From 42da448ea610afa980b5d893ce5982b98507fe3c Mon Sep 17 00:00:00 2001 From: Bryan Newbold Date: Wed, 19 May 2021 16:57:19 -0700 Subject: sitemaps: filter out 12-digit wayback timestamp access URLs, for now --- extra/sitemap/access_urls_query.sh | 2 ++ 1 file changed, 2 insertions(+) (limited to 'extra/sitemap/access_urls_query.sh') diff --git a/extra/sitemap/access_urls_query.sh b/extra/sitemap/access_urls_query.sh index 284b52b..eb34564 100755 --- a/extra/sitemap/access_urls_query.sh +++ b/extra/sitemap/access_urls_query.sh @@ -7,9 +7,11 @@ set -o pipefail # fail if part of a '|' command fails SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )" # 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 \ | jq '[.fulltext.access_type, .fulltext.access_url] | @tsv' -r \ | rg -v '^null' \ + | rg -v 'web.archive.org/web/\d{12}/' \ | $SCRIPT_DIR/transform_access_url.py \ | split --lines 20000 - sitemap-access- -d -a 5 --additional-suffix .txt -- cgit v1.2.3