diff options
| author | Bryan Newbold <bnewbold@archive.org> | 2020-04-10 17:20:40 -0700 |
|---|---|---|
| committer | Bryan Newbold <bnewbold@archive.org> | 2020-04-10 17:20:40 -0700 |
| commit | b123a324d9ac5faad3167052710e71edaadf1bfe (patch) | |
| tree | 4c3efea2a5f58d888056cb46831aae8e9e3a3791 | |
| parent | 44f12b817be9d4ccd4b2c0c6ea07ea401ae86ffa (diff) | |
| download | fatcat-covid19-b123a324d9ac5faad3167052710e71edaadf1bfe.tar.gz fatcat-covid19-b123a324d9ac5faad3167052710e71edaadf1bfe.zip | |
pipeline scripts
| -rwxr-xr-x | bin/make_pdf_derivatives.sh | 4 | ||||
| -rwxr-xr-x | bin/run_pipeline.sh | 61 |
2 files changed, 63 insertions, 2 deletions
diff --git a/bin/make_pdf_derivatives.sh b/bin/make_pdf_derivatives.sh index 8727119..76c9ebf 100755 --- a/bin/make_pdf_derivatives.sh +++ b/bin/make_pdf_derivatives.sh @@ -14,9 +14,9 @@ fi echo "processing: $PDFPATH" if [[ ! -f "$FULLTEXTDIR/pdftotext/$BLOBPATH.txt" ]]; then - pdftotext $PDFPATH $FULLTEXTDIR/pdftotext/$BLOBPATH.txt + pdftotext $PDFPATH $FULLTEXTDIR/pdftotext/$BLOBPATH.txt || true fi if [[ ! -f "$FULLTEXTDIR/thumbnail/$BLOBPATH.png" ]]; then - pdftocairo -png -singlefile -scale-to-x 400 -scale-to-y -1 $PDFPATH $FULLTEXTDIR/thumbnail/$BLOBPATH + pdftocairo -png -singlefile -scale-to-x 400 -scale-to-y -1 $PDFPATH $FULLTEXTDIR/thumbnail/$BLOBPATH || true fi diff --git a/bin/run_pipeline.sh b/bin/run_pipeline.sh new file mode 100755 index 0000000..7ace3b3 --- /dev/null +++ b/bin/run_pipeline.sh @@ -0,0 +1,61 @@ +#!/usr/bin/env bash + + +# invoke this script like: +# +# pipenv shell +# CORDDATE=2020-04-03 ./bin/run_pipeline.sh + +set -e -u -o pipefail + +export TODAY="`TZ=UTC date --iso-8601=date`" + +echo "TODAY=$TODAY" +echo "CORDDATE=$CORDDATE" + +# versbose: +set -x + +mkdir -p metadata fulltext_web + +if [ ! -f metadata/cord19.$CORDDATE.csv ]; then + wget https://archive.org/download/s2-cord19-dataset/cord19.$CORDDATE.csv -O metadata/cord19.$CORDDATE.csv --no-clobber || true +fi + +if [ ! -f metadata/cord19.$CORDDATE.json ]; then + ./covid19_tool.py parse-cord19 metadata/cord19.$CORDDATE.csv > metadata/cord19.$CORDDATE.json +fi + +if [ ! -f metadata/cord19.$CORDDATE.enrich.json ]; then + cat metadata/cord19.$CORDDATE.json | parallel -j10 --linebuffer --round-robin --pipe ./covid19_tool.py enrich-fatcat - | pv -l > metadata/cord19.$CORDDATE.enrich.json +fi + +if [ ! -f metadata/cord19.$CORDDATE.missing.json ]; then + cat metadata/cord19.$CORDDATE.enrich.json | jq 'select(.release_id == null) | .cord19_paper' -c > metadata/cord19.$CORDDATE.missing.json +fi + +if [ ! -f metadata/fatcat_hits.$TODAY.enrich.json ]; then + ./covid19_tool.py query-fatcat | pv -l > metadata/fatcat_hits.$TODAY.enrich.json +fi + +if [ ! -f metadata/combined.$TODAY.enrich.json ]; then + cat metadata/fatcat_hits.$TODAY.enrich.json metadata/cord19.$CORDDATE.enrich.json | ./covid19_tool.py dedupe | pv -l > metadata/combined.$TODAY.enrich.json +fi + +if [ ! -f fatcat_web_$TODAY.log ]; then + cat metadata/combined.$TODAY.enrich.json | jq .fatcat_release -c | parallel -j20 --linebuffer --round-robin --pipe ./bin/deliver_file2disk.py --disk-dir fulltext_web - | pv -l > fatcat_web_$TODAY.log +fi + +if [ ! -f metadata/derivatives.$TODAY.stamp ]; then + ./bin/make_dir_derivatives.sh fulltext_web + touch metadata/derivatives.$TODAY.stamp +fi + +if [ ! -f metadata/combined.$TODAY.fulltext.json ]; then + ./covid19_tool.py enrich-derivatives metadata/combined.$TODAY.enrich.json --base-dir fulltext_web/ | pv -l > metadata/combined.$TODAY.fulltext.json +fi + +echo "## Fulltext Inclusion Counts" +cat metadata/combined.$TODAY.fulltext.json | jq .fulltext_status -r | sort | uniq -c | sort -nr + +./covid19_tool.py transform-es metadata/combined.$TODAY.fulltext.json | pv -l | esbulk -verbose -size 1000 -id fatcat_ident -w 8 -index covid19_fatcat_fulltext -type release |