

259
    elements = partition_md(text=text, metadata_last_modified=expected_last_modification_date)
260

261
    assert elements[0].metadata.last_modified == expected_last_modification_date
262

263

264
def test_partition_md_with_json():
265
    with open(example_doc_path("README.md")) as f:
266
        text = f.read()
267
    elements = partition_md(text=text)
268
    assert_round_trips_through_JSON(elements)
269

270

271
def test_add_chunking_strategy_by_title_on_partition_md(
272
    filename="example-docs/README.md",
273
):
274
    elements = partition_md(filename=filename)
275
    chunk_elements = partition_md(filename, chunking_strategy="by_title")
276
    chunks = chunk_by_title(elements)
277
    assert chunk_elements != elements
278
    assert chunk_elements == chunks
279

280

281
def test_partition_md_element_metadata_has_languages():
282
    filename = "example-docs/README.md"
283
    elements = partition_md(filename=filename)
284
    assert elements[0].metadata.languages == ["eng"]
285

286

287
def test_partition_md_respects_detect_language_per_element():
288
    filename = "example-docs/language-docs/eng_spa_mult.md"
289
    elements = partition_md(filename=filename, detect_language_per_element=True)
290
    langs = [element.metadata.languages for element in elements]
291
    assert langs == [["eng"], ["spa", "eng"], ["eng"], ["eng"], ["spa"]]
292

293

294
def test_partition_md_parse_table():
295
    filename = os.path.join(DIRECTORY, "..", "..", "..", "example-docs", "simple-table.md")
296
    elements = partition_md(filename=filename)
297
    assert len(elements) > 0
298
    assert elements[0].category == ElementType.TABLE
299

Использование cookies

Мы используем файлы cookie в соответствии с Политикой конфиденциальности и Политикой использования cookies.

Нажимая кнопку «Принимаю», Вы даете АО «СберТех» согласие на обработку Ваших персональных данных в целях совершенствования нашего веб-сайта и Сервиса GitVerse, а также повышения удобства их использования.

Запретить использование cookies Вы можете самостоятельно в настройках Вашего браузера.