Address feedback and fix tests

Signed-off-by: Ayan Sinha Mahapatra <ayansmahapatra@gmail.com>

src/licensedcode/models.py

@@ -24,8 +24,11 @@
 from licensedcode._vendor import attr
 from license_expression import ExpressionError
 from license_expression import Licensing
-from saneyaml import load as yaml_load
-from saneyaml import dump as yaml_dump
+from saneyaml import load as saneyaml_load
+from saneyaml import dump as saneyaml_dump
+from yaml import load as yaml_load
+from yaml import dump as yaml_dump
+from yaml import CSafeLoader
 
 from commoncode.fileutils import file_base_name
 from commoncode.fileutils import file_name
@@ -554,7 +557,7 @@ def spdx_keys(self):
             yield key
 
     @staticmethod
-    def validate(licenses, verbose=False, no_dupe_urls=False):
+    def validate(licenses, verbose=False, no_dupe_urls=False, thorough=False):
         """
         Check that the ``licenses`` a mapping of {key: License} are valid.
         Return dictionaries of infos, errors and warnings mapping a license key
@@ -659,14 +662,23 @@ def validate(licenses, verbose=False, no_dupe_urls=False):
                 if not len(all_licenses) == len(set(all_licenses)):
                     warn('Some duplicated URLs')
 
-            # local text consistency
-            text = lic.text
-
-            data = {"text": text}
-            # We are testing whether we can dump as yaml and load from yaml
-            # without failing (i.e. whether the text is yaml safe) 
-            yaml_string = yaml_dump(data, indent=4)
-            loaded_yaml = yaml_load(yaml_string)
+            if thorough:
+                # local text consistency
+                text = lic.text
+
+                data = {"text": text}
+                # We are testing whether we can dump as yaml and load from yaml
+                # without failing (i.e. whether the text is yaml safe)
+                # Using saneyaml
+                try:
+                    yaml_string = saneyaml_dump(data, indent=4)
+                    loaded_yaml = saneyaml_load(yaml_string)
+                except Exception:
+                    errors['GLOBAL'].append(
+                        f'Error invalid YAML text at: {lic.key}, failed during saneyaml.load()'
+                    )
+                # This fails because of missing line break at text end, added by saneyaml_dump
+                # assert text == loaded_yaml["text"]
 
             license_itokens = tuple(index_tokenizer(text))
             if not license_itokens:
@@ -750,9 +762,9 @@ def validate(licenses, verbose=False, no_dupe_urls=False):
 def get_yaml_safe_text(text):
 
     data = {"text": text}
-    yaml_string = yaml_dump(data, indent=4)
+    yaml_string = saneyaml_dump(data, indent=4)
     try:
-        loaded_yaml = yaml_load(yaml_string)
+        loaded_yaml = saneyaml_load(yaml_string)
     except Exception:
         text = text.replace('\n\n', '\n \n')
     return text
@@ -1028,7 +1040,7 @@ def _validate_all_rules(rules, licenses_by_key):
     errors = defaultdict(list)
 
     for rule in rules:
-        for err_msg in rule.validate(licensing):
+        for err_msg in rule.validate(licensing, thorough=True):
             errors[err_msg].append(rule)
     return errors
 
@@ -1717,7 +1729,7 @@ def has_unknown(self):
         # license flag instead
         return self.license_expression and 'unknown' in self.license_expression
 
-    def validate(self, licensing=None):
+    def validate(self, licensing=None, thorough=False):
         """
         Validate this rule using the provided ``licensing`` Licensing and yield
         one error message for each type of error detected.
@@ -1811,6 +1823,17 @@ def validate(self, licensing=None):
                 if len(set(self.referenced_filenames)) != len(self.referenced_filenames):
                     yield 'referenced_filenames cannot contain duplicates.'
 
+        if thorough:
+            text = self.text
+            data = {"text": text}
+            # We are testing whether we can dump as yaml and load from yaml
+            # without failing (i.e. whether the text is yaml safe)
+            try:
+                yaml_string = saneyaml_dump(data, indent=4)
+                loaded_yaml = saneyaml_load(yaml_string)
+            except Exception:
+                yield (f'Error invalid YAML text at: {self.identifier}, failed during saneyaml.load()')
+
     def license_keys(self, unique=True):
         """
         Return a list of license keys for this rule.

tests/formattedcode/data/yaml/package-and-licenses-expected.yaml

@@ -33,6 +33,10 @@ headers:
         platform_version: '#67-Ubuntu SMP Mon Mar 13 14:22:10 UTC 2023'
         python_version: "3.8.10 (default, Mar 13 2023, 10:26:41) \n[GCC 9.4.0]"
       spdx_license_list_version: '3.20'
+      additional_license_directory: /home/ayansinha/nexB/write_access/scancode-extra/tests/licensedcode/data/additional_licenses/additional_dir
+      additional_license_plugins:
+        - /home/ayansinha/nexB/write_access/scancode-extra/venv/lib/python3.8/site-packages/licenses_to_install1
+        - /home/ayansinha/nexB/write_access/scancode-extra/venv/lib/python3.8/site-packages/licenses_to_install2
       files_count: 4
 summary:
   declared_license_expression: apache-2.0 AND (apache-2.0 OR mit)
@@ -85,6 +89,7 @@ packages:
     code_view_url:
     vcs_url:
     copyright:
+    holder:
     declared_license_expression: apache-2.0
     declared_license_expression_spdx: Apache-2.0
     license_detections:
@@ -1137,6 +1142,7 @@ files:
         code_view_url:
         vcs_url:
         copyright:
+        holder:
         declared_license_expression: apache-2.0
         declared_license_expression_spdx: Apache-2.0
         license_detections:

tests/licensedcode/test_models.py

@@ -88,6 +88,7 @@ def test_validate_license_library_data(self):
         errors, warnings, infos = models.License.validate(
             licenses=models.load_licenses(with_deprecated=False),
             verbose=False,
+            thorough=True,
         )
         assert errors == {}
         assert warnings == {}
@@ -185,6 +186,9 @@ def test_rule_from_license_have_text_file_and_data_file_are_computed_correctly(s
 class TestRule(FileBasedTesting):
     test_data_dir = TEST_DATA_DIR
 
+    def test_validate_license_rules_data(self):
+        rules = list(models.get_rules(validate=True))
+
     def test_create_rule_ignore_punctuation(self):
         test_rule = create_rule_from_text_and_expression(text='A one. A two. A three.')
         expected = ['one', 'two', 'three']