diff options
Diffstat (limited to 'extra/sitemap/work_urls_query.sh')
| -rwxr-xr-x | extra/sitemap/work_urls_query.sh | 6 |
1 files changed, 4 insertions, 2 deletions
diff --git a/extra/sitemap/work_urls_query.sh b/extra/sitemap/work_urls_query.sh index c369c35..813596b 100755 --- a/extra/sitemap/work_urls_query.sh +++ b/extra/sitemap/work_urls_query.sh @@ -7,10 +7,12 @@ set -o pipefail # fail if part of a '|' command fails # query for specific works; about 8.6 million circa 2021-04-29 # NOTE: filter out 12-digit access URLs (vs. 14-digit) -fatcat-cli search scholar 'doc_type:work (fulltext.access_type:ia_file OR fulltext.access_type:wayback) (NOT biblio.arxiv_id:*) (NOT biblio.pmcid:*) (NOT biblio.publisher_type:big5) (year:<1926 OR tags:*)' --index-json --limit 0 \ - | pv -l \ +fatcat-cli search scholar 'doc_type:work (fulltext.access_type:ia_file OR fulltext.access_type:wayback) (NOT biblio.arxiv_id:*) (NOT biblio.pmcid:*) ((NOT biblio.publisher_type:big5) OR year:<1926 OR tags:oa)' --index-json --limit 0 \ + | jq -c 'select(.biblio.release_ident == .fulltext.release_ident)' \ + | rg -v '^null' \ | rg -v 'web.archive.org/web/\d{12}/' \ | jq .key -r \ | tr '_' '/' \ | awk '{print "https://scholar.archive.org/" $1}' \ + | pv -l \ | split --lines 20000 - sitemap-works- -d -a 5 --additional-suffix .txt |