Account for multiple OpenAlex works with the same DOI

The dois_from_orcid() now returns a list of unique DOIs, instead of an
iterator of possiblly unique DOIs. This allows a failing test to pass.

The test for openalex_publications_from_dois() was relaxed a bit to look
for 231 or more publications, since a lookup for an individual DOI can
sometimes pull back multiple works.

The number of columns for OpenAlex is now 53 because we added
`institution_assertions`.

rialto_airflow/harvest/openalex.py

@@ -37,7 +37,7 @@ def doi_orcids_pickle(authors_csv, pickle_file, limit=None):
 
 def dois_from_orcid(orcid: str, limit=None):
     """
-    Pass in the ORCID ID and get back an iterator of DOIs for publications authored by that person.
+    Pass in the ORCID ID and get back a list of DOIs for publications authored by that person.
     """
 
     # TODO: I think we can maybe have this function take a list of orcids and
@@ -57,16 +57,19 @@ def dois_from_orcid(orcid: str, limit=None):
     author_id = authors[0]["id"]
 
     # get all the works for the openalex author id
-    work_count = 0
+    dois = set()
     for page in (
         Works().filter(author={"id": author_id}).select(["doi"]).paginate(per_page=200)
     ):
         for pub in page:
             if pub.get("doi"):
-                work_count += 1
-                if limit is not None and work_count > limit:
-                    return
-                yield pub.get("doi").replace("https://doi.org/", "")
+                doi = pub.get("doi").replace("https://doi.org/", "")
+                dois.add(doi)
+                if limit is not None and len(dois) == limit:
+                    return list(dois)
+
+    return list(dois)
+
 
 
 def publications_csv(dois: list, csv_file: str) -> None:

test/harvest/test_openalex.py

@@ -18,7 +18,7 @@ def test_dois_from_orcid_paging():
     # for Shanhui Fan who has a lot of publications (> 1300)
     dois = list(openalex.dois_from_orcid("0000-0002-0081-9732", limit=300))
     assert len(dois) == 300, "paging is limiting to 200 works"
-    assert len(set(dois)) == 300, "the dois are unique"
+    assert len(set(dois)) == len(dois), "the dois are unique"
 
 
 def test_doi_orcids_pickle(tmp_path):
@@ -48,11 +48,12 @@ def test_publications_from_dois():
 
     # look up the publication metadata for them
     pubs = list(openalex.publications_from_dois(dois))
-    assert len(pubs) == 231, "should paginate (page size=200)"
-    assert len(pubs) == len(set([pub["doi"] for pub in pubs])), "DOIs are unique"
+
+    # >= is used because sometimes there can be multiple works for a DOI!
+    assert len(pubs) >= 231, "should paginate (page size=200)"
     assert set(openalex.FIELDS) == set(pubs[0].keys()), "All fields accounted for."
-    assert len(pubs[0].keys()) == 52, "first publication has 52 columns"
-    assert len(pubs[1].keys()) == 52, "second publication has 52 columns"
+    assert len(pubs[0].keys()) == 53, "first publication has 53 columns"
+    assert len(pubs[1].keys()) == 53, "second publication has 53 columns"
 
 
 def test_publications_from_invalid_dois(caplog):