blob: ae9764e6395edef6306d6ec58aa26d5b9ac6d554 (
plain)
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
|
Run in order:
- ISSN
- ORCID
- Crossref
- Manifest
Lots of trouble with encoding; always `export LC_ALL=C.UTF-8`
## Data Sources
Download the following; uncompress the sqlite file, but **do not** uncompress
the others:
https://archive.org/download/crossref_doi_dump_201801/crossref-works.2018-01-21.json.xz
https://archive.org/download/ia_papers_manifest_2018-01-25/index/idents_files_urls.sqlite.gz
https://archive.org/download/ia_journal_metadata_explore_2018-04-05/journal_extra_metadata.csv
https://archive.org/download/issn_issnl_mappings/20180216.ISSN-to-ISSN-L.txt
https://archive.org/download/orcid-dump-2017/public_profiles_API-2.0_2017_10_json.tar.gz
## ISSN
From CSV file:
export LC_ALL=C.UTF-8
time ./client.py import-issn /srv/datasets/journal_extra_metadata.csv
real 2m42.148s
user 0m11.148s
sys 0m0.336s
Pretty quick, a few minutes.
## ORCID
Directly from compressed tarball; takes about 2 hours in production:
tar xf /srv/datasets/public_profiles_API-2.0_2017_10_json.tar.gz -O | jq -c . | grep '"person":' | time parallel -j12 --pipe --round-robin ./client.py import-orcid -
After tuning database, `jq` CPU seems to be bottleneck, so, from pre-extracted
tarball:
tar xf /srv/datasets/public_profiles_API-2.0_2017_10_json.tar.gz -O | jq -c . | rg '"person":' > /srv/datasets/public_profiles_1_2_json.all.json
time parallel --bar --pipepart -j8 -a /srv/datasets/public_profiles_1_2_json.all.json ./client.py import-orcid -
Does not work:
./client.py import-orcid /data/orcid/partial/public_profiles_API-2.0_2017_10_json/3/0000-0001-5115-8623.json
Instead:
cat /data/orcid/partial/public_profiles_API-2.0_2017_10_json/3/0000-0001-5115-8623.json | jq -c . | ./client.py import-orcid -
Or for many files:
find /data/orcid/partial/public_profiles_API-2.0_2017_10_json/3 -iname '*.json' | parallel --bar jq -c . {} | rg '"person":' | ./client.py import-orcid -
### ORCID Performance
for ~9k files:
(python-B2RYrks8) bnewbold@orithena$ time parallel --pipepart -j4 -a /data/orcid/partial/public_profiles_API-2.0_2017_10_json/all.json ./client.py import-orcid -
real 0m15.294s
user 0m28.112s
sys 0m2.408s
=> 636/second
(python-B2RYrks8) bnewbold@orithena$ time ./client.py import-orcid /data/orcid/partial/public_profiles_API-2.0_2017_10_json/all.json
real 0m47.268s
user 0m2.616s
sys 0m0.104s
=> 203/second
For the full batch, on production machine with 12 threads, around 3.8 million records:
3550.76 user
190.16 system
1:40:01 elapsed
=> 644/second
After some simple database tuning:
2177.86 user
145.60 system
56:41.26 elapsed
=> 1117/second
## Crossref
From compressed:
xzcat /srv/datasets/crossref-works.2018-01-21.json.xz | time parallel -j20 --round-robin --pipe ./client.py import-crossref - /srv/datasets/20180216.ISSN-to-ISSN-L.txt
## Manifest
time ./client.py import-manifest /srv/datasets/idents_files_urls.sqlite
[...]
Finished a batch; row 284518671 of 9669646 (2942.39%). Total inserted: 6606900
Finished a batch; row 284518771 of 9669646 (2942.39%). Total inserted: 6606950
Finished a batch; row 284518845 of 9669646 (2942.39%). Total inserted: 6607000
Finished a batch; row 284518923 of 9669646 (2942.39%). Total inserted: 6607050
Done! Inserted 6607075
real 1590m36.626s
user 339m40.928s
sys 19m3.576s
Really sped up once not contending with Crossref import, so don't run these two at the same time.
|
