|
1
|
|
|
"""Vocabulary management functionality for Annif""" |
|
2
|
|
|
|
|
3
|
|
|
import os.path |
|
4
|
|
|
import re |
|
5
|
|
|
import annif |
|
6
|
|
|
import annif.corpus |
|
7
|
|
|
import annif.util |
|
8
|
|
|
from annif.datadir import DatadirMixin |
|
9
|
|
|
from annif.exception import NotInitializedException |
|
10
|
|
|
from annif.util import parse_args |
|
11
|
|
|
|
|
12
|
|
|
logger = annif.logger |
|
13
|
|
|
|
|
14
|
|
|
|
|
15
|
|
|
def get_vocab(vocab_spec, datadir, default_language): |
|
16
|
|
|
match = re.match(r'(\w+)(\((.*)\))?', vocab_spec) |
|
17
|
|
|
if match is None: |
|
18
|
|
|
raise ValueError(f"Invalid vocabulary specification: {vocab_spec}") |
|
19
|
|
|
vocab_id = match.group(1) |
|
20
|
|
|
posargs, kwargs = parse_args(match.group(3)) |
|
21
|
|
|
language = posargs[0] if posargs else default_language |
|
22
|
|
|
|
|
23
|
|
|
return AnnifVocabulary(vocab_id, datadir, language) |
|
24
|
|
|
|
|
25
|
|
|
|
|
26
|
|
View Code Duplication |
class AnnifVocabulary(DatadirMixin): |
|
|
|
|
|
|
27
|
|
|
"""Class representing a subject vocabulary which can be used by multiple |
|
28
|
|
|
Annif projects.""" |
|
29
|
|
|
|
|
30
|
|
|
# defaults for uninitialized instances |
|
31
|
|
|
_subjects = None |
|
32
|
|
|
|
|
33
|
|
|
# constants |
|
34
|
|
|
INDEX_FILENAME_DUMP = "subjects.dump.gz" |
|
35
|
|
|
INDEX_FILENAME_TTL = "subjects.ttl" |
|
36
|
|
|
INDEX_FILENAME_CSV = "subjects.csv" |
|
37
|
|
|
|
|
38
|
|
|
def __init__(self, vocab_id, datadir, language): |
|
39
|
|
|
DatadirMixin.__init__(self, datadir, 'vocabs', vocab_id) |
|
40
|
|
|
self.vocab_id = vocab_id |
|
41
|
|
|
self.language = language |
|
42
|
|
|
self._skos_vocab = None |
|
43
|
|
|
|
|
44
|
|
|
def _create_subject_index(self, subject_corpus): |
|
45
|
|
|
subjects = annif.corpus.SubjectIndex() |
|
46
|
|
|
subjects.load_subjects(subject_corpus) |
|
47
|
|
|
annif.util.atomic_save(subjects, self.datadir, |
|
48
|
|
|
self.INDEX_FILENAME_CSV) |
|
49
|
|
|
return subjects |
|
50
|
|
|
|
|
51
|
|
|
def _update_subject_index(self, subject_corpus): |
|
52
|
|
|
old_subjects = self.subjects |
|
53
|
|
|
new_subjects = annif.corpus.SubjectIndex() |
|
54
|
|
|
new_subjects.load_subjects(subject_corpus) |
|
55
|
|
|
updated_subjects = annif.corpus.SubjectIndex() |
|
56
|
|
|
|
|
57
|
|
|
for old_subject in old_subjects: |
|
58
|
|
|
if new_subjects.contains_uri(old_subject.uri): |
|
59
|
|
|
new_subject = new_subjects[new_subjects.by_uri( |
|
60
|
|
|
old_subject.uri)] |
|
61
|
|
|
else: # subject removed from new corpus |
|
62
|
|
|
new_subject = annif.corpus.Subject(uri=old_subject.uri, |
|
63
|
|
|
labels=None, |
|
64
|
|
|
notation=None) |
|
65
|
|
|
updated_subjects.append(new_subject) |
|
66
|
|
|
for new_subject in new_subjects: |
|
67
|
|
|
if not old_subjects.contains_uri(new_subject.uri): |
|
68
|
|
|
updated_subjects.append(new_subject) |
|
69
|
|
|
annif.util.atomic_save(updated_subjects, self.datadir, |
|
70
|
|
|
self.INDEX_FILENAME_CSV) |
|
71
|
|
|
return updated_subjects |
|
72
|
|
|
|
|
73
|
|
|
@property |
|
74
|
|
|
def subjects(self): |
|
75
|
|
|
if self._subjects is None: |
|
76
|
|
|
path = os.path.join(self.datadir, self.INDEX_FILENAME_CSV) |
|
77
|
|
|
if os.path.exists(path): |
|
78
|
|
|
logger.debug('loading subjects from %s', path) |
|
79
|
|
|
self._subjects = annif.corpus.SubjectIndex.load(path) |
|
80
|
|
|
else: |
|
81
|
|
|
raise NotInitializedException( |
|
82
|
|
|
"subject file {} not found".format(path)) |
|
83
|
|
|
return self._subjects |
|
84
|
|
|
|
|
85
|
|
|
@property |
|
86
|
|
|
def skos(self): |
|
87
|
|
|
"""return the subject vocabulary from SKOS file""" |
|
88
|
|
|
if self._skos_vocab is not None: |
|
89
|
|
|
return self._skos_vocab |
|
90
|
|
|
|
|
91
|
|
|
# attempt to load graph from dump file |
|
92
|
|
|
dumppath = os.path.join(self.datadir, self.INDEX_FILENAME_DUMP) |
|
93
|
|
|
if os.path.exists(dumppath): |
|
94
|
|
|
logger.debug(f'loading graph dump from {dumppath}') |
|
95
|
|
|
try: |
|
96
|
|
|
self._skos_vocab = annif.corpus.SubjectFileSKOS(dumppath) |
|
97
|
|
|
except ModuleNotFoundError: |
|
98
|
|
|
# Probably dump has been saved using a different rdflib version |
|
99
|
|
|
logger.debug('could not load graph dump, using turtle file') |
|
100
|
|
|
else: |
|
101
|
|
|
return self._skos_vocab |
|
102
|
|
|
|
|
103
|
|
|
# graph dump file not found - parse ttl file instead |
|
104
|
|
|
path = os.path.join(self.datadir, self.INDEX_FILENAME_TTL) |
|
105
|
|
|
if os.path.exists(path): |
|
106
|
|
|
logger.debug(f'loading graph from {path}') |
|
107
|
|
|
self._skos_vocab = annif.corpus.SubjectFileSKOS(path) |
|
108
|
|
|
# store the dump file so we can use it next time |
|
109
|
|
|
self._skos_vocab.save_skos(path) |
|
110
|
|
|
return self._skos_vocab |
|
111
|
|
|
|
|
112
|
|
|
raise NotInitializedException(f'graph file {path} not found') |
|
113
|
|
|
|
|
114
|
|
|
def load_vocabulary(self, subject_corpus, force=False): |
|
115
|
|
|
"""Load subjects from a subject corpus and save them into one |
|
116
|
|
|
or more subject index files as well as a SKOS/Turtle file for later |
|
117
|
|
|
use. If force=True, replace the existing subject index completely.""" |
|
118
|
|
|
|
|
119
|
|
|
if not force and os.path.exists( |
|
120
|
|
|
os.path.join(self.datadir, self.INDEX_FILENAME_CSV)): |
|
121
|
|
|
logger.info('updating existing vocabulary') |
|
122
|
|
|
self._subjects = self._update_subject_index(subject_corpus) |
|
123
|
|
|
else: |
|
124
|
|
|
self._subjects = self._create_subject_index(subject_corpus) |
|
125
|
|
|
|
|
126
|
|
|
subject_corpus.save_skos( |
|
127
|
|
|
os.path.join(self.datadir, self.INDEX_FILENAME_TTL)) |
|
128
|
|
|
|
|
129
|
|
|
def as_graph(self): |
|
130
|
|
|
"""return the vocabulary as an rdflib graph""" |
|
131
|
|
|
return self.skos.graph |
|
132
|
|
|
|
NatLibFi /
Annif
Loading history...