Open files that contain U+2029

src/FindBar.qml

@@ -33,7 +33,7 @@ Rectangle {
         replacing = replace
         visible = true
         const area = editor.area
-        if (area.selectionStart !== area.selectionEnd && area.selectedText.indexOf("
") < 0)
+        if (area.selectionStart !== area.selectionEnd && area.selectedText.indexOf("\u2029") < 0)
             findField.text = area.selectedText
         findField.forceActiveFocus()
         findField.selectAll()

src/document.cpp

@@ -71,10 +71,16 @@ Document::Document(QObject *parent) : QObject(parent) {
     connect(&m_debounce, &QTimer::timeout, this, &Document::diskChanged);
 }
 
+// QTextDocument separates its blocks with U+2029, so one in the file would
+// become a line break. It's held in the editor as U+FDEF instead, and turned
+// back on save: a noncharacter, which Unicode sets aside for a program's own
+// use, so text typed or pasted in won't hold one.
+static const QChar ParagraphStandIn(0xFDEF);
+
 // Bytes to editable text, or not ok when saving the text again couldn't give
 // back the same file: invalid UTF-8 (a sequence cut off at the end too), NUL
-// bytes, carriage returns that aren't part of CRLF, or U+2029, which
-// QTextDocument turns into a line break.
+// bytes, carriage returns that aren't part of CRLF, or the U+2029 stand-in
+// itself, which saving would turn into U+2029.
 Document::Decoded Document::decode(const QByteArray &input) {
     Decoded result;
     QByteArray bytes = input;
@@ -91,8 +97,9 @@ Document::Decoded Document::decode(const QByteArray &input) {
     QStringDecoder decoder(QStringConverter::Utf8,
                            QStringConverter::Flag::Stateless | QStringConverter::Flag::ConvertInitialBom);
     QString text = decoder.decode(bytes);
-    if (decoder.hasError() || text.contains(QChar::ParagraphSeparator))
+    if (decoder.hasError() || text.contains(ParagraphStandIn))
         return result;
+    text.replace(QChar::ParagraphSeparator, ParagraphStandIn);
 
     int crlf = 0, lf = 0;
     for (qsizetype i = 0; i < text.size(); ++i) {
@@ -121,7 +128,7 @@ QByteArray Document::encode(const QTextDocument *document, bool bom, bool crlf)
     QString text;
     const QString ending = crlf ? "\r\n" : "\n";
     for (QTextBlock block = document->begin(); block.isValid(); block = block.next()) {
-        text += block.text();
+        text += block.text().replace(ParagraphStandIn, QChar::ParagraphSeparator);
         if (block.next().isValid())
             text += ending;
     }

src/document.h

@@ -37,7 +37,7 @@ class Document : public QObject {
         NotFound,        // the path doesn't exist and its folder doesn't either
         NotRegular,      // a folder, device or pipe
         NotReadable,     // permission denied: belongs to root or someone else
-        NotText,         // not UTF-8, NUL bytes, lone CRs, or U+2029
+        NotText,         // not UTF-8, NUL bytes, lone CRs, or U+FDEF
         TooLarge,        // big enough that TextEdit would be slow; asks first
         MixedEndings,    // opened fine; will be saved with the majority ending
         ChangedOnDisk,   // changed elsewhere while there are unsaved edits

tests/plain_tests.cpp

@@ -269,7 +269,7 @@ private slots:
         QVERIFY(!Document::decode(QByteArray("a\0b", 3)).ok);
         QVERIFY(!Document::decode("caf\xE9\n").ok);           // Latin-1, not UTF-8
         QVERIFY(!Document::decode("a\rb\n").ok);              // lone CR
-        QVERIFY(!Document::decode("a\xE2\x80\xA9" "b").ok);   // U+2029
+        QVERIFY(!Document::decode("a\xEF\xB7\xAF" "b").ok);   // U+FDEF, U+2029's stand-in
         QVERIFY(Document::decode("").ok);
         // A sequence cut off at the end, which a stateful decoder holds back
         // without an error, so saving would drop it.
@@ -295,6 +295,7 @@ private slots:
         QTest::newRow("double bom") << QByteArray("\xEF\xBB\xBF\xEF\xBB\xBF" "x\n");
         QTest::newRow("nbsp and tabs") << QByteArray("a\xC2\xA0" "b\n\tc\n");
         QTest::newRow("line separator") << QByteArray("a\xE2\x80\xA8" "b\n");
+        QTest::newRow("paragraph separator") << QByteArray("a\xE2\x80\xA9" "b\n");
         QTest::newRow("empty") << QByteArray();
         QTest::newRow("only newline") << QByteArray("\n");
         QTest::newRow("unicode") << QByteArray("名前 = \"ø\" # 🎉\n");