sw/qa/filter/md/md.cxx | 60 ++++++++++++++++++++++++++++++++++++++++++
sw/source/filter/md/wrtmd.cxx | 15 ++++++++--
sw/source/filter/md/wrtmd.hxx | 1
3 files changed, 73 insertions(+), 3 deletions(-)
New commits:
commit 51cce2385a7a7cd6980e13e5a2ed889963b1d72c
Author: Miklos Vajna <vmik...@collabora.com>
AuthorDate: Mon Sep 22 08:35:16 2025 +0200
Commit: Caolán McNamara <caolan.mcnam...@collabora.com>
CommitDate: Mon Sep 22 13:48:56 2025 +0200
tdf#167564 sw markdown export: handle multi-para table cells
Save the bugdoc as markdown,
<https://github.github.com/gfm/#tables-extension-> says only "inlines"
are allowed in a table cell, but multiple paragraphs would be block
elements, which is invalid.
An additional problem is that in case the table cell has 3 paragraphs,
then the middle one won't be even detected as an in-table one, since we
generate "cell info" instances for the start and end of the cell.
Fix the problem by tracking cell ends, so we can emit " " between
in-table paragraphs (but not at the end), and also by checking if we're
in a table (and not if we're at cell start/end) to decide when to strip
away unwanted newlines.
With this, the paragraph breaks are lost during export, but you get
valid output and reasonable fallback for paragraphs in table cells.
Change-Id: I58c5226610ce89c6ac1f066ad87765ed8106be29
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/191308
Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoff...@gmail.com>
Reviewed-by: Caolán McNamara <caolan.mcnam...@collabora.com>
diff --git a/sw/qa/filter/md/md.cxx b/sw/qa/filter/md/md.cxx
index 6096c0829b57..e18ed1a50f17 100644
--- a/sw/qa/filter/md/md.cxx
+++ b/sw/qa/filter/md/md.cxx
@@ -619,6 +619,66 @@ CPPUNIT_TEST_FIXTURE(Test, testImageDescTitleExport)
CPPUNIT_ASSERT_EQUAL(aExpected, aActual);
}
+CPPUNIT_TEST_FIXTURE(Test, testMultiParaTableMdExport)
+{
+ // Given a document that has a table, 3 paragraphs in the A1 cell:
+ createSwDoc();
+ SwDocShell* pDocShell = getSwDocShell();
+ SwWrtShell* pWrtShell = pDocShell->GetWrtShell();
+ pWrtShell->Insert(u"A"_ustr);
+ SwInsertTableOptions aInsertTableOptions(SwInsertTableFlags::DefaultBorder,
+ /*nRowsToRepeat=*/0);
+ pWrtShell->InsertTable(aInsertTableOptions, /*nRows=*/3, /*nCols=*/3);
+ pWrtShell->Insert(u"Z"_ustr);
+ pWrtShell->SttPara();
+ pWrtShell->MoveTable(GotoPrevTable, fnTableStart);
+ pWrtShell->Insert(u"A1 first"_ustr);
+ pWrtShell->SplitNode();
+ pWrtShell->Insert(u"A1 second"_ustr);
+ pWrtShell->SplitNode();
+ pWrtShell->Insert(u"A1 third"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"B1"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"C1"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"A2"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"B2"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"C2"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"A3"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"B3"_ustr);
+ pWrtShell->GoNextCell();
+ pWrtShell->Insert(u"C3"_ustr);
+
+ // When saving that to markdown:
+ save(mpFilter);
+
+ // Then make sure the A1 cell still only has inlines:
+ std::string aActual = TempFileToString();
+ std::string aExpected(
+ // clang-format off
+ "A" SAL_NEWLINE_STRING
+ SAL_NEWLINE_STRING
+ "| A1 first A1 second A1 third | B1 | C1 |" SAL_NEWLINE_STRING
+ "| - | - | - |" SAL_NEWLINE_STRING
+ "| A2 | B2 | C2 |" SAL_NEWLINE_STRING
+ "| A3 | B3 | C3 |" SAL_NEWLINE_STRING
+ SAL_NEWLINE_STRING
+ "Z" SAL_NEWLINE_STRING
+ // clang-format on
+ );
+ // Without the accompanying fix in place, this test would have failed with:
+ // - Expected: A1 first A1 second A1 third
+ // - Actual : A1 first
A1 second
A1 third
+ // i.e. multiple paragraphs were not merged into a single paragraph to
form just a list of
+ // inline blocks, as required by the spec.
+ CPPUNIT_ASSERT_EQUAL(aExpected, aActual);
+}
+
CPPUNIT_PLUGIN_IMPLEMENT();
/* vim:set shiftwidth=4 softtabstop=4 expandtab cinoptions=b1,g0,N-s
cinkeys+=0=break: */
diff --git a/sw/source/filter/md/wrtmd.cxx b/sw/source/filter/md/wrtmd.cxx
index 50bf3810d917..67f02bf926d3 100644
--- a/sw/source/filter/md/wrtmd.cxx
+++ b/sw/source/filter/md/wrtmd.cxx
@@ -472,8 +472,10 @@ void OutMarkdown_SwTextNode(SwMDWriter& rWrt, const
SwTextNode& rNode, bool bFir
{
std::optional<SwMDCellInfo> oCellInfo;
std::stack<SwMDTableInfo>& rTableInfos = rWrt.GetTableInfos();
+ bool bInTable = false;
if (!rTableInfos.empty())
{
+ bInTable = true;
SwMDTableInfo& aTableInfo = rTableInfos.top();
auto it = aTableInfo.aCellInfos.find(rNode.GetIndex());
if (it != aTableInfo.aCellInfos.end())
@@ -499,7 +501,7 @@ void OutMarkdown_SwTextNode(SwMDWriter& rWrt, const
SwTextNode& rNode, bool bFir
if (!rNodeText.isEmpty())
{
// Paragraphs separate by empty lines
- if (!bFirst && !oCellInfo)
+ if (!bFirst && !bInTable)
rWrt.Strm().WriteUnicodeOrByteText(u"" SAL_NEWLINE_STRING);
const SwFormatColl* pFormatColl = rNode.GetFormatColl();
@@ -724,7 +726,13 @@ void OutMarkdown_SwTextNode(SwMDWriter& rWrt, const
SwTextNode& rNode, bool bFir
}
}
- if (!oCellInfo || bRowEnd)
+ bool bCellEnd = oCellInfo && oCellInfo->bCellEnd;
+ if (bInTable && !bCellEnd)
+ {
+ // Separator is a space between two in-table-cell paragraphs.
+ rWrt.Strm().WriteUnicodeOrByteText(u" ");
+ }
+ else if (!bInTable || bRowEnd)
{
rWrt.Strm().WriteUnicodeOrByteText(u"" SAL_NEWLINE_STRING);
}
@@ -767,9 +775,10 @@ void OutMarkdown_SwTableNode(SwMDWriter& rWrt, const
SwTableNode& rTableNode)
{
rStartInfo.bRowStart = true;
}
+ SwMDCellInfo& rEndInfo = aTableInfo.aCellInfos[pEnd->GetIndex() -
1];
+ rEndInfo.bCellEnd = true;
if (nBox == pLine->GetTabBoxes().size() - 1)
{
- SwMDCellInfo& rEndInfo =
aTableInfo.aCellInfos[pEnd->GetIndex() - 1];
rEndInfo.bRowEnd = true;
if (nLine == 0)
{
diff --git a/sw/source/filter/md/wrtmd.hxx b/sw/source/filter/md/wrtmd.hxx
index d3f6d15db58f..77b6e965c792 100644
--- a/sw/source/filter/md/wrtmd.hxx
+++ b/sw/source/filter/md/wrtmd.hxx
@@ -33,6 +33,7 @@
struct SwMDCellInfo
{
bool bCellStart = false;
+ bool bCellEnd = false;
bool bRowStart = false;
bool bRowEnd = false;
bool bFirstRowEnd = false;
