fix: deduplicate internetMediaType elements

Author: phette23

collection-export/strict-mods.js

@@ -795,6 +795,53 @@ export function convertSpeakerReleaseDetail(doc) {
     return doc
 }
 
+/**
+ * Deduplicate internetMediaType elements within physicalDescription
+ * Items with multiple attachments often have duplicate MIME types listed.
+ * Keep only the first occurrence of each unique value, preserving order.
+ *
+ * @param {Document} doc - XML DOM document
+ * @returns {Document} Modified document
+ */
+export function deduplicateInternetMediaType(doc) {
+    if (!doc) {
+        return doc
+    }
+
+    const physicalDescriptions = safeSelect(XPATH_CONTEXTS.PHYSICAL_DESCRIPTION, doc)
+    const select = xpath.useNamespaces({})
+
+    for (let physDesc of physicalDescriptions) {
+        const mediaTypes = select('internetMediaType', physDesc)
+        
+        if (mediaTypes.length <= 1) {
+            continue
+        }
+
+        const seen = new Set()
+        const toRemove = []
+
+        for (let mediaType of mediaTypes) {
+            const value = mediaType.textContent.trim()
+            
+            if (seen.has(value)) {
+                // Duplicate, mark for removal
+                toRemove.push(mediaType)
+            } else {
+                // First occurrence, keep it
+                seen.add(value)
+            }
+        }
+
+        // Remove duplicates
+        for (let element of toRemove) {
+            physDesc.removeChild(element)
+        }
+    }
+
+    return doc
+}
+
 /**
  * Wrap copyInformation in holdingSimple element
  * MODS standard requires: location/holdingSimple/copyInformation
@@ -1270,6 +1317,9 @@ export function toStrictMODS(xmlString) {
     // Wrap recordInfo/languageOfCataloging with languageTerm and move authority attribute
     wrapTextWithChild(doc, `${XPATH_CONTEXTS.RECORD_INFO}/${ELEMENT_NAMES.LANGUAGE_OF_CATALOGING}`, ELEMENT_NAMES.LANGUAGE_TERM, [ATTRIBUTES.AUTHORITY])
 
+    // Deduplicate internetMediaType in physicalDescription
+    deduplicateInternetMediaType(doc)
+
     // Remove all empty elements (no text, no children with text)
     removeEmptyElements(doc)
 

collection-export/strict-mods.test.js

@@ -3218,6 +3218,62 @@ describe('Strict MODS Conversion', () => {
                 'Should not have direct title under relatedItem')
         })
 
+        it('should deduplicate internetMediaType in physicalDescription', () => {
+            const input = `<xml><mods>
+                <titleInfo><title>Test</title></titleInfo>
+                <physicalDescription>
+                    <internetMediaType>image/tiff</internetMediaType>
+                    <internetMediaType>image/jpeg</internetMediaType>
+                    <internetMediaType>image/tiff</internetMediaType>
+                    <internetMediaType>image/jpeg</internetMediaType>
+                    <digitalOrigin>reformatted digital</digitalOrigin>
+                </physicalDescription>
+            </mods></xml>`
+            const result = toStrictMODS(input)
+
+            // Should only have one of each type
+            const tiffMatches = result.match(/<internetMediaType>image\/tiff<\/internetMediaType>/g) || []
+            const jpegMatches = result.match(/<internetMediaType>image\/jpeg<\/internetMediaType>/g) || []
+            
+            assert.strictEqual(tiffMatches.length, 1, 'Should have exactly one image/tiff')
+            assert.strictEqual(jpegMatches.length, 1, 'Should have exactly one image/jpeg')
+        })
+
+        it('should deduplicate internetMediaType while preserving order', () => {
+            const input = `<xml><mods>
+                <titleInfo><title>Test</title></titleInfo>
+                <physicalDescription>
+                    <internetMediaType>image/tiff</internetMediaType>
+                    <internetMediaType>image/jpeg</internetMediaType>
+                    <internetMediaType>application/pdf</internetMediaType>
+                    <internetMediaType>image/tiff</internetMediaType>
+                    <internetMediaType>image/jpeg</internetMediaType>
+                </physicalDescription>
+            </mods></xml>`
+            const result = toStrictMODS(input)
+
+            // Should have each type once, in order of first appearance
+            const types = [...result.matchAll(/<internetMediaType>(.*?)<\/internetMediaType>/g)].map(m => m[1])
+            
+            assert.deepStrictEqual(types, ['image/tiff', 'image/jpeg', 'application/pdf'],
+                'Should preserve order of first appearance')
+        })
+
+        it('should handle physicalDescription with no internetMediaType', () => {
+            const input = `<xml><mods>
+                <titleInfo><title>Test</title></titleInfo>
+                <physicalDescription>
+                    <digitalOrigin>born digital</digitalOrigin>
+                    <extent>10 pages</extent>
+                </physicalDescription>
+            </mods></xml>`
+            const result = toStrictMODS(input)
+
+            assert.ok(result.includes('<digitalOrigin>born digital</digitalOrigin>'), 'Should preserve digitalOrigin')
+            assert.ok(result.includes('<extent>10 pages</extent>'), 'Should preserve extent')
+        })
+
+
         it('should move formBroad and formSpecific to genre', () => {
             const input = `<xml><mods>
                 <titleInfo><title>Test Item</title></titleInfo>