TIKA-4744 - fix wordml tags

Author: tballison

tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-microsoft-module/src/main/java/org/apache/tika/parser/microsoft/xml/AbstractXML2003Parser.java

@@ -38,6 +38,7 @@
 import org.apache.tika.sax.EmbeddedContentHandler;
 import org.apache.tika.sax.TaggedContentHandler;
 import org.apache.tika.sax.TeeContentHandler;
+import org.apache.tika.sax.XHTMLBalancingHandler;
 import org.apache.tika.sax.XHTMLContentHandler;
 import org.apache.tika.utils.XMLReaderUtils;
 
@@ -88,7 +89,15 @@ public void parse(TikaInputStream tis, ContentHandler handler, Metadata metadata
         final XHTMLContentHandler xhtml = new XHTMLContentHandler(handler, metadata, context);
         xhtml.startDocument();
 
-        TaggedContentHandler tagged = new TaggedContentHandler(xhtml);
+        // Wrap xhtml in a balancing handler so that if the inner SAX parser
+        // aborts mid-element (e.g., on truncated input -- common in crawled
+        // sources clipped to 1MB), we can drain the open elements before
+        // the finally block calls xhtml.endDocument(). Without this drain
+        // the trailing </body></html> mismatches the unclosed <p>/<tr>/etc.
+        // left on the wire, and StrictXHTMLValidator masks the real parse
+        // error with a misleading well-formedness exception.
+        final XHTMLBalancingHandler balancer = new XHTMLBalancingHandler(xhtml);
+        TaggedContentHandler tagged = new TaggedContentHandler(balancer);
         tis.setCloseShield();
         try {
             //need to get new SAXParser because
@@ -99,6 +108,8 @@ public void parse(TikaInputStream tis, ContentHandler handler, Metadata metadata
                             getContentHandler(tagged, metadata, context)));
         } catch (SAXException e) {
             WriteLimitReachedException.throwIfWriteLimitReached(e);
+            // Close anything the aborted parse left open, then propagate.
+            balancer.drainOpenElements();
             throw new TikaException("XML parse error", e);
         } finally {
             tis.removeCloseShield();

tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-microsoft-module/src/main/java/org/apache/tika/parser/microsoft/xml/WordMLParser.java

@@ -121,13 +121,27 @@ public void startElement(String uri, String localName, String qName, Attributes
                         //close p if already in a p to prevent nested <p>
                         if (inP) {
                             handler.endElement(XHTMLContentHandler.XHTML, P, P);
+                            inP = false;
                         }
-                        inP = true;
+                    } else if (html.equals(TABLE) && inP) {
+                        // WordML allows <w:tbl> nested inside <w:p>. XHTML
+                        // does not allow a block-level table inside an
+                        // inline <p>: the inline <p> would still be on the
+                        // SAX stack when the table's <tr>/<td> children
+                        // (and their inner <w:p>'s) emit, and the next
+                        // "close p before nested p" event would land on top
+                        // of <td>/<tr> instead of <p>. Close the open <p>
+                        // before opening the table.
+                        handler.endElement(XHTMLContentHandler.XHTML, P, P);
+                        inP = false;
                     }
                     handler.startElement(XHTMLContentHandler.XHTML, html, html, EMPTY_ATTRS);
                     if (html.equals(TABLE)) {
                         handler.startElement(XHTMLContentHandler.XHTML, TBODY, TBODY, EMPTY_ATTRS);
                     }
+                    if (P.equals(localName)) {
+                        inP = true;
+                    }
                 }
                 if (BR.equals(localName)) {
                     handler.characters(NEWLINE, 0, 1);

tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-microsoft-module/src/test/java/org/apache/tika/parser/microsoft/xml/XML2003ParserTest.java

@@ -63,6 +63,21 @@ public void testBasicExcel() throws Exception {
 
     }
 
+    @Test //TIKA-4744
+    public void testTableInsideParagraph() throws Exception {
+        // WordML allows <w:tbl> nested inside <w:p>. The XHTML emission must
+        // close the <p> before opening the <table>, otherwise the inline
+        // <p> stays on the SAX stack while <tr>/<td> emit -- and the inner
+        // paragraphs inside cells trip the "close p before nested p" logic
+        // emitting </p> while <td> is topmost.
+        // getXML wraps the handler in StrictXHTMLValidator, so any imbalance
+        // would throw before the assertions.
+        XMLResult r = getXML("testWORDML_tableInsideParagraph.doc");
+        assertEquals("application/vnd.ms-wordml", r.metadata.get(Metadata.CONTENT_TYPE));
+        // Structural sanity: tables emit with tbody and rows.
+        assertContains("<table><tbody>", r.xml);
+    }
+
     @Test
     @Timeout(60000)
     public void testMultiThreaded() throws Exception {