Merge pull request #11950 from IQSS/11900-improved-cvoc-value-validation

BugFix: controlled vocab values validation

Author: sekmiller

src/main/java/edu/harvard/iq/dataverse/DatasetFieldValidator.java

@@ -59,9 +59,25 @@ public boolean isValid(DatasetField value, ConstraintValidatorContext context) {
         }
 
         // if value is not primitive or not empty
-        if (!dsfType.isPrimitive() || !StringUtils.isBlank(value.getValue())) {
+        // For controlled vocabulary fields, check that actual CV values are selected,
+        // not just that datasetFieldValues contains something (which might be an invalid N/A placeholder)
+        // See https://github.com/IQSS/dataverse/issues/11900
+        if (!dsfType.isPrimitive()) {
             return true;
         }
+        
+        if (dsfType.isControlledVocabulary()) {
+            // For CV fields, check if there are actual controlled vocabulary values selected
+            if (value.getControlledVocabularyValues() != null && !value.getControlledVocabularyValues().isEmpty()) {
+                return true;
+            }
+            // If no CV values, fall through to required field check below
+        } else {
+            // For non-CV primitive fields, check if value is not blank
+            if (!StringUtils.isBlank(value.getValue())) {
+                return true;
+            }
+        }
 
         if (value.isRequired()) { 
             String errorMessage = null;

src/main/java/edu/harvard/iq/dataverse/DatasetPage.java

@@ -4039,8 +4039,8 @@ public String save() {
             dataset.setOwner(ownerId != null ? dataverseService.find(ownerId) : null);
         }
         // Validate
-        Set<ConstraintViolation> constraintViolations = workingVersion.validate();
-        if (!constraintViolations.isEmpty()) {
+        workingVersion.validate(); // add validation messages to dataset fields
+        if (!workingVersion.isValid()) {
             FacesContext.getCurrentInstance().validationFailed();
             return "";
         }

src/main/java/edu/harvard/iq/dataverse/engine/command/impl/AbstractDatasetCommand.java

@@ -110,10 +110,14 @@ protected void validateOrDie(DatasetVersion dsv, Boolean lenient) throws Command
         Set<ConstraintViolation> constraintViolations = dsv.validate();
         if (!constraintViolations.isEmpty()) {
             if (lenient) {
-                // populate invalid fields with N/A
+                // populate invalid primitive fields with N/A
+                // Note: controlled vocabulary fields should NOT get N/A values in datasetfieldvalue,
+                // as this creates an inconsistent state where the CV field appears valid but is empty.
+                // See https://github.com/IQSS/dataverse/issues/11900
                 constraintViolations.stream()
                     .filter(cv -> cv.getRootBean() instanceof DatasetField)
                     .map(cv -> ((DatasetField) cv.getRootBean()))
+                    .filter(f -> !f.getDatasetFieldType().isControlledVocabulary())
                     .forEach(f -> f.setSingleValue(DatasetField.NA_VALUE));
 
             } else {