Add validation for invalid html tags

app/error_messages.py

@@ -1,4 +1,5 @@
 DUMB_QUOTES_FOUND = "Found dumb quotes(s) in schema text"
+HTML_FOUND = "Found invalid HTML in schema text"
 INVALID_WHITESPACE_FOUND = "Found invalid white space(s) in schema text"
 DUPLICATE_ID_FOUND = "Duplicate id found"
 FOR_LIST_NEVER_POPULATED = "for_list is not populated by any ListCollector blocks or supplementary data sources"

app/validators/questionnaire_validator.py

@@ -35,6 +35,7 @@ def validate(self):
         self.validate_duplicates()
         self.validate_smart_quotes()
         self.validate_white_spaces()
+        self.validate_html()
         self.validate_answer_references()
         self.validate_list_references()
 
@@ -114,6 +115,47 @@ def validate_smart_quotes(self):
                         pointer=translatable_item.pointer,
                     )
 
+    def validate_html(self):
+        html_strings = []
+        schema_object = SurveySchema(self.schema_element)
+
+        # pylint: disable=invalid-string-quote
+        html_regex = re.compile(r"<[^>]*>")
+
+        for translatable_item in schema_object.translatable_items:
+            schema_text = translatable_item.value
+
+            values_to_check = [schema_text]
+
+            if isinstance(schema_text, dict):
+                values_to_check = schema_text.values()
+
+            html_strings.extend(
+                {"pointer": translatable_item.pointer, "text": schema_text}
+                for schema_text in values_to_check
+                if schema_text and html_regex.search(schema_text)
+            )
+        if html_strings:
+            self.check_invalid_html_tags(html_strings)
+
+    def check_invalid_html_tags(self, html_strings):
+        strong = re.compile(r"<strong>(?:(?!</strong>).)*</strong>")
+        anchor = re.compile(r"<a [^>]*>.*?</a>")
+        all_tags = re.compile(
+            r"<([a-z0-9]+)(?=[\s>])(?:[^>=]|='[^']*'|=\"[^\"]*\"|=[^'\"\s]*)*\s?/?>"
+        )
+
+        for html_string in html_strings:
+            if len(strong.findall(html_string["text"])) + len(
+                anchor.findall(html_string["text"])
+            ) != len(all_tags.findall(html_string["text"])):
+
+                self.add_error(
+                    error_messages.HTML_FOUND,
+                    pointer=html_string["pointer"],
+                    text=html_string["text"],
+                )
+
     def validate_white_spaces(self):
         schema_object = SurveySchema(self.schema_element)
 

tests/schemas/invalid/test_invalid_html_in_schema_text.json

@@ -0,0 +1,133 @@
+{
+  "mime_type": "application/json/ons/eq",
+  "language": "en",
+  "schema_version": "0.0.1",
+  "data_version": "0.0.3",
+  "survey_id": "144",
+  "theme": "default",
+  "title": "Test invalid html",
+  "legal_basis": "Notice is given under section 999 of the Test Act 2000",
+  "metadata": [
+    {
+      "name": "user_id",
+      "type": "string"
+    },
+    {
+      "name": "period_id",
+      "type": "string"
+    },
+    {
+      "name": "ru_name",
+      "type": "string"
+    },
+    {
+      "name": "ru_ref",
+      "type": "string"
+    },
+    {
+      "name": "trad_as",
+      "type": "string",
+      "optional": true
+    }
+  ],
+  "questionnaire_flow": {
+    "type": "Linear",
+    "options": {}
+  },
+  "sections": [
+    {
+      "id": "introduction-section",
+      "title": "Introduction",
+      "groups": [
+        {
+          "id": "introduction-group",
+          "title": "General Business Information",
+          "blocks": [
+            {
+              "id": "introduction",
+              "type": "Introduction",
+              "primary_content": [
+                {
+                  "id": "business-details",
+                  "title": "Introduction with valid and invalid HTML",
+                  "contents": [
+                    {
+                      "guidance": {
+                        "contents": [
+                          {
+                            "title": "<invalid>Coronavirus (COVID-19) guidance</invalid>",
+                            "description": "<strong>Explain your figures</strong> in the comment section to minimise us contacting you and to help us tell an industry story"
+                          }
+                        ]
+                      }
+                    }
+                  ]
+                }
+              ]
+            },
+            {
+              "type": "Interstitial",
+              "id": "intersitital-one",
+              "content": {
+                "title": "Page with invalid html",
+                "contents": [
+                  {
+                    "description": "<p>You have successfully completed this section<p>"
+                  }
+                ]
+              }
+            },
+            {
+              "type": "Interstitial",
+              "id": "interstitial-two",
+              "content": {
+                "title": "Page with link",
+                "contents": [
+                  {
+                    "description": "<a href='link'>Anchor</a>"
+                  }
+                ]
+              }
+            },
+            {
+              "type": "Interstitial",
+              "id": "interstitial-three",
+              "content": {
+                "title": "<strong>Page with mixed invalid tags<strong>",
+                "contents": [
+                  {
+                    "description": "<h1>Title</h1><em>Not valid tag</em>"
+                  }
+                ]
+              }
+            },
+            {
+              "type": "Interstitial",
+              "id": "interstitial-four",
+              "content": {
+                "title": "Valid double <strong>strong</strong> with another <strong>strong</strong>.",
+                "contents": [
+                  {
+                    "description": "<strong>Title</strong><em>Not valid tag</em>"
+                  }
+                ]
+              }
+            },
+            {
+              "type": "Interstitial",
+              "id": "interstitial-five",
+              "content": {
+                "title": "Valid double anchor.",
+                "contents": [
+                  {
+                    "description": "<a href='link'>Title</a> and <a href='link-2'>Not valid tag</a>"
+                  }
+                ]
+              }
+            }
+          ]
+        }
+      ]
+    }
+  ]
+}

tests/schemas/valid/test_introduction_with_guidance.json

@@ -72,7 +72,7 @@
                 "title": "Section complete",
                 "contents": [
                   {
-                    "description": "<p>You have successfully completed this section</p>"
+                    "description": "You have successfully completed this section"
                   }
                 ]
               }

tests/test_questionnaire_validator.py

@@ -340,6 +340,43 @@ def test_invalid_whitespaces_in_schema():
     assert validator.errors == expected_error_messages
 
 
+def test_invalid_html_in_schema():
+    filename = "schemas/invalid/test_invalid_html_in_schema_text.json"
+
+    validator = QuestionnaireValidator(_open_and_load_schema_file(filename))
+
+    expected_error_messages = [
+        {
+            "message": error_messages.HTML_FOUND,
+            "pointer": "/sections/0/groups/0/blocks/3/content/title",
+            "text": "<strong>Page with mixed invalid tags<strong>",
+        },
+        {
+            "message": error_messages.HTML_FOUND,
+            "pointer": "/sections/0/groups/0/blocks/1/content/contents/0/description",
+            "text": "<p>You have successfully completed this section<p>",
+        },
+        {
+            "message": error_messages.HTML_FOUND,
+            "pointer": "/sections/0/groups/0/blocks/3/content/contents/0/description",
+            "text": "<h1>Title</h1><em>Not valid tag</em>",
+        },
+        {
+            "message": error_messages.HTML_FOUND,
+            "pointer": "/sections/0/groups/0/blocks/4/content/contents/0/description",
+            "text": "<strong>Title</strong><em>Not valid tag</em>",
+        },
+        {
+            "message": error_messages.HTML_FOUND,
+            "pointer": "/sections/0/groups/0/blocks/0/primary_content/0/contents/0/guidance/contents/0/title",
+            "text": "<invalid>Coronavirus (COVID-19) guidance</invalid>",
+        },
+    ]
+    validator.validate_html()
+
+    assert validator.errors == expected_error_messages
+
+
 def test_invalid_answer_type_for_question_summary_concatenation():
     filename = "schemas/invalid/test_invalid_answer_type_for_question_summary.json"