Merge pull request #753 from MihaZupan/perf-nov23-3

A few more perf improvements

src/Markdig/Helpers/EntityHelper.cs

@@ -31,6 +31,8 @@
 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
+using System.Text;
+
 namespace Markdig.Helpers;
 
 /// <summary>
@@ -57,41 +59,31 @@ public static class EntityHelper
     /// <returns>The unicode character set or <c>null</c> if the entity was not recognized.</returns>
     public static string DecodeEntity(int utf32)
     {
-        if (!CharHelper.IsInInclusiveRange(utf32, 1, 1114111) || CharHelper.IsInInclusiveRange(utf32, 55296, 57343))
+        if (utf32 == 0 || !UnicodeUtility.IsValidUnicodeScalar((uint)utf32))
             return CharHelper.ReplacementCharString;
 
-        if (utf32 < 65536)
+        if (UnicodeUtility.IsBmpCodePoint((uint)utf32))
             return char.ToString((char)utf32);
 
-        utf32 -= 65536;
-        return new string(
-#if NETSTANDARD2_1_OR_GREATER || NETCOREAPP3_1_OR_GREATER
-            stackalloc
-#else
-            new
-#endif
-            char[]
-        {
-            (char)((uint)utf32 / 1024 + 55296),
-            (char)((uint)utf32 % 1024 + 56320)
-        });
+        UnicodeUtility.GetUtf16SurrogatesFromSupplementaryPlaneScalar((uint)utf32, out char high, out char low);
+        return new string([high, low]);
     }
 
     internal static void DecodeEntity(int utf32, ref ValueStringBuilder sb)
     {
-        if (!CharHelper.IsInInclusiveRange(utf32, 1, 1114111) || CharHelper.IsInInclusiveRange(utf32, 55296, 57343))
+        if (utf32 == 0 || !UnicodeUtility.IsValidUnicodeScalar((uint)utf32))
         {
             sb.Append(CharHelper.ReplacementChar);
         }
-        else if (utf32 < 65536)
+        else if (UnicodeUtility.IsBmpCodePoint((uint)utf32))
         {
             sb.Append((char)utf32);
         }
         else
         {
-            utf32 -= 65536;
-            sb.Append((char)((uint)utf32 / 1024 + 55296));
-            sb.Append((char)((uint)utf32 % 1024 + 56320));
+            UnicodeUtility.GetUtf16SurrogatesFromSupplementaryPlaneScalar((uint)utf32, out char high, out char low);
+            sb.Append(high);
+            sb.Append(low);
         }
     }
 

src/Markdig/Helpers/ThrowHelper.cs

@@ -80,7 +80,7 @@ public static void CheckDepthLimit(int depth, bool useLargeLimit = false)
         if (depth > limit)
             DepthLimitExceeded();
 
-        [MethodImpl(MethodImplOptions.NoInlining)]
+        [DoesNotReturn]
         static void DepthLimitExceeded() => throw new ArgumentException("Markdown elements in the input are too deeply nested - depth limit exceeded. Input is most likely not sensible or is a very large table.");
     }
 

src/Markdig/Helpers/UnicodeUtility.cs

@@ -0,0 +1,30 @@
+// Copyright (c) Alexandre Mutel. All rights reserved.
+// This file is licensed under the BSD-Clause 2 license.
+// See the license.txt file in the project root for more information.
+
+using System.Diagnostics;
+using System.Runtime.CompilerServices;
+
+namespace System.Text;
+
+// Based on https://github.com/dotnet/runtime/blob/main/src/libraries/System.Private.CoreLib/src/System/Text/UnicodeUtility.cs
+internal static class UnicodeUtility
+{
+    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+    public static bool IsBmpCodePoint(uint value) => value <= 0xFFFFu;
+
+    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+    public static bool IsValidUnicodeScalar(uint value)
+    {
+        return ((value - 0x110000u) ^ 0xD800u) >= 0xFFEF0800u;
+    }
+
+    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+    public static void GetUtf16SurrogatesFromSupplementaryPlaneScalar(uint value, out char highSurrogateCodePoint, out char lowSurrogateCodePoint)
+    {
+        Debug.Assert(IsValidUnicodeScalar(value) && IsBmpCodePoint(value));
+
+        highSurrogateCodePoint = (char)((value + ((0xD800u - 0x40u) << 10)) >> 10);
+        lowSurrogateCodePoint = (char)((value & 0x3FFu) + 0xDC00u);
+    }
+}

src/Markdig/Parsers/FencedBlockParserBase.cs

@@ -40,7 +40,8 @@ public abstract class FencedBlockParserBase : BlockParser, IAttributesParseable
 /// <seealso cref="BlockParser" />
 public abstract class FencedBlockParserBase<T> : FencedBlockParserBase where T : Block, IFencedBlock
 {
-    private static readonly TransformedStringCache _infoStringCache = new(static infoString => HtmlHelper.Unescape(infoString));
+    private static readonly TransformedStringCache s_infoStringCache = new(static infoString => HtmlHelper.Unescape(infoString));
+    private static readonly TransformedStringCache s_argumentsStringCache = new(static argumentsString => HtmlHelper.Unescape(argumentsString));
     private TransformedStringCache? _infoPrefixCache;
 
     /// <summary>
@@ -176,7 +177,7 @@ public static bool RoundtripInfoParser(BlockProcessor blockProcessor, ref String
 
     end:
         fenced.TriviaAfterFencedChar = afterFence;
-        fenced.Info = _infoStringCache.Get(info.AsSpan());
+        fenced.Info = s_infoStringCache.Get(info.AsSpan());
         fenced.UnescapedInfo = info;
         fenced.TriviaAfterInfo = afterInfo;
         fenced.Arguments = HtmlHelper.Unescape(arg.ToString());
@@ -197,71 +198,47 @@ public static bool RoundtripInfoParser(BlockProcessor blockProcessor, ref String
     /// <returns><c>true</c> if parsing of the line is successfull; <c>false</c> otherwise</returns>
     public static bool DefaultInfoParser(BlockProcessor state, ref StringSlice line, IFencedBlock fenced, char openingCharacter)
     {
-        // An info string cannot contain any backticks (unless it is a tilde block)
         int firstSpace = -1;
-        if (openingCharacter == '`')
+        ReadOnlySpan<char> lineSpan = line.AsSpan();
+
+        if (!lineSpan.IsEmpty)
         {
-            for (int i = line.Start; i <= line.End; i++)
+            if (openingCharacter == '`')
             {
-                char c = line.Text[i];
-                if (c == '`')
-                {
-                    return false;
-                }
+                firstSpace = lineSpan.IndexOfAny(' ', '\t', '`');
 
-                if (firstSpace < 0 && c.IsSpaceOrTab())
+                // An info string cannot contain any backticks (unless it is a tilde block)
+                if (firstSpace >= 0 && lineSpan.Slice(firstSpace).Contains('`'))
                 {
-                    firstSpace = i;
+                    return false;
                 }
             }
-        }
-        else
-        {
-            for (int i = line.Start; i <= line.End; i++)
+            else
             {
-                if (line.Text[i].IsSpaceOrTab())
-                {
-                    firstSpace = i;
-                    break;
-                }
+                firstSpace = lineSpan.IndexOfAny(' ', '\t');
             }
         }
 
         StringSlice infoStringSlice;
-        string? argString = null;
 
-        if (firstSpace > 0)
+        if (firstSpace >= 0)
         {
+            firstSpace += line.Start;
             infoStringSlice = new StringSlice(line.Text, line.Start, firstSpace - 1);
 
-            // Skip any spaces after info string
-            firstSpace++;
-            while (firstSpace <= line.End)
-            {
-                char c = line[firstSpace];
-                if (c.IsSpaceOrTab())
-                {
-                    firstSpace++;
-                }
-                else
-                {
-                    break;
-                }
-            }
-
             var argStringSlice = new StringSlice(line.Text, firstSpace, line.End);
             argStringSlice.Trim();
-            argString = argStringSlice.ToString();
+            fenced.Arguments = s_argumentsStringCache.Get(argStringSlice.AsSpan());
         }
         else
         {
             infoStringSlice = line;
+            fenced.Arguments = string.Empty;
         }
 
         infoStringSlice.Trim();
 
-        fenced.Info = _infoStringCache.Get(infoStringSlice.AsSpan());
-        fenced.Arguments = HtmlHelper.Unescape(argString);
+        fenced.Info = s_infoStringCache.Get(infoStringSlice.AsSpan());
 
         return true;
     }
@@ -303,17 +280,19 @@ public override BlockState TryOpen(BlockProcessor processor)
         // Try to parse any attached attributes
         TryParseAttributes?.Invoke(processor, ref line, fenced);
 
-        // If the info parser was not successfull, early exit
+        // If the info parser was not successful, early exit
         if (InfoParser != null && !InfoParser(processor, ref line, fenced, matchChar))
         {
             return BlockState.None;
         }
 
         // Add the language as an attribute by default
-        if (!string.IsNullOrEmpty(fenced.Info))
+        string? info = fenced.Info;
+
+        if (!string.IsNullOrEmpty(info))
         {
             Debug.Assert(_infoPrefixCache is not null || InfoPrefix is null);
-            string infoWithPrefix = _infoPrefixCache?.Get(fenced.Info!) ?? fenced.Info!;
+            string infoWithPrefix = _infoPrefixCache?.Get(info!) ?? info!;
             fenced.GetAttributes().AddClass(infoWithPrefix);
         }
 
@@ -329,34 +308,32 @@ public override BlockState TryOpen(BlockProcessor processor)
     public override BlockState TryContinue(BlockProcessor processor, Block block)
     {
         var fence = (IFencedBlock)block;
-        var openingCount = fence.OpeningFencedCharCount;
 
         // Match if we have a closing fence
         var line = processor.Line;
         var sourcePosition = processor.Start;
         var closingCount = line.CountAndSkipChar(fence.FencedChar);
-        var diff = openingCount - closingCount;
 
         char c = line.CurrentChar;
-        var lastFenceCharPosition = processor.Start + closingCount;
 
         // If we have a closing fence, close it and discard the current line
         // The line must contain only fence opening character followed only by whitespaces.
         var startBeforeTrim = line.Start;
-        var endBeforeTrim = line.End;
-        var trimmed = line.TrimEnd();
-        if (diff <= 0 && !processor.IsCodeIndent && (c == '\0' || c.IsWhitespace()) && trimmed)
+
+        if (fence.OpeningFencedCharCount <= closingCount &&
+            !processor.IsCodeIndent &&
+            (c == '\0' || c.IsWhitespace()) &&
+            line.TrimEnd())
         {
             block.UpdateSpanEnd(startBeforeTrim - 1);
 
-            var fencedBlock = (IFencedBlock)block;
-            fencedBlock.ClosingFencedCharCount = closingCount;
+            fence.ClosingFencedCharCount = closingCount;
 
             if (processor.TrackTrivia)
             {
-                fencedBlock.NewLine = processor.Line.NewLine;
-                fencedBlock.TriviaBeforeClosingFence = processor.UseTrivia(sourcePosition - 1);
-                fencedBlock.TriviaAfter = new StringSlice(processor.Line.Text, lastFenceCharPosition, endBeforeTrim);
+                fence.NewLine = line.NewLine;
+                fence.TriviaBeforeClosingFence = processor.UseTrivia(sourcePosition - 1);
+                fence.TriviaAfter = new StringSlice(line.Text, processor.Start + closingCount, processor.Line.End);
             }
 
             // Don't keep the last line

src/Markdig/Parsers/MarkdownParser.cs

@@ -53,34 +53,11 @@ public static MarkdownDocument Parse(string text, MarkdownPipeline? pipeline = n
         {
             blockProcessor.Open(document);
 
-            ProcessBlocks(blockProcessor, new LineReader(text));
+            ProcessBlocks(blockProcessor, text);
 
             if (pipeline.TrackTrivia)
             {
-                Block? lastBlock = blockProcessor.LastBlock;
-                if (lastBlock is null && document.Count == 0)
-                {
-                    // this means we have unassigned characters
-                    var noBlocksFoundBlock = new EmptyBlock(null);
-                    List<StringSlice> linesBefore = blockProcessor.UseLinesBefore();
-                    noBlocksFoundBlock.LinesAfter = new List<StringSlice>();
-                    if (linesBefore != null)
-                    {
-                        noBlocksFoundBlock.LinesAfter.AddRange(linesBefore);
-                    }
-
-                    document.Add(noBlocksFoundBlock);
-                }
-                else if (lastBlock != null && blockProcessor.LinesBefore != null)
-                {
-                    // this means we're out of lines, but still have unassigned empty lines.
-                    // thus, we'll assign the empty unsassigned lines to the last block
-                    // of the document.
-                    var rootMostContainerBlock = Block.FindRootMostContainerParent(lastBlock);
-                    rootMostContainerBlock.LinesAfter ??= new List<StringSlice>();
-                    var linesBefore = blockProcessor.UseLinesBefore();
-                    rootMostContainerBlock.LinesAfter.AddRange(linesBefore);
-                }
+                ProcessBlocksTrivia(blockProcessor, document);
             }
 
             // At this point the LineIndex is the same as the number of lines in the document
@@ -117,12 +94,15 @@ private static string FixupZero(string text)
         return text.Replace('\0', CharHelper.ReplacementChar);
     }
 
-    private static void ProcessBlocks(BlockProcessor blockProcessor, LineReader lineReader)
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static void ProcessBlocks(BlockProcessor blockProcessor, string text)
     {
+        var lineReader = new LineReader(text);
+
         while (true)
         {
-            // Get the precise position of the begining of the line
-            var lineText = lineReader.ReadLine();
+            // Get the precise position of the beginning of the line
+            StringSlice lineText = lineReader.ReadLine();
 
             // If this is the end of file and the last line is empty
             if (lineText.Text is null)
@@ -132,9 +112,39 @@ private static void ProcessBlocks(BlockProcessor blockProcessor, LineReader line
 
             blockProcessor.ProcessLine(lineText);
         }
+
         blockProcessor.CloseAll(true);
     }
 
+    private static void ProcessBlocksTrivia(BlockProcessor blockProcessor, MarkdownDocument document)
+    {
+        Block? lastBlock = blockProcessor.LastBlock;
+        if (lastBlock is null && document.Count == 0)
+        {
+            // this means we have unassigned characters
+            var noBlocksFoundBlock = new EmptyBlock(null);
+            List<StringSlice> linesBefore = blockProcessor.UseLinesBefore();
+            noBlocksFoundBlock.LinesAfter = [];
+            if (linesBefore != null)
+            {
+                noBlocksFoundBlock.LinesAfter.AddRange(linesBefore);
+            }
+
+            document.Add(noBlocksFoundBlock);
+        }
+        else if (lastBlock != null && blockProcessor.LinesBefore != null)
+        {
+            // this means we're out of lines, but still have unassigned empty lines.
+            // thus, we'll assign the empty unsassigned lines to the last block
+            // of the document.
+            var rootMostContainerBlock = Block.FindRootMostContainerParent(lastBlock);
+            rootMostContainerBlock.LinesAfter ??= [];
+            var linesBefore = blockProcessor.UseLinesBefore();
+            rootMostContainerBlock.LinesAfter.AddRange(linesBefore);
+        }
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
     private static void ProcessInlines(InlineProcessor inlineProcessor, MarkdownDocument document)
     {
         // "stackless" processor

src/Markdig/Polyfills/IndexOfHelpers.cs

@@ -41,6 +41,13 @@ public static int IndexOfAny(this ReadOnlySpan<char> span, string values)
         return -1;
     }
 #endif
+
+#if !NET6_0_OR_GREATER
+    public static bool Contains<T>(this ReadOnlySpan<T> span, T value) where T : IEquatable<T>
+    {
+        return span.IndexOf(value) >= 0;
+    }
+#endif
 }
 
 #endif