edge case name and doi handling

Author: mccallc

dvcurator/dataverse.py

@@ -15,13 +15,21 @@ def get_metadata(doi, token=None, host=None):
 	:return: Metadata block from Dataverse API, or None if error
 	:rtype: list[str]
 	"""
-	import requests, dvcurator.hosts
+	import re, requests, dvcurator.hosts
 
 	doi = doi.strip()
 	if not doi.startswith("doi:"):
 		print("Error: DOIs should start with \"doi:\"")
 		return None
 
+	# check for invalid DOI
+	doi_suffix = doi[4:]  # Remove the "doi:" prefix
+	doi_pattern = r"^10\.\d{4,9}/[-._;()/:A-Z0-9]+$"
+	if not re.match(doi_pattern, doi_suffix, re.IGNORECASE):
+		print(f"Error: Invalid DOI format: {doi}")
+		return None
+
+	
 	# Scrape data and metadata from dataverse
 	host = dvcurator.hosts.qdr_dataverse if not host else host 
 	dataset_url = host + '/api/datasets/:persistentId/?persistentId=' + doi

dvcurator/rename.py

@@ -15,7 +15,7 @@ def project_name(citation):
     :rtype: String
 
     """
-    import re
+    import re, unicodedata
 
     author1_last_name = citation['author'][0]['authorName']['value'].split(', ')[0]
     title = citation['title']
@@ -40,6 +40,10 @@ def project_name(citation):
     # replace multiple spaces with one space
     folder_name = re.sub(r'\s+', ' ', folder_name).strip()
     folder_name = folder_name.strip()  # Remove leading/trailing whitespace
+
+    # revert accented characters to their ascii equivalents. 
+    # also totally strips unicode characters
+    folder_name = unicodedata.normalize('NFKD', folder_name).encode('ascii', 'ignore').decode('ascii')
 
     return folder_name
 

test.py

@@ -37,8 +37,12 @@ def test_new_step(self):
 class TestDataverseAPI(unittest.TestCase):
 
 	def test_citation(self):
+		# these should all fail
 		self.assertIsNone(dataverse.get_metadata("foobar", host=harvard_host))
 		self.assertIsNone(dataverse.get_metadata("doi:foobar", host=harvard_host))
+		self.assertIsNone(dataverse.get_metadata("doi:10.5064/F6FHTB9H"))
+		self.assertIsNone(dataverse.get_metadata("doi:doi:doi:10.5064/F6FHTB9E"))
+		# now let's test a success case
 		metadata = dataverse.get_metadata(harvard_doi, host=harvard_host)
 		citation = dataverse.get_citation(metadata)
 		self.assertIsNotNone(citation)
@@ -76,6 +80,7 @@ def test_check(self):
 
 	def test_search(self):
 		self.assertTrue(github.search_existing("Karcher - Anonymous Peer Review", repo="QualitativeDataRepository/testing-demos"))
+		self.assertFalse(github.search_existing("Nobody - Project that doesnt exist", repo="QualitativeDataRepository/testing-demos"))
 
 	def test_version(self):
 		self.assertFalse(github.check_version())
@@ -105,6 +110,11 @@ def test_projectname(self):
 		os.makedirs(new_folder_path)
 		self.assertTrue(os.path.exists(new_folder_path))  # Ensure it was successfully made
 		d.cleanup()
+		# test accented character removal
+		metadata = dataverse.get_metadata("doi:10.5064/F6FHTB9E")
+		citation = dataverse.get_citation(metadata)
+		self.assertIsNotNone(citation)
+		self.assertEqual(rename.project_name(citation), "Rabello Sodre - Memories about Colegio Sao Vicente")
 
 	def test_rename(self):
 		f = tempfile.TemporaryDirectory()