Merge branch 'bnewbold-shadow-import'

2 files changed, 71 insertions, 0 deletions

diff --git a/python/tests/files/example_shadow.json b/python/tests/files/example_shadow.json
new file mode 100644
index 00000000..3386f481
--- /dev/null
+++ b/python/tests/files/example_shadow.json
@@ -0,0 +1,10 @@
+{"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
new file mode 100644
index 00000000..70a918d2
--- /dev/null
+++ b/python/tests/import_shadow.py
@@ -0,0 +1,61 @@
+
+import json
+import pytest
+from fatcat_tools.importers import ShadowLibraryImporter, JsonLinePusher
+from fixtures import api
+
+
+@pytest.fixture(scope="function")
+def shadow_importer(api):
+    yield ShadowLibraryImporter(api)
+
+# TODO: use API to check that entities actually created...
+def test_shadow_importer_basic(shadow_importer):
+    with open('tests/files/example_shadow.json', 'r') as f:
+        JsonLinePusher(shadow_importer, f).run()
+
+def test_shadow_importer(shadow_importer):
+    last_index = shadow_importer.api.get_changelog(limit=1)[0].index
+    with open('tests/files/example_shadow.json', 'r') as f:
+        shadow_importer.bezerk_mode = True
+        counts = JsonLinePusher(shadow_importer, f).run()
+    assert counts['insert'] == 2
+    assert counts['exists'] == 0
+    assert counts['skip'] == 8
+
+    # fetch most recent editgroup
+    change = shadow_importer.api.get_changelog_entry(index=last_index+1)
+    eg = change.editgroup
+    assert eg.description
+    assert "shadow library" in eg.description.lower()
+    assert eg.extra['git_rev']
+    assert "fatcat_tools.ShadowLibraryImporter" in eg.extra['agent']
+
+    # re-insert; should skip
+    with open('tests/files/example_shadow.json', 'r') as f:
+        shadow_importer.reset()
+        shadow_importer.bezerk_mode = False
+        counts = JsonLinePusher(shadow_importer, f).run()
+    assert counts['insert'] == 0
+    assert counts['exists'] == 2
+    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 == "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":
+                assert u.url.startswith("https://link.springer.com/content/pdf/10.1007%2Fs11626-008-9119-8.pdf")
+            if u.rel == "webarchive":
+                assert u.url.startswith("https://web.archive.org/")
+                assert "20180729135948" in u.url
+        assert len(f.release_ids) == 1
+