Missing invariant check txt-2 on Narrative.div

[Rob5045]

According to the Fhir specification there should be a check on the content of the div to make sure it does not contains white-space only (invariant txt-2).

Link to specification:
https://www.hl7.org/fhir/narrative-definitions.html#Narrative.div

Version used:

• FHIR Version: R4
• Version: 3.4.0

[marcovisserFurore]

Hi @Rob5045,

Looking at the profiles that we ship with the package Hl7.Fhir.Specification.R4, I see that the element Narrative.div has the following constraints:

<constraint>
  <key value="txt-1" />
  <severity value="error" />
  <human value="The narrative SHALL contain only the basic html formatting elements and attributes described in chapters 7-11 (except section 4 of chapter 9) and 15 of the HTML 4.0 standard, &lt;a&gt; elements (either name or href), images and internally contained style attributes" />
  <expression value="htmlChecks()" />
  <xpath value="not(descendant-or-self::*[not(local-name(.)=('a', 'abbr', 'acronym', 'b', 'big', 'blockquote', 'br', 'caption', 'cite', 'code', 'col', 'colgroup', 'dd', 'dfn', 'div', 'dl', 'dt', 'em', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'hr', 'i', 'img', 'li', 'ol', 'p', 'pre', 'q', 'samp', 'small', 'span', 'strong', 'sub', 'sup', 'table', 'tbody', 'td', 'tfoot', 'th', 'thead', 'tr', 'tt', 'ul', 'var'))]) and not(descendant-or-self::*/@*[not(name(.)=('abbr', 'accesskey', 'align', 'alt', 'axis', 'bgcolor', 'border', 'cellhalign', 'cellpadding', 'cellspacing', 'cellvalign', 'char', 'charoff', 'charset', 'cite', 'class', 'colspan', 'compact', 'coords', 'dir', 'frame', 'headers', 'height', 'href', 'hreflang', 'hspace', 'id', 'lang', 'longdesc', 'name', 'nowrap', 'rel', 'rev', 'rowspan', 'rules', 'scope', 'shape', 'span', 'src', 'start', 'style', 'summary', 'tabindex', 'title', 'type', 'valign', 'value', 'vspace', 'width'))])" />
</constraint>
<constraint>
  <key value="txt-2" />
  <severity value="error" />
  <human value="The narrative SHALL have some non-whitespace content" />
  <expression value="htmlChecks()" />
  <xpath value="descendant::text()[normalize-space(.)!=''] or descendant::h:img[@src]" />
</constraint>

So txt-2 is performing the FhirPath function htmlChecks(), which is a function of the extended FhirPath set.

In code you can use it like this :

  var symbolTable = new SymbolTable();
  symbolTable.AddStandardFP();
  symbolTable.AddFhirExtensions();
  
  var fpCompiler = new FhirPathCompiler(symbolTable);

  var compiledExpression = fpCompiler.Compile("htmlChecks()");

  bool succes = compiledExpression .Predicate(instance);

The validator (part of the SDK) already uses this extra extension AddFhirExtensions.

[Rob5045]

Hi Marco,

I wrote a unit test to see if I could produce an error when validating html with white-space content (i.e. trigger txt-2). I was not able to. Here is the unit test (it runs in the context of BasicValidationTests):

        [Theory]
        [InlineData("Rubbish", false)] // Should trigger txt-1 (OK)
        [InlineData("<div xmlns=\"http://www.w3.org/1999/xhtml\"> </div>", false)] // Should trigger txt-2 (NOT OK)
        [InlineData("<div xmlns=\"http://www.w3.org/1999/xhtml\">Not empty</div>", true)] // Valid (OK)
        public async T.Task NarrativeTest(string xml, bool expected)
        {
            output.WriteLine(xml);

            var narrative = new Narrative { Div = xml, Status = Narrative.NarrativeStatus.Generated };

            // DotNetAttributeValidation (used in Forge)
            var list = new List<ValidationResult>();
            var result = DotNetAttributeValidation.TryValidate(narrative, list);
            output.WriteLine($"DotNetAttributeValidation:{result} Expected:{expected}");

            // Validator
            var narrativeSd = await _asyncSource.FindStructureDefinitionForCoreTypeAsync(FHIRAllTypes.Narrative);
            var outcome = _validator.Validate(narrative, narrativeSd);
            output.WriteLine($"Validator:{outcome.Success} Expected:{expected}");

            // htmlChecks()
            var symbolTable = new SymbolTable();
            symbolTable.AddStandardFP();
            symbolTable.AddFhirExtensions();
            var fpCompiler = new FhirPathCompiler(symbolTable);
            var compiledExpression = fpCompiler.Compile("htmlChecks()");
            var typedElement = ElementNode.ForPrimitive(narrative.Div);
            var context = EvaluationContext.CreateDefault();
            bool success = compiledExpression.Predicate(typedElement, context);
            output.WriteLine($"htmlChecks:{success} Expected:{expected}");
        }

Maybe I'm doing something wrong here?

[marcovisserFurore]

Thanks @Rob5045 for the extensive unit tests.

I had a quick look at the source code where we validate the xml in the Narrative to the fhir.xsd and xml.xsd. Those XML schemas should alarm us of any improper use of xml. Although, I see also that we ignore whitespaces before we give the xml to the schema validator. That smells a bit, so I will investigate it a bit deeper. Thanks again for the unit test, that's always a great help for our team!