Fix roundtrip lang literal conversion

sheet_to_triples/rdf.py

@@ -4,6 +4,7 @@
 """Context specific RDF behaviours for Visual Meaning graphs."""
 
 import functools
+import json
 import operator
 import re
 
@@ -55,8 +56,14 @@ def from_identifier(self, value):
         value = _norm(value)
         # if ends with language tag, create a Literal with the appropriate lang
         if re.search(self.lang_match, value):
-            inner = value[1:-4] if value[0] == '"' else value[:-3]
-            return rdflib.Literal(inner, lang=value[-2:])
+            lang = value[-2:]
+            value_without_lang = value[:-3]
+            if value_without_lang.startswith('"') and value_without_lang.endswith('"'):
+                inner = json.loads(value_without_lang)
+            else:
+                # {"k": "v"}@en compatibility
+                inner = value_without_lang
+            return rdflib.Literal(inner, lang=lang)
         return rdflib.Literal(value)
 
 
@@ -138,9 +145,12 @@ def _maybe_from_literal(maybe_literal):
     them in the json serialisation for both strings and arrays or objects.
     """
     if getattr(maybe_literal, 'language', None):
+        # For JSON arrays/objects, str() gives the JSON representation
         if maybe_literal[:1] + maybe_literal[-1:] in ('[]', '{}'):
             return str(maybe_literal) + '@' + maybe_literal.language
-        return maybe_literal.n3()
+        # For regular strings, use json.dumps to properly escape
+        # This ensures roundtrip compatibility with from_identifier
+        return json.dumps(str(maybe_literal)) + '@' + maybe_literal.language
     return str(maybe_literal)
 
 

sheet_to_triples/tests/test_rdf.py

@@ -4,6 +4,7 @@
 """Unittests for the rdf.py module of sheet-to-triples."""
 
 import io
+import json
 import unittest
 from unittest import mock
 
@@ -158,15 +159,20 @@ def test_update_model_terms_language_tags(self):
             # Not testing rdflib.Literal(1) as should maybe change behaviour?
             ('_s', '_p', rdflib.Literal('o')),
             ('_s', '_p', rdflib.Literal('o', lang='en')),
+            ('_s', '_p', rdflib.Literal(
+                'thing: "quotes"\nand newlines\tand tabs', lang='en')),
             ('_s', '_p', rdflib.Literal('\u043e', lang='bg')),
+            ('_s', '_p', rdflib.Literal('л', lang='bg')),
             ('_s', '_p', rdflib.Literal('', lang='en')),
             ('_s', '_p', rdflib.Literal('["o1", "o2"]', lang='en')),
             ('_s', '_p', rdflib.Literal('{"o": "v"}', lang='en')),
         ]
         rdf.update_model_terms(model, triples)
         expected = [
-            'o:_1', 'o', '"o"@en', '"\u043e"@bg', '""@en', '["o1", "o2"]@en',
-            '{"o": "v"}@en',
+            'o:_1', 'o', '"o"@en',
+            '"thing: \\"quotes\\"\\nand newlines\\tand tabs"@en',
+            '"\\u043e"@bg', '"\\u043b"@bg', '""@en',
+            '["o1", "o2"]@en', '{"o": "v"}@en',
         ]
         self.assertEqual([t['obj'] for t in model['terms']], expected)
 
@@ -422,3 +428,32 @@ def test_from_identifier_lang_obj(self):
             self.resolver.from_identifier('{"k": "v"}@en'),
             rdflib.Literal('{"k": "v"}', lang='en')
         )
+
+    def test_from_identifier_lang_roundtrip(self):
+        """
+            Test the roundtrip workflow:
+            string -> from_identifier -> Literal -> update_model_terms -> string.
+        """
+        test_texts = [
+            ('simple text', 'en'),
+            ('text\nwith\nnewlines', 'en'),
+            ('text with "quotes"', 'fr'),
+            ('text\twith\ttabs', 'de'),
+            ('path\\with\\backslashes', 'es'),
+            ('mixed: "quotes"\nand newlines\tand tabs', 'it'),
+            ('', 'en'),
+            ('о', 'bg'), # cyrillic o
+            ('л', 'bg'),
+        ]
+
+        for text, lang in test_texts:
+            with self.subTest(text=text, lang=lang):
+                initial_string = json.dumps(text) + '@' + lang
+                literal = self.resolver.from_identifier(initial_string)
+
+                model = {'terms': []}
+                triples = [('_s', '_p', literal)]
+                rdf.update_model_terms(model, triples)
+
+                resulting_string = model['terms'][0]['obj']
+                self.assertEqual(initial_string, resulting_string)