|
1
|
|
|
"""Support for document corpora in JSON format""" |
|
2
|
|
|
|
|
3
|
|
|
import functools |
|
4
|
|
|
import json |
|
5
|
|
|
import os.path |
|
6
|
|
|
from importlib.resources import files |
|
7
|
|
|
|
|
8
|
|
|
import annif |
|
9
|
|
|
from annif.vocab import SubjectIndex |
|
10
|
|
|
|
|
11
|
|
|
from .types import Document, SubjectSet |
|
12
|
|
|
|
|
13
|
|
|
logger = annif.logger |
|
14
|
|
|
|
|
15
|
|
|
|
|
16
|
|
|
@functools.lru_cache(maxsize=1) |
|
17
|
|
|
def _get_json_schema(schema_name): |
|
18
|
|
|
schema_path = files("annif.schemas").joinpath(schema_name) |
|
19
|
|
|
with schema_path.open("r", encoding="utf-8") as schema_file: |
|
20
|
|
|
return json.load(schema_file) |
|
21
|
|
|
|
|
22
|
|
|
|
|
23
|
|
|
def _subjects_to_subject_set(subjects, subject_index, language): |
|
24
|
|
|
subject_ids = [] |
|
25
|
|
|
for subj in subjects: |
|
26
|
|
|
if "uri" in subj: |
|
27
|
|
|
subject_ids.append(subject_index.by_uri(subj["uri"])) |
|
28
|
|
|
else: |
|
29
|
|
|
subject_ids.append(subject_index.by_label(subj["label"], language)) |
|
30
|
|
|
return SubjectSet(subject_ids) |
|
31
|
|
|
|
|
32
|
|
|
|
|
33
|
|
|
def json_to_document( |
|
34
|
|
|
filename: str, |
|
35
|
|
|
json_data: str, |
|
36
|
|
|
subject_index: SubjectIndex | None, |
|
37
|
|
|
language: str, |
|
38
|
|
|
require_subjects: bool, |
|
39
|
|
|
) -> Document | None: |
|
40
|
|
|
|
|
41
|
|
|
import jsonschema |
|
42
|
|
|
|
|
43
|
|
|
try: |
|
44
|
|
|
data = json.loads(json_data) |
|
45
|
|
|
except json.JSONDecodeError as err: |
|
46
|
|
|
logger.warning(f"JSON parsing failed for file {filename}: {err}") |
|
47
|
|
|
return None |
|
48
|
|
|
|
|
49
|
|
|
try: |
|
50
|
|
|
jsonschema.validate(instance=data, schema=_get_json_schema("document.json")) |
|
51
|
|
|
except jsonschema.ValidationError as err: |
|
52
|
|
|
logger.warning(f"JSON validation failed for file {filename}: {err.message}") |
|
53
|
|
|
return None |
|
54
|
|
|
|
|
55
|
|
|
if require_subjects or (subject_index is not None and "subjects" in data): |
|
56
|
|
|
subject_set = _subjects_to_subject_set( |
|
57
|
|
|
data.get("subjects", []), subject_index, language |
|
58
|
|
|
) |
|
59
|
|
|
if not subject_set: |
|
60
|
|
|
return None |
|
61
|
|
|
else: |
|
62
|
|
|
subject_set = None |
|
63
|
|
|
|
|
64
|
|
|
return Document( |
|
65
|
|
|
text=data.get("text", ""), |
|
66
|
|
|
metadata=data.get("metadata", {}), |
|
67
|
|
|
subject_set=subject_set, |
|
68
|
|
|
file_path=filename, |
|
69
|
|
|
document_id=data.get("document_id", None), |
|
70
|
|
|
) |
|
71
|
|
|
|
|
72
|
|
|
|
|
73
|
|
|
def json_file_to_document( |
|
74
|
|
|
filename: str, |
|
75
|
|
|
subject_index: SubjectIndex | None, |
|
76
|
|
|
language: str, |
|
77
|
|
|
require_subjects: bool, |
|
78
|
|
|
) -> Document | None: |
|
79
|
|
|
if os.path.getsize(filename) == 0: |
|
80
|
|
|
logger.warning(f"Skipping empty file {filename}") |
|
81
|
|
|
return None |
|
82
|
|
|
|
|
83
|
|
|
with open(filename, "r", encoding="utf-8") as jsonfile: |
|
84
|
|
|
json_data = jsonfile.read() |
|
85
|
|
|
|
|
86
|
|
|
return json_to_document( |
|
87
|
|
|
filename, json_data, subject_index, language, require_subjects |
|
88
|
|
|
) |
|
89
|
|
|
|
NatLibFi /
Annif
