Close a frontmatter block at YAML's ... document end marker [patch]

Frontmatter.Test/DelimiterLineTests.cs

@@ -191,4 +191,75 @@ public void HasFrontmatter_DelimiterWithoutLineEnding_IsNotFrontmatter()
 		Assert.IsFalse(Frontmatter.HasFrontmatter("\uFEFF"));
 		Assert.IsFalse(Frontmatter.HasFrontmatter(string.Empty));
 	}
+
+	private static readonly string PandocDocument =
+		$"---{Nl}title: A{Nl}...{Nl}# Heading{Nl}{Nl}Important paragraph.{Nl}{Nl}---{Nl}{Nl}More text{Nl}";
+
+	[TestMethod]
+	public void ExtractFrontmatter_BlockClosedByDocumentEndMarker_ReadsTheBlock()
+	{
+		Dictionary<string, object>? frontmatter = Frontmatter.ExtractFrontmatter(PandocDocument);
+
+		Assert.IsNotNull(frontmatter);
+		Assert.HasCount(1, frontmatter);
+		Assert.AreEqual("A", frontmatter["title"]);
+	}
+
+	[TestMethod]
+	public void ExtractBody_BlockClosedByDocumentEndMarker_KeepsTheBodyBeforeALaterRule()
+	{
+		string body = Frontmatter.ExtractBody(PandocDocument);
+
+		Assert.AreEqual($"# Heading{Nl}{Nl}Important paragraph.{Nl}{Nl}---{Nl}{Nl}More text", body);
+	}
+
+	[TestMethod]
+	public void RemoveFrontmatter_BlockClosedByDocumentEndMarker_KeepsTheBodyBeforeALaterRule()
+	{
+		string result = Frontmatter.RemoveFrontmatter(PandocDocument);
+
+		Assert.AreEqual($"# Heading{Nl}{Nl}Important paragraph.{Nl}{Nl}---{Nl}{Nl}More text{Nl}", result);
+	}
+
+	[TestMethod]
+	public void ReplaceFrontmatter_BlockClosedByDocumentEndMarker_ReplacesOnlyTheBlock()
+	{
+		string result = Frontmatter.ReplaceFrontmatter(PandocDocument, new() { { "title", "B" } });
+
+		Assert.AreEqual($"---{Nl}title: B{Nl}---{Nl}# Heading{Nl}{Nl}Important paragraph.{Nl}{Nl}---{Nl}{Nl}More text{Nl}", result);
+	}
+
+	[TestMethod]
+	public void ExtractFrontmatter_BlockClosedByDocumentEndMarkerWithNoLaterRule_ReadsTheBlock()
+	{
+		string input = $"---{Nl}title: A{Nl}...   {Nl}Body{Nl}";
+
+		Dictionary<string, object>? frontmatter = Frontmatter.ExtractFrontmatter(input);
+
+		Assert.IsNotNull(frontmatter);
+		Assert.AreEqual("A", frontmatter["title"]);
+		Assert.AreEqual("Body", Frontmatter.ExtractBody(input));
+	}
+
+	[TestMethod]
+	public void ExtractFrontmatter_DocumentEndMarkerOnTheFirstLine_IsNotAnOpener()
+	{
+		string input = $"...{Nl}title: A{Nl}---{Nl}Body{Nl}";
+
+		Assert.IsFalse(Frontmatter.HasFrontmatter(input));
+		Assert.IsNull(Frontmatter.ExtractFrontmatter(input));
+		Assert.AreEqual(input, Frontmatter.RemoveFrontmatter(input));
+	}
+
+	[TestMethod]
+	public void ExtractFrontmatter_IndentedDocumentEndMarkerInsideAValue_DoesNotCloseTheBlock()
+	{
+		string input = $"---{Nl}notes: |{Nl}  ...{Nl}title: A{Nl}---{Nl}Body{Nl}";
+
+		Dictionary<string, object>? frontmatter = Frontmatter.ExtractFrontmatter(input);
+
+		Assert.IsNotNull(frontmatter);
+		Assert.AreEqual("A", frontmatter["title"]);
+		Assert.AreEqual("Body", Frontmatter.ExtractBody(input));
+	}
 }

Frontmatter/Frontmatter.cs

@@ -15,6 +15,7 @@
 	/// The delimiter that marks the beginning and end of a frontmatter section.
 	/// </summary>
 	private const string FrontmatterDelimiter = "---";
+	private const string DocumentEndMarker = "...";
 
 	/// <summary>
 	/// The byte order mark a document may begin with when it was decoded without stripping it.
@@ -136,7 +137,7 @@
 		}
 
 		List<Dictionary<string, object>> frontmatterObjects = ExtractFrontmatterObjects(input, out _);
 		return frontmatterObjects.Count > 0 ? frontmatterObjects.First() : null;
 	}
 
 	/// <summary>
@@ -283,7 +284,7 @@
 		}
 
 		// Then add any remaining properties that weren't in the standard order
 		foreach (KeyValuePair<string, object> property in frontmatter)
 		{
 			if (!sortedFrontmatter.ContainsKey(property.Key))
 			{
@@ -333,7 +334,7 @@
 				continue;
 			}
 
 			if (YamlSerializer.TryParseYamlObject(section, out Dictionary<string, object>? frontmatterObject) && frontmatterObject != null)
 			{
 				frontmatterObjects.Add(frontmatterObject);
 			}
@@ -368,7 +369,9 @@
 	/// <remarks>
 	/// Delimiters are recognised only as whole lines, so a <c>---</c> inside a value or a markdown
 	/// horizontal rule in the body is never mistaken for one. The first line opens a block, which closes
-	/// at the next delimiter line, including one at the very end of the document. Further blocks are
+	/// at the next delimiter line, including one at the very end of the document. YAML's document end
+	/// marker <c>...</c>, which Pandoc metadata blocks use, also closes a block but never opens one; a
+	/// rewritten document always closes its frontmatter with <c>---</c>. Further blocks are
 	/// consumed only while each opens on the line straight after the previous one closed and holds
 	/// frontmatter rather than body text (see <see cref="IsFollowOnBlock"/>); the body is everything after
 	/// the last block consumed.
@@ -390,12 +393,13 @@
 		List<(int Start, int End)> lines = SplitLines(input);
 		lines[0] = (OpeningLineStart(input), lines[0].End);
 		bool IsDelimiterAt(int index) => IsDelimiterLine(input[lines[index].Start..lines[index].End]);
+		bool IsCloserAt(int index) => IsClosingDelimiterLine(input[lines[index].Start..lines[index].End]);
 
 		int next = 0;
 		while (next < lines.Count && IsDelimiterAt(next))
 		{
 			int close = next + 1;
-			while (close < lines.Count && !IsDelimiterAt(close))
+			while (close < lines.Count && !IsCloserAt(close))
 			{
 				close++;
 			}
@@ -478,7 +482,7 @@
 			lines.Add((start, i));
 			if (c == '\r' && i + 1 < input.Length && input[i + 1] == '\n')
 			{
 				i++;
 			}
 
 			start = i + 1;
@@ -521,6 +525,15 @@
 	/// <returns>True if the line is a frontmatter delimiter, false otherwise.</returns>
 	private static bool IsDelimiterLine(string line) => line.TrimEnd() == FrontmatterDelimiter;
 
+	/// <summary>
+	/// Checks whether a line can close a frontmatter block: a delimiter, or YAML's document end marker
+	/// <c>...</c>, either allowing trailing whitespace.
+	/// </summary>
+	/// <param name="line">The line to check, without its line ending.</param>
+	/// <returns>True if the line closes a frontmatter block, false otherwise.</returns>
+	private static bool IsClosingDelimiterLine(string line) =>
+		IsDelimiterLine(line) || line.TrimEnd() == DocumentEndMarker;
+
 	/// <summary>
 	/// Combines two frontmatter dictionaries into a single dictionary.
 	/// </summary>
@@ -539,7 +552,7 @@
 		}
 
 		// Then, add properties from dictionary b that don't exist in a
 		foreach (KeyValuePair<string, object> kvp in b)
 		{
 			if (!combinedFrontmatterObject.ContainsKey(kvp.Key))
 			{