perf(mtconnect): sample chunks parse from bytes, no per-chunk string copies

src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect/MTConnectAgentClient.cs

@@ -231,7 +231,7 @@ public sealed class MTConnectAgentClient : IMTConnectAgentClient, IDisposable
         var delivered = 0L;
         while (true)
         {
-            var part = await reader.ReadNextPartAsync(live).ConfigureAwait(false);
+            var part = await reader.ReadNextPartBytesAsync(live).ConfigureAwait(false);
             if (part is null)
             {
                 // The caller cancelling is the ONE contracted way out; anything else is reported.
@@ -244,7 +244,7 @@ public sealed class MTConnectAgentClient : IMTConnectAgentClient, IDisposable
             // Agent's real keep-alive is a well-formed observation-free MTConnectStreams document,
             // and that one IS yielded — it advances the sequence, so swallowing it would strand the
             // caller's `from` behind the Agent's buffer.
-            if (string.IsNullOrWhiteSpace(part))
+            if (IsFramingOnly(part))
             {
                 continue;
             }
@@ -562,6 +562,24 @@ public sealed class MTConnectAgentClient : IMTConnectAgentClient, IDisposable
         return parsed.GetLeftPart(UriPartial.Path).TrimEnd('/');
     }
 
+    /// <summary>
+    ///     Whether a multipart part carries nothing but transport filler — the byte-level equivalent
+    ///     of the <c>IsNullOrWhiteSpace</c> check the string-shaped framer used to run, and the same
+    ///     ASCII set the streams parser trims off a chunk before parsing it.
+    /// </summary>
+    private static bool IsFramingOnly(ReadOnlySpan<byte> part)
+    {
+        foreach (var value in part)
+        {
+            if (value is not ((byte)' ' or (byte)'\t' or (byte)'\r' or (byte)'\n'))
+            {
+                return false;
+            }
+        }
+
+        return true;
+    }
+
     /// <summary>
     ///     Frames a <c>multipart/x-mixed-replace</c> body into individual part payloads, reading
     ///     incrementally and never buffering more than the part currently in flight.
@@ -610,10 +628,16 @@ public sealed class MTConnectAgentClient : IMTConnectAgentClient, IDisposable
         public MTConnectStreamEndReason EndReason { get; private set; } = MTConnectStreamEndReason.ConnectionClosed;
 
         /// <summary>
-        ///     Reads the next part's payload, or <c>null</c> once the closing delimiter is seen or
-        ///     the Agent closes the connection — see <see cref="EndReason"/> for which.
+        ///     Reads the next part's payload as its raw bytes, or <c>null</c> once the closing
+        ///     delimiter is seen or the Agent closes the connection — see <see cref="EndReason"/>
+        ///     for which.
         /// </summary>
-        public async Task<string?> ReadNextPartAsync(CancellationToken ct)
+        /// <remarks>
+        ///     Deliberately <b>not</b> decoded to a string: the one consumer is the streams parser,
+        ///     which reads UTF-8 off a stream itself. Decoding here would materialise every chunk of
+        ///     a never-ending stream twice for a document that is parsed once and dropped.
+        /// </remarks>
+        public async Task<byte[]?> ReadNextPartBytesAsync(CancellationToken ct)
         {
             if (_finished)
             {
@@ -652,7 +676,7 @@ public sealed class MTConnectAgentClient : IMTConnectAgentClient, IDisposable
                 body = await ReadToNextDelimiterAsync(ct).ConfigureAwait(false);
             }
 
-            return body is null ? null : Encoding.UTF8.GetString(body);
+            return body;
         }
 
         private void WarnOnceAboutMissingContentLength()
@@ -669,7 +693,7 @@ public sealed class MTConnectAgentClient : IMTConnectAgentClient, IDisposable
                 uri);
         }
 
-        private string? Finish(MTConnectStreamEndReason reason)
+        private byte[]? Finish(MTConnectStreamEndReason reason)
         {
             _finished = true;
             EndReason = reason;

src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect/MTConnectStreamsParser.cs

@@ -1,4 +1,5 @@
 using System.Globalization;
+using System.Runtime.InteropServices;
 using System.Xml;
 using System.Xml.Linq;
 
@@ -121,6 +122,33 @@ internal static class MTConnectStreamsParser
         return Build(document);
     }
 
+    /// <summary>
+    ///     Parses one <c>/sample</c> chunk straight from the bytes the multipart framer produced.
+    /// </summary>
+    /// <remarks>
+    ///     The <c>/sample</c> hot path used to materialise every chunk twice — once as a decoded
+    ///     string, once again as the trimmed copy handed to the parser — for a document that is read
+    ///     once and thrown away. Parsing from the bytes drops both copies; <see cref="XDocument"/>
+    ///     does its own UTF-8 decoding off the stream. The framing tolerance the string overload's
+    ///     <c>Trim</c> provided is preserved by <see cref="TrimFraming"/>, which slices rather than
+    ///     copies.
+    /// </remarks>
+    /// <param name="xml">The chunk's raw bytes; a UTF-8 BOM and surrounding framing whitespace are tolerated.</param>
+    /// <exception cref="InvalidDataException">As for <see cref="Parse(string)"/>.</exception>
+    public static MTConnectStreamsResult Parse(ReadOnlyMemory<byte> xml)
+    {
+        var trimmed = TrimFraming(xml);
+        if (trimmed.IsEmpty)
+        {
+            throw new InvalidDataException(
+                "MTConnect streams response was empty; expected an MTConnectStreams XML document.");
+        }
+
+        using var stream = AsStream(trimmed);
+
+        return Parse(stream);
+    }
+
     /// <summary>Parses a <c>/current</c> response read from a stream.</summary>
     /// <param name="stream">A stream positioned at the start of the XML document.</param>
     /// <exception cref="InvalidDataException">As for <see cref="Parse(string)"/>.</exception>
@@ -141,6 +169,55 @@ internal static class MTConnectStreamsParser
         return Build(document);
     }
 
+    /// <summary>
+    ///     Slices off what the string overload's <c>Trim('', ' ', '\t', '\r', '\n')</c> removed:
+    ///     a UTF-8 byte-order mark and the multipart framing's surrounding line breaks, either of
+    ///     which <see cref="XmlReader"/> rejects ahead of the XML declaration.
+    /// </summary>
+    private static ReadOnlyMemory<byte> TrimFraming(ReadOnlyMemory<byte> xml)
+    {
+        var span = xml.Span;
+        var start = 0;
+        var end = span.Length;
+
+        while (start < end)
+        {
+            if (end - start >= 3 && span[start] == 0xEF && span[start + 1] == 0xBB && span[start + 2] == 0xBF)
+            {
+                start += 3;
+
+                continue;
+            }
+
+            if (!IsFramingWhitespace(span[start]))
+            {
+                break;
+            }
+
+            start++;
+        }
+
+        while (end > start && IsFramingWhitespace(span[end - 1]))
+        {
+            end--;
+        }
+
+        return xml[start..end];
+    }
+
+    private static bool IsFramingWhitespace(byte value) =>
+        value is (byte)' ' or (byte)'\t' or (byte)'\r' or (byte)'\n';
+
+    /// <summary>
+    ///     Views the (already trimmed) bytes as a stream without copying them where the memory is
+    ///     array-backed — which it always is on the <c>/sample</c> path, whose framer hands over the
+    ///     part's own <c>byte[]</c>.
+    /// </summary>
+    private static MemoryStream AsStream(ReadOnlyMemory<byte> xml) =>
+        MemoryMarshal.TryGetArray(xml, out ArraySegment<byte> segment) && segment.Array is not null
+            ? new MemoryStream(segment.Array, segment.Offset, segment.Count, writable: false)
+            : new MemoryStream(xml.ToArray(), writable: false);
+
     private static MTConnectStreamsResult Build(XDocument document)
     {
         var root = document.Root

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Tests/MTConnectStreamsParseTests.cs

@@ -294,6 +294,113 @@ public sealed class MTConnectStreamsParseTests
         fromStream.Observations.ShouldBe(fromString.Observations);
     }
 
+    // ---------------------------------------------------- the byte overload the /sample pump feeds on
+
+    [Theory]
+    [InlineData(CurrentFixture)]
+    [InlineData(SampleFixture)]
+    [InlineData(GapFixture)]
+    [InlineData(UnavailableFixture)]
+    public void Parse_from_bytes_matches_the_string_overload_on_every_fixture(string fixture)
+    {
+        // The /sample pump hands the parser the chunk's raw bytes; a chunk parsed from bytes must be
+        // indistinguishable from the same chunk round-tripped through a string.
+        var fromBytes = MTConnectStreamsParser.Parse(File.ReadAllBytes(fixture));
+
+        var fromString = MTConnectStreamsParser.Parse(File.ReadAllText(fixture));
+        fromBytes.InstanceId.ShouldBe(fromString.InstanceId);
+        fromBytes.FirstSequence.ShouldBe(fromString.FirstSequence);
+        fromBytes.NextSequence.ShouldBe(fromString.NextSequence);
+        fromBytes.Observations.ShouldBe(fromString.Observations);
+    }
+
+    [Theory]
+    [InlineData("\r\n")]
+    [InlineData("\n")]
+    [InlineData(" \t")]
+    [InlineData("\r\n\r\n")]
+    public void Parse_from_bytes_tolerates_framing_whitespace_AHEAD_of_the_xml_declaration(string residue)
+    {
+        // THE load-bearing leg of the byte path's trim, and the one that is not falsifiable by
+        // accident: every fixture opens with an <?xml ...?> declaration, and XmlReader rejects even
+        // one space ahead of one ("no whitespace characters are allowed to appear before it"). The
+        // string overload trimmed it; a byte path that fed the frame straight to XDocument.Load
+        // would turn an Agent that writes a blank line after the part headers into a hard parse
+        // failure on every chunk. (Trailing whitespace and a leading BOM, by contrast, XmlReader
+        // handles natively — the cases below pin that rather than the trim.)
+        var bytes = Encoding.UTF8.GetBytes(residue).Concat(File.ReadAllBytes(SampleFixture)).ToArray();
+
+        var withResidue = MTConnectStreamsParser.Parse(bytes);
+
+        withResidue.NextSequence.ShouldBe(113);
+        withResidue.Observations.ShouldBe(MTConnectStreamsParser.Parse(File.ReadAllText(SampleFixture)).Observations);
+    }
+
+    [Theory]
+    [InlineData(CurrentFixture)]
+    [InlineData(SampleFixture)]
+    public void Parse_from_bytes_tolerates_a_utf8_byte_order_mark_ahead_of_the_declaration(string fixture)
+    {
+        // A chunk lifted out of a multipart frame can carry a BOM; the string path trimmed ''
+        // for exactly this reason, so the byte path must handle the EF BB BF that decodes to it.
+        var bytes = new byte[] { 0xEF, 0xBB, 0xBF }.Concat(File.ReadAllBytes(fixture)).ToArray();
+
+        var withBom = MTConnectStreamsParser.Parse(bytes);
+
+        withBom.Observations.ShouldBe(MTConnectStreamsParser.Parse(File.ReadAllText(fixture)).Observations);
+    }
+
+    [Theory]
+    [InlineData("\r\n")]
+    [InlineData("\r\n\r\n")]
+    [InlineData("\n")]
+    [InlineData(" \t\r\n")]
+    public void Parse_from_bytes_tolerates_the_multipart_frames_trailing_line_break(string residue)
+    {
+        // The other half of what the string path's Trim covered: the framing's trailing line break.
+        var bytes = File.ReadAllBytes(SampleFixture).Concat(Encoding.UTF8.GetBytes(residue)).ToArray();
+
+        var withResidue = MTConnectStreamsParser.Parse(bytes);
+
+        withResidue.NextSequence.ShouldBe(113);
+        withResidue.Observations.ShouldBe(MTConnectStreamsParser.Parse(File.ReadAllText(SampleFixture)).Observations);
+    }
+
+    [Fact]
+    public void Parse_from_bytes_tolerates_a_bom_behind_leading_whitespace()
+    {
+        // Whitespace THEN a BOM defeats XmlReader's encoding sniffing outright ("Data at the root
+        // level is invalid"), so the trim has to interleave the two rather than strip a BOM only at
+        // offset 0 — which is exactly what the string overload's Trim('', ' ', …) set did.
+        var bytes = " \r\n"u8.ToArray()
+            .Concat(new byte[] { 0xEF, 0xBB, 0xBF })
+            .Concat(File.ReadAllBytes(CurrentFixture))
+            .Concat("\r\n"u8.ToArray())
+            .ToArray();
+
+        MTConnectStreamsParser.Parse(bytes).Observations
+            .ShouldBe(MTConnectStreamsParser.Parse(File.ReadAllText(CurrentFixture)).Observations);
+    }
+
+    [Theory]
+    [InlineData(new byte[0])]
+    [InlineData(new byte[] { (byte)' ', (byte)'\r', (byte)'\n' })]
+    [InlineData(new byte[] { 0xEF, 0xBB, 0xBF })]
+    public void Parse_from_bytes_throws_on_an_empty_payload(byte[] bytes)
+    {
+        // Same failure family as the string path: never a silently empty result.
+        var ex = Should.Throw<InvalidDataException>(() => MTConnectStreamsParser.Parse(bytes));
+
+        ex.Message.ShouldContain("empty");
+    }
+
+    [Fact]
+    public void Parse_from_bytes_throws_on_malformed_xml()
+    {
+        Should.Throw<InvalidDataException>(
+            () => MTConnectStreamsParser.Parse("not xml at all"u8.ToArray()));
+    }
+
     // ------------------------------------------------------------- fails loudly, never silently empty
 
     [Theory]