improved project name handling

Author: mccallc

dvcurator/rename.py

@@ -24,13 +24,22 @@ def project_name(citation):
     # What did this line do?
     # title = re.match("^(.+?\\s){1,5}", title).group(0).rstrip()
     title = re.sub("^[^a-zA-Z]?", "", title) # get rid of any beginning non-letter chars
+    # limit to only the content before the first colon
     title = re.sub(":.+", '', title)
-    
+    # if there's more than 5 words, limit it to the first 5
+    words = title.split()
+    if len(words) > 5:
+        title = ' '.join(words[:5])
+
     folder_name = author1_last_name + " - " + title
 
-    special_characters = ['!','#','$','%', '&','@','[',']',']','_',':',';',"'"]
-    for i in special_characters:
-        folder_name = folder_name.replace(i,'')
+    # remove special characters that are not allowed in folder names
+    invalid_chars = r'[<>:"/\\|?*\'\[\]\{\}!#$%&@;=+,]'
+    folder_name = re.sub(invalid_chars, '', folder_name)
+
+    # replace multiple spaces with one space
+    folder_name = re.sub(r'\s+', ' ', folder_name).strip()
+    folder_name = folder_name.strip()  # Remove leading/trailing whitespace
 
     return folder_name
 

test.py

@@ -54,12 +54,18 @@ def test_download(self):
 		self.assertTrue(os.path.isdir(path))
 		self.assertTrue(os.path.exists(os.path.join(path, os.pardir, os.pardir, "Original Deposit.zip")))
 		self.assertTrue(os.path.exists(os.path.join(path, os.pardir, os.pardir, "Original metadata.json")))
-
 		self.assertTrue(os.path.exists(os.path.join(path, "README_VandeVusse-Mueller.txt")))
-	
 		# Try again, we should fail the second time
 		self.assertIsNone(dataverse.download_dataset(metadata, f.name))
+		f.cleanup()
 
+		# try a harder example, with messed up filenames
+		f = tempfile.TemporaryDirectory()
+		metadata = dataverse.get_metadata("doi:10.5064/F68G8HMM")
+		path = dataverse.download_dataset(metadata, f.name)
+		self.assertTrue(os.path.isdir(path))
+		self.assertTrue(os.path.exists(os.path.join(path, os.pardir, os.pardir, "Original Deposit.zip")))
+		self.assertTrue(os.path.exists(os.path.join(path, os.pardir, os.pardir, "Original metadata.json")))
 		f.cleanup()
 
 class TestGithubAPI(unittest.TestCase):
@@ -77,10 +83,28 @@ def test_version(self):
 class TestRename(unittest.TestCase):
 
 	def test_projectname(self):
+		# test limiting to before the first colon and removing "Data for:"
 		metadata = dataverse.get_metadata("doi:10.5064/F6AQGERV")
 		citation = dataverse.get_citation(metadata)
 		self.assertIsNotNone(citation)
 		self.assertEqual(rename.project_name(citation), "Haney - Child Support Adjudication")
+		# try the limiting to 5 words feature
+		metadata = dataverse.get_metadata("doi:10.5064/F6ZXIJS5")
+		citation = dataverse.get_citation(metadata)
+		self.assertIsNotNone(citation)
+		self.assertEqual(rename.project_name(citation), "Guastaferro - Adapting a Selective Parent-Focused Child")
+		# test special character removal
+		metadata = dataverse.get_metadata("doi:10.5064/F6MBCJ8M")
+		citation = dataverse.get_citation(metadata)
+		self.assertIsNotNone(citation)
+		self.assertEqual(rename.project_name(citation), "Berntzen - Monster or Hero Far-right Responses")
+		# make a temporary directory named the citation and test it exists
+		d = tempfile.TemporaryDirectory()
+		new_folder_name = rename.project_name(citation)
+		new_folder_path = os.path.join(d.name, new_folder_name, "QDR Prepared", "1_extract")
+		os.makedirs(new_folder_path)
+		self.assertTrue(os.path.exists(new_folder_path))  # Ensure it was successfully made
+		d.cleanup()
 
 	def test_rename(self):
 		f = tempfile.TemporaryDirectory()