diff options
| -rw-r--r-- | python/fatcat_tools/importers/shadow.py | 10 | ||||
| -rw-r--r-- | python/tests/files/example_shadow.json | 22 | ||||
| -rw-r--r-- | python/tests/import_shadow.py | 14 |
3 files changed, 28 insertions, 18 deletions
diff --git a/python/fatcat_tools/importers/shadow.py b/python/fatcat_tools/importers/shadow.py index 21a18837..cfe1b1cf 100644 --- a/python/fatcat_tools/importers/shadow.py +++ b/python/fatcat_tools/importers/shadow.py @@ -43,6 +43,16 @@ class ShadowLibraryImporter(EntityImporter): self.default_link_rel = kwargs.get("default_link_rel", "web") def want(self, raw_record): + """ + Only want to import records with complete file-level metadata + """ + fm = raw_record['file_meta'] + if not (fm['mimetype'] and fm['md5hex'] and fm['sha256hex'] and fm['size_bytes']): + self.counts['skip-file-meta-incomplete'] += 1 + return False + if fm['mimetype'] != 'application/pdf': + self.counts['skip-not-pdf'] += 1 + return False return True def parse_record(self, obj): diff --git a/python/tests/files/example_shadow.json b/python/tests/files/example_shadow.json index f84a61a5..3386f481 100644 --- a/python/tests/files/example_shadow.json +++ b/python/tests/files/example_shadow.json @@ -1,12 +1,10 @@ -{"shadow":{"shadow_corpus":"scimag","shadow_id":"8149931","sha1hex":"000008bc38cb80636b647b38653fc1574936c03e","doi":"10.1371/journal.pmed.0020124","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"000008bc38cb80636b647b38653fc1574936c03e","sha256hex":"18b341119bbbf297a7dfa21aca86211da446617600baa153df70b4209c2c6e84","md5hex":"629e84885be85bc8d88345b98cffa0b0","size_bytes":39955,"mimetype":null},"cdx":{"url":"https://link.springer.com/content/pdf/10.1007%2Fs11626-008-9119-8.pdf","datetime":"20180729135948","sha1hex":"000008bc38cb80636b647b38653fc1574936c03e","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"UNPAYWALL-PDF-CRAWL-2018-07-20180729132538992-15980-16048-wbgrp-svc281/UNPAYWALL-PDF-CRAWL-2018-07-20180729135708800-16009-11693~wbgrp-svc281.us.archive.org~8443.warc.gz","warc_csize":32497,"warc_offset":105265425,"row_created":"2019-08-09T23:25:44.571943+00:00"}} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"33139096","sha1hex":"00000c4296e2c5f8f70ab265c683235fbf5e354b","doi":"10.0000/cyberleninka.ru/article/n/analiz-primeneniya-fazochastotnyh-algoritmov-proslezhivaniya-signalov-dlya-izmereniya-urovnya-zhidkosti-v-neftedobyvayuschih","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"00000c4296e2c5f8f70ab265c683235fbf5e354b","sha256hex":"99f15c58c2343f46c8cae75ff01c11b1b9e3c6d2f57189ec78df94e234b2c633","md5hex":"488681b249f6e9292bcde1fab1422550","size_bytes":182449,"mimetype":null},"cdx":{"url":"http://www.lib.tpu.ru/fulltext/v/Bulletin_TPU/2011/v319/i5/12.pdf","datetime":"20180412144307","sha1hex":"00000c4296e2c5f8f70ab265c683235fbf5e354b","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"OA-JOURNAL-TESTCRAWL-TWO-2018-20180412133030095-00799-00808-wbgrp-svc284/OA-JOURNAL-TESTCRAWL-TWO-2018-20180412142334247-00807-23249~wbgrp-svc284.us.archive.org~8443.warc.gz","warc_csize":126165,"warc_offset":924893749,"row_created":"2019-08-09T05:16:39.785581+00:00"}} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"33139096","sha1hex":"00000c4296e2c5f8f70ab265c683235fbf5e354b","doi":null,"pmid":"54321","isbn13":null},"file_meta":{"sha1hex":"00000c4296e2c5f8f70ab265c683235fbf5e354b","sha256hex":"99f15c58c2343f46c8cae75ff01c11b1b9e3c6d2f57189ec78df94e234b2c633","md5hex":"488681b249f6e9292bcde1fab1422550","size_bytes":182449,"mimetype":null},"cdx":{"url":"https://cyberleninka.ru/article/n/analiz-primeneniya-fazochastotnyh-algoritmov-proslezhivaniya-signalov-dlya-izmereniya-urovnya-zhidkosti-v-neftedobyvayuschih.pdf","datetime":"20180506175847","sha1hex":"00000c4296e2c5f8f70ab265c683235fbf5e354b","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"OA-JOURNAL-TESTCRAWL-TWO-2018-20180506171133875-05766-05775-wbgrp-svc284/OA-JOURNAL-TESTCRAWL-TWO-2018-20180506174415763-05771-23249~wbgrp-svc284.us.archive.org~8443.warc.gz","warc_csize":126144,"warc_offset":532659301,"row_created":"2019-08-09T05:16:39.785581+00:00"}} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"21389391","sha1hex":"00000d5508d7d7106560ade65c33c628c54d7c75","doi":"10.1038/nn.3419","pmid":"23727820","isbn13":null},"file_meta":{"sha1hex":"00000d5508d7d7106560ade65c33c628c54d7c75","sha256hex":"8c48dd68b974ed117f839dc88db44884e7e1df9ddef30f26c541437d7f390d96","md5hex":"c2a43160b62ef0f13256c789270ec2a9","size_bytes":1375452,"mimetype":null},"cdx":{"url":"https://www.janelia.org/sites/default/files/Library/nn.3419.pdf","datetime":"20170829032635","sha1hex":"00000d5508d7d7106560ade65c33c628c54d7c75","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"SEMSCHOLAR-PDF-CRAWL-2017-08-04-20170829031124939-00100-00109-wbgrp-svc284/SEMSCHOLAR-PDF-CRAWL-2017-08-04-20170829032404137-00107-3480~wbgrp-svc284.us.archive.org~8443.warc.gz","warc_csize":973733,"warc_offset":262621802,"row_created":"2019-08-09T05:21:39.486744+00:00"}} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"16462885","sha1hex":"0000102db78329a149d3b6319f6ccf8cc90483e2","doi":"10.1016/j.cell.2007.04.022","pmid":"17482536","isbn13":null},"file_meta":{"sha1hex":"0000102db78329a149d3b6319f6ccf8cc90483e2","sha256hex":null,"md5hex":"995e7145d09d50eadccf322780e474d3","size_bytes":206812,"mimetype":"application/pdf"},"cdx":{"url":"http://publisher-connector.core.ac.uk/resourcesync/data/elsevier/pdf/212/aHR0cDovL2FwaS5lbHNldmllci5jb20vY29udGVudC9hcnRpY2xlL3BpaS9zMDA5Mjg2NzQwNzAwNTI4NA%3D%3D.pdf","datetime":"20170925031039","sha1hex":"0000102db78329a149d3b6319f6ccf8cc90483e2","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"TARGETED-PDF-CRAWL-2017-08-04-20170925022437721-01046-01055-wbgrp-svc284/TARGETED-PDF-CRAWL-2017-08-04-20170925024811552-01048-15075~wbgrp-svc284.us.archive.org~8443.warc.gz","warc_csize":188232,"warc_offset":677858259,"row_created":"2019-08-10T03:02:21.656961+00:00"}} -{"shadow":{"shadow_corpus":"scimag","shadow_id":"12703034","sha1hex":"0000002922264275f11cca7b1c3fb662070d0dd7","doi":"10.1007/s11061-011-9281-1","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000002922264275f11cca7b1c3fb662070d0dd7","sha256hex":"b4728210cc0f70d8a8f8c39bd97fcbbab3eaca4309ac4bdfbce5df3b66c82f79","md5hex":"debd8db178fa08a7a0aaec6e42832a8e","size_bytes":206121,"mimetype":null},"cdx":null} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"51052483","sha1hex":"00000119fa780ce368ebd96563afdb3eebb90ad3","doi":"10.1191/0266355403gh289oa","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"00000119fa780ce368ebd96563afdb3eebb90ad3","sha256hex":"57ce460db4410b9bfaf500ed652fd29e64d46b40c17e28f1156ba03736edf91b","md5hex":"96133eec3a6c533993213e7bdf446251","size_bytes":164344,"mimetype":null},"cdx":null} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"2476283","sha1hex":"0000017a31547caf347fab66282a40831b9ceb08","doi":"10.1016/0042-207x(62)90512-2","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000017a31547caf347fab66282a40831b9ceb08","sha256hex":"e8d0c607b024ff6ffd58a35f76c454844b70ad19fe3f78a573af1ae53f53ad9d","md5hex":"b53318522b9f35a42b7e53f150fe70b2","size_bytes":116735,"mimetype":null},"cdx":null} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"8760871","sha1hex":"000001abf3dbf936d5053d14f41699722531b8c6","doi":"10.1016/s0042-207x(79)80945-8","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"000001abf3dbf936d5053d14f41699722531b8c6","sha256hex":"8a69b4a6dff98682ad43e7d4139221c1557c1bd202b615490af8a2c7dcbb71d2","md5hex":"29e1cfac8ecfbc8be57a1ec8b465c4be","size_bytes":138218,"mimetype":null},"cdx":null} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"11473618","sha1hex":"0000022e387be46ef797f6686d36c9899cbd6856","doi":"10.1038/ng.2339","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000022e387be46ef797f6686d36c9899cbd6856","sha256hex":"a72517e8e72d78bc07a6ef7ff3a6d1d3e04325df986cb8f1bbb4e809f7a9dbdd","md5hex":"9cb8a6e056c9cc740d3bed0c50cd53dc","size_bytes":80992,"mimetype":null},"cdx":null} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"47301218","sha1hex":"0000029209536bda5f22e5110e573c5bd8ceb43a","doi":"10.2307/23406551","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000029209536bda5f22e5110e573c5bd8ceb43a","sha256hex":"315f1d39a00ccf256fa15d92a14869dbda48d31500989aaacb11368f906a5827","md5hex":"8141b42ec3bb41fa87099633a1b61d93","size_bytes":305236,"mimetype":null},"cdx":null} - {"shadow":{"shadow_corpus":"scimag","shadow_id":"30603850","sha1hex":"000002c1abd521f18aa23d9e8f464e697e218ab1","doi":"10.1109/spire.1998.712983","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"000002c1abd521f18aa23d9e8f464e697e218ab1","sha256hex":"777e2c472e9d2fec3bbd26bad788562cf1e08e5850315c25cfb6e46d38e7e4af","md5hex":"3a3c92fabaf6cf437bb596d9e9255ff6","size_bytes":113768,"mimetype":null},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"12703034","sha1hex":"0000002922264275f11cca7b1c3fb662070d0dd7","doi":"10.1371/journal.pmed.0020124","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000002922264275f11cca7b1c3fb662070d0dd7","sha256hex":"b4728210cc0f70d8a8f8c39bd97fcbbab3eaca4309ac4bdfbce5df3b66c82f79","md5hex":"debd8db178fa08a7a0aaec6e42832a8e","size_bytes":206121,"mimetype":"application/pdf"},"cdx":{"url":"https://link.springer.com/content/pdf/10.1007%2Fs11626-008-9119-8.pdf","datetime":"20180729135948","sha1hex":"0000002922264275f11cca7b1c3fb662070d0dd7","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"UNPAYWALL-PDF-CRAWL-2018-07-20180729132538992-15980-16048-wbgrp-svc281/UNPAYWALL-PDF-CRAWL-2018-07-20180729135708800-16009-11693~wbgrp-svc281.us.archive.org~8443.warc.gz","warc_csize":32497,"warc_offset":105265425,"row_created":"2019-08-09T23:25:44.571943+00:00"}} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"51052483","sha1hex":"00000119fa780ce368ebd96563afdb3eebb90ad3","doi":"10.1191/0266355403gh289oa","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"00000119fa780ce368ebd96563afdb3eebb90ad3","sha256hex":"57ce460db4410b9bfaf500ed652fd29e64d46b40c17e28f1156ba03736edf91b","md5hex":"96133eec3a6c533993213e7bdf446251","size_bytes":164344,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"2476283","sha1hex":"0000017a31547caf347fab66282a40831b9ceb08","doi":"10.1016/0042-207x(62)90512-2","pmid":"54321","isbn13":null},"file_meta":{"sha1hex":"0000017a31547caf347fab66282a40831b9ceb08","sha256hex":"e8d0c607b024ff6ffd58a35f76c454844b70ad19fe3f78a573af1ae53f53ad9d","md5hex":"b53318522b9f35a42b7e53f150fe70b2","size_bytes":116735,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"8760871","sha1hex":"000001abf3dbf936d5053d14f41699722531b8c6","doi":"10.1016/s0042-207x(79)80945-8","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"000001abf3dbf936d5053d14f41699722531b8c6","sha256hex":"8a69b4a6dff98682ad43e7d4139221c1557c1bd202b615490af8a2c7dcbb71d2","md5hex":"29e1cfac8ecfbc8be57a1ec8b465c4be","size_bytes":138218,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"11473618","sha1hex":"0000022e387be46ef797f6686d36c9899cbd6856","doi":"10.1038/ng.2339","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000022e387be46ef797f6686d36c9899cbd6856","sha256hex":"a72517e8e72d78bc07a6ef7ff3a6d1d3e04325df986cb8f1bbb4e809f7a9dbdd","md5hex":"9cb8a6e056c9cc740d3bed0c50cd53dc","size_bytes":80992,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"47301218","sha1hex":"0000029209536bda5f22e5110e573c5bd8ceb43a","doi":"10.2307/23406551","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"0000029209536bda5f22e5110e573c5bd8ceb43a","sha256hex":"315f1d39a00ccf256fa15d92a14869dbda48d31500989aaacb11368f906a5827","md5hex":"8141b42ec3bb41fa87099633a1b61d93","size_bytes":305236,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"30603850","sha1hex":"000002c1abd521f18aa23d9e8f464e697e218ab1","doi":"10.1109/spire.1998.712983","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"000002c1abd521f18aa23d9e8f464e697e218ab1","sha256hex":"777e2c472e9d2fec3bbd26bad788562cf1e08e5850315c25cfb6e46d38e7e4af","md5hex":"3a3c92fabaf6cf437bb596d9e9255ff6","size_bytes":113768,"mimetype":"application/pdf"},"cdx":{"url":"http://proteomics.bioprojects.org/pavel/papers/SST_versus_EST_in_gene_recognition..pdf","datetime":"20081121222143","sha1hex":"000002c1abd521f18aa23d9e8f464e697e218ab1","cdx_sha1hex":null,"mimetype":"application/pdf","warc_path":"1227992340180_31-c/1227992509265_9.arc.gz","warc_csize":61212,"warc_offset":62956683,"row_created":"2020-01-07T02:06:33.965383+00:00"}} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"9311918","sha1hex":"000002d4f7d4174451e4214475d5ba59f1f6a593","doi":"10.1111/j.1439-0507.2008.01572.x","pmid":"18721331","isbn13":null},"file_meta":{"sha1hex":"000002d4f7d4174451e4214475d5ba59f1f6a593","sha256hex":"713758ce0417f604c0a4b0bf5b5eea571a9b08ca4cc81a98d602c43f42abfe37","md5hex":"0df123e6305c617ffd38ebef90b1e318","size_bytes":178664,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"7757772","sha1hex":"000002f8966a4c5547f8a47f43661fcc3edc34ea","doi":"10.1007/s10464-011-9424-3","pmid":"21287262","isbn13":null},"file_meta":{"sha1hex":"000002f8966a4c5547f8a47f43661fcc3edc34ea","sha256hex":"ee1bce27134ae55b3d67f9b31f66571e41ac496fc3fb526dec2d53513b8f6deb","md5hex":"e72c5cf3d61635821e78ca0306c98887","size_bytes":337857,"mimetype":"application/pdf"},"cdx":null} +{"shadow":{"shadow_corpus":"scimag","shadow_id":"74272862","sha1hex":"000003a94022be58305ccc2a018a6359eeb226db","doi":"10.1002/slct.201802783","pmid":null,"isbn13":null},"file_meta":{"sha1hex":"000003a94022be58305ccc2a018a6359eeb226db","sha256hex":"f277eefc7b1466df814a7a892ab8e2e7f08db1faae0bf73b893211e5f5b37193","md5hex":"27534b8494f54ba5de47c16fb2590b04","size_bytes":1372272,"mimetype":"application/pdf"},"cdx":null} diff --git a/python/tests/import_shadow.py b/python/tests/import_shadow.py index 30e1724f..70a918d2 100644 --- a/python/tests/import_shadow.py +++ b/python/tests/import_shadow.py @@ -21,7 +21,7 @@ def test_shadow_importer(shadow_importer): counts = JsonLinePusher(shadow_importer, f).run() assert counts['insert'] == 2 assert counts['exists'] == 0 - assert counts['skip'] == 10 + assert counts['skip'] == 8 # fetch most recent editgroup change = shadow_importer.api.get_changelog_entry(index=last_index+1) @@ -38,16 +38,18 @@ def test_shadow_importer(shadow_importer): counts = JsonLinePusher(shadow_importer, f).run() assert counts['insert'] == 0 assert counts['exists'] == 2 - assert counts['skip'] == 10 + assert counts['skip'] == 8 def test_shadow_dict_parse(shadow_importer): with open('tests/files/example_shadow.json', 'r') as f: raw = json.loads(f.readline()) f = shadow_importer.parse_record(raw) - assert f.sha1 == "000008bc38cb80636b647b38653fc1574936c03e" - assert f.md5 == "629e84885be85bc8d88345b98cffa0b0" - assert f.mimetype == None # "application/pdf" - assert f.size == 39955 + + assert f.sha1 == "0000002922264275f11cca7b1c3fb662070d0dd7" + assert f.md5 == "debd8db178fa08a7a0aaec6e42832a8e" + assert f.sha256 == "b4728210cc0f70d8a8f8c39bd97fcbbab3eaca4309ac4bdfbce5df3b66c82f79" + assert f.mimetype == "application/pdf" + assert f.size == 206121 assert len(f.urls) == 2 for u in f.urls: if u.rel == "publisher": |