Merge pull request #3048 from SixLabors/js/png-iptc

Add proper IptcProfile support in PNG

Author: JimBobSquarePants

src/ImageSharp/Formats/Png/PngConstants.cs

@@ -62,6 +62,21 @@ internal static class PngConstants
     /// </summary>
     public const int MinTextKeywordLength = 1;
 
+    /// <summary>
+    /// Specifies the keyword used to identify the Exif raw profile in image metadata.
+    /// </summary>
+    public const string ExifRawProfileKeyword = "Raw profile type exif";
+
+    /// <summary>
+    /// Specifies the profile keyword used to identify raw IPTC metadata within image files.
+    /// </summary>
+    public const string IptcRawProfileKeyword = "Raw profile type iptc";
+
+    /// <summary>
+    /// The IPTC resource id in Photoshop IRB. 0x0404 (big endian).
+    /// </summary>
+    public const ushort AdobeIptcResourceId = 0x0404;
+
     /// <summary>
     /// Gets the header bytes identifying a Png.
     /// </summary>
@@ -100,4 +115,31 @@ internal static class PngConstants
         (byte)'m',
         (byte)'p'
     ];
+
+    /// <summary>
+    /// Gets the ASCII bytes for the "Photoshop 3.0" identifier used in some PNG metadata payloads.
+    /// This value is null-terminated.
+    /// </summary>
+    public static ReadOnlySpan<byte> AdobePhotoshop30 =>
+    [
+        (byte)'P',
+        (byte)'h',
+        (byte)'o',
+        (byte)'t',
+        (byte)'o',
+        (byte)'s',
+        (byte)'h',
+        (byte)'o',
+        (byte)'p',
+        (byte)' ',
+        (byte)'3',
+        (byte)'.',
+        (byte)'0',
+        0
+    ];
+
+    /// <summary>
+    /// Gets the ASCII bytes for the "8BIM" signature used in Photoshop resources.
+    /// </summary>
+    public static ReadOnlySpan<byte> EightBim => [(byte)'8', (byte)'B', (byte)'I', (byte)'M'];
 }

src/ImageSharp/Formats/Png/PngDecoderCore.cs

@@ -21,6 +21,7 @@
 using SixLabors.ImageSharp.Metadata.Profiles.Cicp;
 using SixLabors.ImageSharp.Metadata.Profiles.Exif;
 using SixLabors.ImageSharp.Metadata.Profiles.Icc;
+using SixLabors.ImageSharp.Metadata.Profiles.Iptc;
 using SixLabors.ImageSharp.Metadata.Profiles.Xmp;
 using SixLabors.ImageSharp.PixelFormats;
 
@@ -1440,14 +1441,19 @@ private void ReadCompressedTextChunk(ImageMetadata baseMetadata, PngMetadata met
     /// object unmodified.</returns>
     private static bool TryReadTextChunkMetadata(ImageMetadata baseMetadata, string chunkName, string chunkText)
     {
-        if (chunkName.Equals("Raw profile type exif", StringComparison.OrdinalIgnoreCase) &&
+        if (chunkName.Equals(PngConstants.ExifRawProfileKeyword, StringComparison.OrdinalIgnoreCase) &&
             TryReadLegacyExifTextChunk(baseMetadata, chunkText))
         {
             // Successfully parsed legacy exif data from text
             return true;
         }
 
-        // TODO: "Raw profile type iptc", potentially others?
+        if (chunkName.Equals(PngConstants.IptcRawProfileKeyword, StringComparison.OrdinalIgnoreCase) &&
+            TryReadLegacyIptcTextChunk(baseMetadata, chunkText))
+        {
+            // Successfully parsed legacy iptc data from text
+            return true;
+        }
 
         // No special chunk data identified
         return false;
@@ -1571,6 +1577,214 @@ private static bool TryReadLegacyExifTextChunk(ImageMetadata metadata, string da
         return true;
     }
 
+    /// <summary>
+    /// Reads iptc data encoded into a text chunk with the name "Raw profile type iptc".
+    /// This convention is used by ImageMagick/exiftool/exiv2/digiKam and stores a byte-count
+    /// followed by hex-encoded bytes.
+    /// </summary>
+    /// <param name="metadata">The <see cref="ImageMetadata"/> to store the decoded iptc tags into.</param>
+    /// <param name="data">The contents of the "Raw profile type iptc" text chunk.</param>
+    private static bool TryReadLegacyIptcTextChunk(ImageMetadata metadata, string data)
+    {
+        // Preserve first IPTC found.
+        if (metadata.IptcProfile != null)
+        {
+            return true;
+        }
+
+        ReadOnlySpan<char> dataSpan = data.AsSpan().TrimStart();
+
+        // Must start with the "iptc" identifier (case-insensitive).
+        // Common real-world format (ImageMagick/ExifTool) is:
+        // "IPTC profile\n      <len>\n<hex...>"
+        if (dataSpan.Length < 4 || !StringEqualsInsensitive(dataSpan[..4], "iptc".AsSpan()))
+        {
+            return false;
+        }
+
+        // Skip the remainder of the first line ("IPTC profile", etc).
+        int firstLineEnd = dataSpan.IndexOf('\n');
+        if (firstLineEnd < 0)
+        {
+            return false;
+        }
+
+        dataSpan = dataSpan[(firstLineEnd + 1)..].TrimStart();
+
+        // Next line contains the decimal byte length (often indented).
+        int dataLengthEnd = dataSpan.IndexOf('\n');
+        if (dataLengthEnd < 0)
+        {
+            return false;
+        }
+
+        int dataLength;
+        try
+        {
+            dataLength = ParseInt32(dataSpan[..dataLengthEnd]);
+        }
+        catch
+        {
+            return false;
+        }
+
+        if (dataLength <= 0)
+        {
+            return false;
+        }
+
+        // Skip to the hex-encoded data.
+        dataSpan = dataSpan[(dataLengthEnd + 1)..].Trim();
+
+        byte[] iptcBlob = new byte[dataLength];
+
+        try
+        {
+            int written = 0;
+
+            for (; written < dataLength;)
+            {
+                ReadOnlySpan<char> lineSpan = dataSpan;
+
+                int newlineIndex = dataSpan.IndexOf('\n');
+                if (newlineIndex != -1)
+                {
+                    lineSpan = dataSpan[..newlineIndex];
+                }
+
+                // Important: handle CRLF and any incidental whitespace.
+                lineSpan = lineSpan.Trim(); // removes ' ', '\t', '\r', '\n', etc.
+
+                if (!lineSpan.IsEmpty)
+                {
+                    written += HexConverter.HexStringToBytes(lineSpan, iptcBlob.AsSpan()[written..]);
+                }
+
+                if (newlineIndex == -1)
+                {
+                    break;
+                }
+
+                dataSpan = dataSpan[(newlineIndex + 1)..];
+            }
+
+            if (written != dataLength)
+            {
+                return false;
+            }
+        }
+        catch
+        {
+            return false;
+        }
+
+        // Prefer IRB extraction if this is Photoshop-style data (8BIM resource blocks).
+        byte[] iptcPayload = TryExtractIptcFromPhotoshopIrb(iptcBlob, out byte[] extracted)
+            ? extracted
+            : iptcBlob;
+
+        metadata.IptcProfile = new IptcProfile(iptcPayload);
+        return true;
+    }
+
+    /// <summary>
+    /// Attempts to extract IPTC metadata from a Photoshop Image Resource Block (IRB) contained within the specified
+    /// data buffer.
+    /// </summary>
+    /// <remarks>This method scans the provided data for a Photoshop IRB block containing IPTC metadata and
+    /// extracts it if present. The method does not validate the contents of the IPTC data beyond locating the
+    /// appropriate resource block.</remarks>
+    /// <param name="data">A read-only span of bytes containing the Photoshop IRB data to search for embedded IPTC metadata.</param>
+    /// <param name="iptcBytes">When this method returns, contains the extracted IPTC metadata as a byte array if found; otherwise, an undefined
+    /// value.</param>
+    /// <returns><see langword="true"/> if IPTC metadata is successfully extracted from the IRB data; otherwise, <see langword="false"/>.</returns>
+    private static bool TryExtractIptcFromPhotoshopIrb(ReadOnlySpan<byte> data, out byte[] iptcBytes)
+    {
+        iptcBytes = default!;
+
+        ReadOnlySpan<byte> adobePhotoshop30 = PngConstants.AdobePhotoshop30;
+
+        // Some writers include the "Photoshop 3.0\0" header, some store just IRB blocks.
+        if (data.Length >= adobePhotoshop30.Length && data[..adobePhotoshop30.Length].SequenceEqual(adobePhotoshop30))
+        {
+            data = data[adobePhotoshop30.Length..];
+        }
+
+        ReadOnlySpan<byte> eightBim = PngConstants.EightBim;
+        ushort adobeIptcResourceId = PngConstants.AdobeIptcResourceId;
+        while (data.Length >= 12)
+        {
+            if (!data[..4].SequenceEqual(eightBim))
+            {
+                return false;
+            }
+
+            data = data[4..];
+
+            // Resource ID (2 bytes, big endian)
+            if (data.Length < 2)
+            {
+                return false;
+            }
+
+            ushort resourceId = (ushort)((data[0] << 8) | data[1]);
+            data = data[2..];
+
+            // Pascal string name (1-byte length, then bytes), padded to even.
+            if (data.Length < 1)
+            {
+                return false;
+            }
+
+            int nameLen = data[0];
+            int nameFieldLen = 1 + nameLen;
+            if ((nameFieldLen & 1) != 0)
+            {
+                nameFieldLen++; // pad to even
+            }
+
+            if (data.Length < nameFieldLen + 4)
+            {
+                return false;
+            }
+
+            data = data[nameFieldLen..];
+
+            // Resource data size (4 bytes, big endian)
+            int size = (data[0] << 24) | (data[1] << 16) | (data[2] << 8) | data[3];
+            data = data[4..];
+
+            if (size < 0 || data.Length < size)
+            {
+                return false;
+            }
+
+            ReadOnlySpan<byte> payload = data[..size];
+
+            // Data is padded to even.
+            int advance = size;
+            if ((advance & 1) != 0)
+            {
+                advance++;
+            }
+
+            if (resourceId == adobeIptcResourceId)
+            {
+                iptcBytes = payload.ToArray();
+                return true;
+            }
+
+            if (data.Length < advance)
+            {
+                return false;
+            }
+
+            data = data[advance..];
+        }
+
+        return false;
+    }
+
     /// <summary>
     /// Reads the color profile chunk. The data is stored similar to the zTXt chunk.
     /// </summary>