Fix Bug 66263 — Add support for reading SDT row in tables

[hostedbygnome]

The issue is that there's an SDT at the same level as the table row.
Added support for this case.

CTR elements inside CTSdtRun and CTRow elements inside CTSdtRow are processed recursively.

[pjfanning]

• doesn't even compile
• no tests

[pjfanning]

would it be possible to provide a test?

[hostedbygnome]

would it be possible to provide a test?

yes, sorry for the delay

[hostedbygnome]

Is there anything else required in pr?
@pjfanning

[Copilot]

Pull request overview

Adds support for handling Word SDT (content control) structures that appear at the same level as table rows and runs, so their inner content can be discovered during parsing and text extraction (Bug 66263).

Changes:

• Update XWPFTable to iterate table child elements and recursively process CTSdtRow to extract inner CTRow content.
• Update XWPFParagraph to additionally process CTR elements inside CTSdtBlock / CTSdtRun content.
• Add regression tests and new sample .docx files for SDT rows in tables and SDT runs in paragraphs.

Reviewed changes

Copilot reviewed 4 out of 6 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
poi-ooxml/src/main/java/org/apache/poi/xwpf/usermodel/XWPFTable.java	Cursor-based traversal of table children; recursive SDT-row handling
poi-ooxml/src/main/java/org/apache/poi/xwpf/usermodel/XWPFParagraph.java	Processes CTRs nested inside SDT content when building run lists
poi-ooxml/src/test/java/org/apache/poi/xwpf/usermodel/TestXWPFTable.java	New unit tests for SDT-row table behavior and sample-doc regression
poi-ooxml/src/test/java/org/apache/poi/xwpf/usermodel/TestXWPFParagraph.java	New unit tests for SDT run handling + sample-doc regression
test-data/document/Bug66263-table.docx	Added sample document covering SDT row in table
test-data/document/Bug66263-paragraph.docx	Added sample document covering SDT runs in paragraphs

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The constructor now adds CTRow instances extracted from CTSdtRow into tableRows. Many mutation/access methods (e.g., removeRow, insertNewTableRow, addRow, getRow(int), getNumberOfRows) assume tableRows indexes map 1:1 to ctTbl’s top-level w:tr array. With SDT-derived rows included, removeRow(pos) will remove the wrong underlying w:tr (or throw) and getRow(int)/getNumberOfRows() become inconsistent with getRows(). Consider either (a) keeping tableRows aligned to top-level w:tr only and handling SDT rows separately for text extraction, or (b) refactoring row operations to remove/insert via the actual row element’s cursor (or SDT wrapper) rather than ctTbl.removeTr(pos) and updating getRow/getNumberOfRows semantics accordingly.

[Copilot]

collectCTRowsInnerSdtRow returns immediately when sdtContent.getTrList() is non-empty, which skips any nested w:sdt elements that may appear alongside w:tr within the same CTSdtContentRow. This will miss rows in mixed-content SDTs and also loses document order between tr and nested sdt. Iterate through the SDT content children in order (e.g., via an XmlCursor over sdtContent) and collect both CTRow and nested CTSdtRow recursively without the early return.

Suggested change

	List<CTRow> rowsInnerSdtContent = sdtContent.getTrList();
	if (!rowsInnerSdtContent.isEmpty()) {
	rows.addAll(rowsInnerSdtContent);
	return;
	}
	for (CTSdtRow innerSdt : sdtContent.getSdtList()) {
	collectCTRowsInnerSdtRow(innerSdt, rows);
	XmlCursor cursor = sdtContent.newCursor();
	try {
	if (!cursor.toFirstChild()) {
	return;
	}
	do {
	XmlObject child = cursor.getObject();
	if (child instanceof CTRow) {
	rows.add((CTRow) child);
	} else if (child instanceof CTSdtRow) {
	collectCTRowsInnerSdtRow((CTSdtRow) child, rows);
	}
	} while (cursor.toNextSibling());
	} finally {
	cursor.dispose();

[Copilot]

The "empty table" check only looks at sizeOfTrArray() == 0. A table that contains only w:sdt row wrappers has no top-level w:tr, but it isn’t actually empty; createEmptyTable(table) will inject an extra blank row when using the default constructor (initRow=true). Consider extending the condition to also check for the presence of SDT rows (e.g., sizeOfSdtArray()==0/getSdtList().isEmpty()), or otherwise detect whether the table already contains any row-like content before creating a default row.

[Copilot]

processCTRs adds XWPFRun objects for CTRs inside SDT content into the paragraph’s runs list. Many editing APIs (insertNewRun, insertNewHyperlinkRun, etc.) assume every entry in runs corresponds to a CTR whose cursor is in the paragraph (or hyperlink/field) container; for CTRs nested under CTSdtContentRun, insertNewProvidedRun will fail the isCursorInParagraph check and return null for positions that point at these SDT-derived runs. Either avoid adding nested SDT CTRs into runs (keep SDT content accessible via XWPFSDT/iruns), or update the insertion/removal logic and cursor positioning to support runs whose parent is SDT content so that the runs list remains a valid edit surface.

Suggested change

	private void processCTRs(List<CTR> ctrs) {
	if (ctrs == null) {
	return;
	}
	for (CTR ctr : ctrs) {
	if (ctr.getRPr() != null) {
	private boolean isRunInParagraphEditContainer(CTR ctr) {
	try (XmlCursor cursor = ctr.newCursor()) {
	if (!cursor.toParent()) {
	return false;
	}
	XmlObject parent = cursor.getObject();
	return parent instanceof CTP
	|| parent instanceof CTHyperlink
	|| parent instanceof CTSimpleField;
	}
	}
	private void processCTRs(List<CTR> ctrs) {
	if (ctrs == null) {
	return;
	}
	for (CTR ctr : ctrs) {
	if (ctr.getRPr() != null && isRunInParagraphEditContainer(ctr)) {

[Copilot]

These imports are unused in this test class (CTSdtBlock, CTSdtContentBlock) and will cause compilation to fail with "unused import" in Java. Remove them or add coverage that actually uses SDT blocks.

Suggested change

	import org.openxmlformats.schemas.wordprocessingml.x2006.main.CTSdtBlock;
	import org.openxmlformats.schemas.wordprocessingml.x2006.main.CTSdtContentBlock;