aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorBryan Newbold <bnewbold@robocracy.org>2021-11-30 18:00:27 -0800
committerBryan Newbold <bnewbold@robocracy.org>2021-11-30 18:00:27 -0800
commit979c9b392ea24c09234d0f26e06150c5fb861411 (patch)
tree8912620c8089076bcba00c4f76a896a168ec76d6
parent3cd594f951728165556849bdff1c35ce1d7daab1 (diff)
downloadfatcat-979c9b392ea24c09234d0f26e06150c5fb861411.tar.gz
fatcat-979c9b392ea24c09234d0f26e06150c5fb861411.zip
initial 'far-future' release date updates
-rw-r--r--extra/bulk_edits/2021-11-30_far_future_dates.md212
1 files changed, 212 insertions, 0 deletions
diff --git a/extra/bulk_edits/2021-11-30_far_future_dates.md b/extra/bulk_edits/2021-11-30_far_future_dates.md
new file mode 100644
index 00000000..15485057
--- /dev/null
+++ b/extra/bulk_edits/2021-11-30_far_future_dates.md
@@ -0,0 +1,212 @@
+
+This is an experiment in using `fatcat-cli` for bulk updates of metadata.
+
+As a simple initial cleanup, looking at far-future release entities, which have
+dates beyond year 2100 (or even beyond year 5000). These are the result of not
+checking for reasonable date values during initial catalog bootstrapping,
+combined with poor upstream metadata.
+
+ fatcat-cli search release --count 'year:>2100'
+ # 279622
+
+ fatcat-cli search release --count 'year:>5000'
+ # 123035
+
+That is a lot of bad releases!
+
+## Source of Bad Metadata
+
+Where were these imported from?
+
+9999 seems to be pretty much just JALC:
+
+ fatcat-cli search release --count 'year:>5000' doi_registrar:jalc
+ # 122279
+
+ fatcat-cli search release --count 'year:9999' doi_registrar:jalc
+ # 122279
+
+ fatcat-cli search release --count 'year:9999' '!doi_registrar:jalc'
+ # 380
+
+A handful from longtail OA import:
+
+ fatcat-cli search release --count 'year:>5000' is_longtail_oa:true
+ 84
+
+ fatcat-cli search release --count 'year:>2100' is_longtail_oa:true
+ 669
+
+A significant number from datacite:
+
+ fatcat-cli search release --count 'year:>5000' doi_registrar:datacite
+ 672
+
+ fatcat-cli search release --count 'year:>2100' doi_registrar:datacite
+ 156359
+
+And a handful from crossref:
+
+ fatcat-cli search release --count 'year:>2100' '!doi_registrar:datacite' '!doi_registrar:jalc' '!is_longtail_oa:true'
+ 314
+
+ fatcat-cli search release 'year:>2100' '!doi_registrar:datacite' '!doi_registrar:jalc' '!is_longtail_oa:true' --index-json --limit 0 | jq .doi_registrar | sort | uniq -c
+ Got 314 hits in 81ms
+ 314 "crossref"
+
+How about no DOI and not longtail OA?
+
+ fatcat-cli search release --count 'year:>2100' is_longtail_oa:false '!doi:*'
+
+## QA (2021-11-17)
+
+ export FATCAT_API_HOST=https://api.qa.fatcat.wiki
+ export FATCAT_AUTH_WORKER_CLEANUP=[...]
+ export FATCAT_API_AUTH_TOKEN=$FATCAT_AUTH_WORKER_CLEANUP
+ # using prod search index
+
+ fatcat-cli --version
+ fatcat-cli 0.1.6
+
+
+ bnewbold@orithena$ fatcat-cli status
+ API Version: 0.5.0 (local)
+ API host: https://api.qa.fatcat.wiki [successfully connected]
+ Last changelog: 5326062
+ API auth token: [configured]
+ Account: cleanup-bot [bot] [admin] [active]
+ editor_vvnmtzskhngxnicockn4iavyxq
+
+Year `9999` seems popular and is probably a place-holder; skip those for later re-import/update.
+
+The QA catalog is out of date from prod, so need to filter to only older
+releases using `doc_index_ts` (as a hack).
+
+We want to re-filter locally to ensure that the fields are actually bogus. `jq` works for this.
+
+ # NOTE: in update chain, important to *NOT* use `--hide` here
+ fatcat-cli search releases 'year:>5000' '!year:9999' 'doc_index_ts:<2021-05-01' --entity-json --limit 20 \
+ | jq 'select(.release_year > 5000)' -c \
+ | fatcat-cli batch update release release_year= release_date=
+ # Got 376 hits in 28ms
+ # editgroup_q2rqwoaa4zdmhmlmyrtcc5tnia
+
+Editgroup came through with agent, `agent_version` correct. Only 5 edits. Newer
+presumably because entities have been updated in prod. Description was "part of
+a fatcat-cli batch operation", should be able to modify this (CLI arg? env
+variable?).
+
+Trying again with `--limit 200`. Got 2x editgroups.
+
+Trying different search parameters:
+
+ fatcat-cli search releases 'year:>2100' is_longtail_oa:true 'doc_index_ts:<2021-05-01' --entity-json --limit 100 \
+ | jq 'select(.release_year > 2100)' -c \
+ | wc -l
+ # Got 669 hits in 312ms
+ # 100
+
+Try doing 300 updates, with auto-accept:
+
+ fatcat-cli search releases 'year:>2100' is_longtail_oa:true 'doc_index_ts:<2021-05-01' --entity-json --limit 300 \
+ | jq 'select(.release_year > 2100)' -c \
+ | pv -l \
+ | fatcat-cli batch update release release_year= release_date= --auto-accept
+
+## Production (2021-11-30)
+
+Scope for this initial update:
+
+- longtail releases
+- Crossref
+- other non-DOI releases (?)
+- possibly JALC releases (more than 100k)
+
+
+ export FATCAT_API_HOST=https://api.fatcat.wiki
+ export FATCAT_AUTH_WORKER_CLEANUP=[...]
+ export FATCAT_API_AUTH_TOKEN=$FATCAT_AUTH_WORKER_CLEANUP
+
+ fatcat-cli --version
+ # fatcat-cli 0.1.6
+
+ fatcat-cli status
+ API Version: 0.5.0 (local)
+ API host: https://api.fatcat.wiki [successfully connected]
+ Last changelog: 5639457
+ API auth token: [configured]
+ Account: cleanup-bot [bot] [admin] [active]
+ editor_vvnmtzskhngxnicockn4iavyxq
+
+Start with longtail releases:
+
+ fatcat-cli search releases 'year:>2100' is_longtail_oa:true 'doc_index_ts:<2021-05-01' --entity-json --limit 100 \
+ | jq 'select(.release_year > 2100)' -c \
+ | wc -l
+ # Got 669 hits in 375ms
+ # 100
+
+ fatcat-cli search releases 'year:>2100' is_longtail_oa:true 'doc_index_ts:<2021-05-01' '!doi:*' --entity-json --limit 100 \
+ | jq 'select(.release_year > 2100)' -c \
+ | head -n50 \
+ | pv -l \
+ | fatcat-cli batch update release release_year= release_date= --description "Cleanup of 'far-future' release dates"
+ # editgroup_6sjgeun22bexlc5ujxhy6ui3sm
+
+No dates in this batch, only years. Diff looks good. Accepted.
+
+Run all these (few hundred, auto-accept):
+
+ fatcat-cli search releases 'year:>2100' is_longtail_oa:true 'doc_index_ts:<2021-11-20' '!doi:*' --entity-json --limit 0 \
+ | jq 'select(.release_year > 2100)' -c \
+ | pv -l \
+ | fatcat-cli batch update release release_year= release_date= --description "Cleanup of 'far-future' release dates" --auto-accept
+ # Got 579 hits in 101ms
+ # 579 0:00:15 [38.0 /s]
+
+Next, Crossref:
+
+ fatcat-cli search releases 'year:>2100' doi_registrar:crossref 'doc_index_ts:<2021-11-20' --count
+ # 314
+
+Small batch, no auto-accept:
+
+ fatcat-cli search releases 'year:>2100' doi_registrar:crossref 'doc_index_ts:<2021-11-20' --entity-json --limit 20 \
+ | jq 'select(.release_year > 2100)' -c \
+ | pv -l \
+ | fatcat-cli batch update release release_year= release_date= --description "Cleanup of 'far-future' release dates"
+ # editgroup_u75a4jdgyra4pbqixt344uus4e
+
+Looks good. Accept, then run full batch:
+
+ fatcat-cli search releases 'year:>2100' doi_registrar:crossref 'doc_index_ts:<2021-11-20' --entity-json --limit 0 \
+ | jq 'select(.release_year > 2100)' -c \
+ | pv -l \
+ | fatcat-cli batch update release release_year= release_date= --description "Cleanup of 'far-future' release dates" --auto-accept
+ # Got 294 hits in 133ms
+ # 294 0:00:05 [49.5 /s]
+
+Any other non-DOI releases?
+
+ fatcat-cli search releases 'year:>2100' '!doi:*' --count
+ # 0
+
+
+ fatcat-cli search releases 'year:>2100' 'doi_registrar:jalc' --count
+ # 122280
+
+Most of the remaining are JALC or datacite, and are large updates. Let's do
+just the subset which are `in_ia`:
+
+ fatcat-cli search releases 'year:>2100' in_ia:true --count
+ # 23
+
+ fatcat-cli search releases 'year:>2100' in_ia:true 'doc_index_ts:<2021-11-20' --entity-json --limit 50 \
+ | jq 'select(.release_year > 2100)' -c \
+ | pv -l \
+ | fatcat-cli batch update release release_year= release_date= --description "Cleanup of 'far-future' release dates"
+ # Got 23 hits in 54ms
+ # 13
+ # editgroup_jg35r4tr7bfavcvksh4k7k3siq
+
+Accepted.