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
|
import sys
import json
import base64
import sqlite3
import itertools
import fatcat_openapi_client
from .common import EntityImporter, clean, make_rel_url, SANE_MAX_RELEASES, SANE_MAX_URLS
def b32_hex(s):
s = s.strip().split()[0].lower()
if s.startswith("sha1:"):
s = s[5:]
if len(s) != 32:
return s
return base64.b16encode(base64.b32decode(s.upper())).lower().decode('utf-8')
ARABESQUE_MATCH_WHERE_CLAUSE='WHERE hit = 1 AND identifier IS NOT NULL'
class ArabesqueMatchImporter(EntityImporter):
"""
Importer for arabesque crawl report .sqlite files, which contain
file/identifier matches based on URL/identifier seedlists.
Uses a SQL query to iterate through table looking for rows with:
- GROBID status 200
- known SHA-1 (base32 format)
- known timestamp
Looks up release (by identifier) and file (by SHA-1). If no release exists,
bail.
If no file exists, create one from metadata (using both direct and wayback
URLs), link to release, and insert.
If file exists, optionally update it:
- if no release match, match to the release
- if new URL not included, add it (and wayback)
Config options:
- default URL rel
- crawl id (for editgroup metadata)
- identifier type
TODO:
- a mode to insert bare files even if identifier not known?
"""
def __init__(self, api, extid_type, require_grobid=True, **kwargs):
eg_desc = kwargs.get('editgroup_description',
"Match web crawl files to releases based on identifier/URL seedlist")
eg_extra = kwargs.get('editgroup_extra', dict())
eg_extra['agent'] = eg_extra.get('agent', 'fatcat_tools.ArabesqueMatchImporter')
if kwargs.get('crawl_id'):
eg_extra['crawl_id'] = kwargs.get('crawl_id')
super().__init__(api,
editgroup_description=eg_desc,
editgroup_extra=eg_extra,
**kwargs)
assert extid_type in ('doi', 'pmcid', 'pmid')
self.extid_type = extid_type
self.default_link_rel = kwargs.get("default_link_rel", "web")
assert self.default_link_rel
self.default_mimetype = kwargs.get("default_mimetype", None)
self.do_updates = kwargs.get("do_updates", False)
self.require_grobid = require_grobid
if self.require_grobid:
print("Requiring GROBID status == 200")
else:
print("NOT checking GROBID status column")
def want(self, row):
if self.require_grobid and not row['postproc_status'] == "200":
return False
if (row['hit'] == True
and row['final_sha1']
and row['final_timestamp'] and row['final_timestamp'] != "-"
and row['final_mimetype']
and row['hit'] == True
and row['identifier']):
return True
else:
return False
def parse_record(self, row):
extid = row['identifier'].strip()
# check/cleanup DOI
if self.extid_type == 'doi':
extid = extid.lower()
extid.replace('http://doi.org/', '')
extid.replace('https://doi.org/', '')
if extid.startswith('doi:'):
extid = extid[4:]
if not extid.startswith('10.'):
self.counts['skip-extid-invalid']
return None
# lookup extid
try:
re = self.api.lookup_release(**{self.extid_type: extid})
except fatcat_openapi_client.rest.ApiException as err:
if err.status == 404:
# bail on 404 (release not in DB)
self.counts['skip-extid-not-found'] += 1
return None
elif err.status == 400:
self.counts['skip-extid-invalid'] += 1
return None
else:
raise err
url = make_rel_url(row['final_url'], self.default_link_rel)
if not url:
self.counts['skip-url'] += 1
return None
if not row['final_timestamp']:
self.counts['skip-missing-timestamp'] += 1
return None
wayback = "https://web.archive.org/web/{}/{}".format(
row['final_timestamp'],
row['final_url'])
urls = [url, ("webarchive", wayback)]
urls = [fatcat_openapi_client.FileUrl(rel=rel, url=url) for (rel, url) in urls]
if len(urls) > SANE_MAX_URLS:
self.counts['skip-too-many-url'] += 1
return None
fe = fatcat_openapi_client.FileEntity(
sha1=b32_hex(row['final_sha1']),
mimetype=row['final_mimetype'] or self.default_mimetype,
release_ids=[re.ident],
urls=urls,
)
return fe
def try_update(self, fe):
# lookup sha1, or create new entity
existing = None
try:
existing = self.api.lookup_file(sha1=fe.sha1)
except fatcat_openapi_client.rest.ApiException as err:
if err.status != 404:
raise err
if not existing:
return True
if (fe.release_ids[0] in existing.release_ids) and existing.urls:
# TODO: could still, in theory update with the new URL?
self.counts['exists'] += 1
return False
if not self.do_updates:
self.counts['skip-update-disabled'] += 1
return False
if existing.ident in [e.ident for e in self._edits_inflight]:
self.counts['skip-update-inflight'] += 1
return False
# TODO: this code path never gets hit because of the check above
if set(fe.release_ids) == set(existing.release_ids):
existing_urls = set([u.url for u in existing.urls])
new_urls = set([u.url for u in fe.urls])
if existing_urls.issuperset(new_urls):
self.counts['skip-update-nothing-new'] += 1
return False
# merge the existing into this one and update
existing.urls = list(set([(u.rel, u.url) for u in fe.urls + existing.urls]))
existing.urls = [fatcat_openapi_client.FileUrl(rel=rel, url=url) for (rel, url) in existing.urls]
if len(existing.urls) > SANE_MAX_URLS:
self.counts['skip-update-too-many-url'] += 1
return None
existing.release_ids = list(set(fe.release_ids + existing.release_ids))
if len(existing.release_ids) > SANE_MAX_RELEASES:
self.counts['skip-update-too-many-url'] += 1
return None
existing.mimetype = existing.mimetype or fe.mimetype
edit = self.api.update_file(self.get_editgroup_id(), existing.ident, existing)
self._edits_inflight.append(edit)
self.counts['update'] += 1
return False
def insert_batch(self, batch):
self.api.create_file_auto_batch(fatcat_openapi_client.FileAutoBatch(
editgroup=fatcat_openapi_client.Editgroup(
description=self.editgroup_description,
extra=self.editgroup_extra),
entity_list=batch))
|