Expose stream, seek, and span archive input modes

Compression.Core/README.md

@@ -204,7 +204,7 @@ Use the concrete version you intend to consume; this document does not predict a
 
 <!-- API:BEGIN generated by Hawkynt/RepositoryTemplate/package-readme — edit the XML docs in source, not here -->
 
-Every public and protected member of all 681 types, generated from the built assembly and its XML documentation, is in [REFERENCE.md](https://github.com/Hawkynt/CompressionWorkbench/blob/main/Compression.Core/REFERENCE.md).
+Every public and protected member of all 682 types, generated from the built assembly and its XML documentation, is in [REFERENCE.md](https://github.com/Hawkynt/CompressionWorkbench/blob/main/Compression.Core/REFERENCE.md).
 
 <!-- API:END -->
 

Compression.Registry/ArchiveFormatOperationsExtensions.cs

@@ -0,0 +1,96 @@
+namespace Compression.Registry;
+
+/// <summary>
+/// Makes the default archive input-mode members available when a descriptor is
+/// referenced by its concrete type. Default interface members are otherwise only
+/// in the member set of an <see cref="IArchiveFormatOperations"/> reference.
+/// </summary>
+public static class ArchiveFormatOperationsExtensions {
+  /// <inheritdoc cref="IArchiveFormatOperations.ListStreaming"/>
+  public static List<ArchiveEntryInfo> ListStreaming(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string? password)
+    => operations.ListStreaming(archive, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ExtractStreaming"/>
+  public static void ExtractStreaming(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string outputDir,
+    string? password,
+    string[]? files)
+    => operations.ExtractStreaming(archive, outputDir, password, files);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ListSeekable"/>
+  public static List<ArchiveEntryInfo> ListSeekable(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string? password)
+    => operations.ListSeekable(archive, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ExtractSeekable"/>
+  public static void ExtractSeekable(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string outputDir,
+    string? password,
+    string[]? files)
+    => operations.ExtractSeekable(archive, outputDir, password, files);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ListSpan"/>
+  public static List<ArchiveEntryInfo> ListSpan(
+    this IArchiveFormatOperations operations,
+    ReadOnlySpan<byte> archive,
+    string? password)
+    => operations.ListSpan(archive, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ExtractSpan"/>
+  public static void ExtractSpan(
+    this IArchiveFormatOperations operations,
+    ReadOnlySpan<byte> archive,
+    string outputDir,
+    string? password,
+    string[]? files)
+    => operations.ExtractSpan(archive, outputDir, password, files);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.OpenEntryStreaming"/>
+  public static Stream OpenEntryStreaming(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string entryName,
+    string? password)
+    => operations.OpenEntryStreaming(archive, entryName, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.OpenEntrySeekable"/>
+  public static Stream OpenEntrySeekable(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string entryName,
+    string? password)
+    => operations.OpenEntrySeekable(archive, entryName, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.OpenEntrySpan"/>
+  public static Stream OpenEntrySpan(
+    this IArchiveFormatOperations operations,
+    ReadOnlySpan<byte> archive,
+    string entryName,
+    string? password)
+    => operations.OpenEntrySpan(archive, entryName, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ExtractEntryToMemoryStreaming"/>
+  public static byte[] ExtractEntryToMemoryStreaming(
+    this IArchiveFormatOperations operations,
+    Stream archive,
+    string entryName,
+    string? password)
+    => operations.ExtractEntryToMemoryStreaming(archive, entryName, password);
+
+  /// <inheritdoc cref="IArchiveFormatOperations.ExtractEntryToMemorySpan"/>
+  public static byte[] ExtractEntryToMemorySpan(
+    this IArchiveFormatOperations operations,
+    ReadOnlySpan<byte> archive,
+    string entryName,
+    string? password)
+    => operations.ExtractEntryToMemorySpan(archive, entryName, password);
+}

Compression.Registry/IArchiveFormatOperations.cs

@@ -8,13 +8,94 @@ namespace Compression.Registry;
 /// defragment, shrink, input constraints) are separate opt-in interfaces so callers can
 /// discover them at the type level.
 /// </summary>
+/// <remarks>
+/// <para>
+/// The historical <see cref="List(Stream,string?)"/> / <c>Extract(Stream, ...)</c>
+/// methods remain the descriptor's native compatibility surface. The explicit input-mode methods
+/// below make the three archive-reading models available uniformly to every archive and
+/// pseudo-archive: forward-only stream input, required-seek input, and in-memory
+/// <see cref="ReadOnlySpan{T}"/> input.
+/// </para>
+/// <para>
+/// Existing descriptors automatically gain all three modes. A descriptor can override the
+/// default interface implementations when it has a genuinely streaming parser, a native
+/// random-access reader, or a zero-copy span parser. Until then, forward-only streams are
+/// spooled to a temporary seekable file and spans are copied once into an owned memory stream;
+/// neither fallback imposes a whole-archive managed-array limit on stream input.
+/// </para>
+/// </remarks>
 public interface IArchiveFormatOperations {
-  /// <summary>List all entries in the archive.</summary>
+  /// <summary>List all entries in the archive using the descriptor's native stream path.</summary>
   List<ArchiveEntryInfo> List(Stream stream, string? password);
 
-  /// <summary>Extract entries from the archive to an output directory.</summary>
+  /// <summary>Extract entries from the archive to an output directory using the descriptor's native stream path.</summary>
   void Extract(Stream stream, string outputDir, string? password, string[]? files);
 
+  /// <summary>
+  /// Lists entries from a forward-only or seekable stream. This is the libarchive-style
+  /// streaming entry point: callers do not need to provide seek capability.
+  /// </summary>
+  public virtual List<ArchiveEntryInfo> ListStreaming(Stream archive, string? password) {
+    using var lease = SeekableArchiveInputLease.Open(archive);
+    return this.ListSeekable(lease.Stream, password);
+  }
+
+  /// <summary>
+  /// Extracts entries from a forward-only or seekable stream. Descriptors with a native
+  /// one-pass parser should override this method; the default spools only when necessary.
+  /// </summary>
+  public virtual void ExtractStreaming(Stream archive, string outputDir, string? password, string[]? files) {
+    using var lease = SeekableArchiveInputLease.Open(archive);
+    this.ExtractSeekable(lease.Stream, outputDir, password, files);
+  }
+
+  /// <summary>
+  /// Lists entries through the explicit seek-based path. The supplied stream must support
+  /// seeking; it is rewound before the descriptor's native reader is invoked.
+  /// </summary>
+  public virtual List<ArchiveEntryInfo> ListSeekable(Stream archive, string? password) {
+    ArgumentNullException.ThrowIfNull(archive);
+    if (!archive.CanRead)
+      throw new ArgumentException("Archive input must be readable.", nameof(archive));
+    if (!archive.CanSeek)
+      throw new ArgumentException("Seek-based archive input must support seeking.", nameof(archive));
+    archive.Position = 0;
+    return this.List(archive, password);
+  }
+
+  /// <summary>
+  /// Extracts entries through the explicit seek-based path. The supplied stream must support
+  /// seeking; it is rewound before the descriptor's native reader is invoked.
+  /// </summary>
+  public virtual void ExtractSeekable(Stream archive, string outputDir, string? password, string[]? files) {
+    ArgumentNullException.ThrowIfNull(archive);
+    if (!archive.CanRead)
+      throw new ArgumentException("Archive input must be readable.", nameof(archive));
+    if (!archive.CanSeek)
+      throw new ArgumentException("Seek-based archive input must support seeking.", nameof(archive));
+    archive.Position = 0;
+    this.Extract(archive, outputDir, password, files);
+  }
+
+  /// <summary>
+  /// Lists entries from an in-memory archive image. The default compatibility bridge copies
+  /// the span once because a <see cref="Stream"/> cannot safely retain a borrowed span; native
+  /// span parsers should override this method to remain allocation-free.
+  /// </summary>
+  public virtual List<ArchiveEntryInfo> ListSpan(ReadOnlySpan<byte> archive, string? password) {
+    using var stream = new MemoryStream(archive.ToArray(), writable: false);
+    return this.ListSeekable(stream, password);
+  }
+
+  /// <summary>
+  /// Extracts entries from an in-memory archive image. Native span parsers can override this
+  /// method to avoid the compatibility copy used by the default implementation.
+  /// </summary>
+  public virtual void ExtractSpan(ReadOnlySpan<byte> archive, string outputDir, string? password, string[]? files) {
+    using var stream = new MemoryStream(archive.ToArray(), writable: false);
+    this.ExtractSeekable(stream, outputDir, password, files);
+  }
+
   /// <summary>
   /// Opens a single entry as a read-only <see cref="Stream"/> bounded to that
   /// entry's logical bytes — physically incapable of reading slack space,
@@ -29,7 +110,7 @@ public interface IArchiveFormatOperations {
   /// </para>
   /// <para>
   /// The default implementation intentionally does <b>not</b> materialize a
-  /// <c>byte[]</c>. It asks <see cref="Extract"/> for the selected entry in an
+  /// <c>byte[]</c>. It asks <c>Extract(Stream, ...)</c> for the selected entry in an
   /// isolated temporary directory, opens the resulting file as a seekable
   /// stream, and deletes that tree on dispose. This gives every descriptor a
   /// large-file-safe streaming fallback even before it grows a native per-entry
@@ -44,13 +125,61 @@ public virtual Stream OpenEntry(Stream archive, string entryName, string? passwo
     return new BoundedEntryStream(extracted, extracted.Length, leaveOpen: false);
   }
 
+  /// <summary>
+  /// Opens one entry from a forward-only or seekable archive source. A temporary spool, when
+  /// required, stays alive until the returned entry stream is disposed.
+  /// </summary>
+  public virtual Stream OpenEntryStreaming(Stream archive, string entryName, string? password) {
+    ArgumentException.ThrowIfNullOrWhiteSpace(entryName);
+    var lease = SeekableArchiveInputLease.Open(archive);
+    try {
+      var entry = this.OpenEntrySeekable(lease.Stream, entryName, password);
+      return new OwnedArchiveEntryStream(entry, lease);
+    } catch {
+      lease.Dispose();
+      throw;
+    }
+  }
+
+  /// <summary>
+  /// Opens one entry through the explicit seek-based path. The archive is rewound before the
+  /// descriptor-specific entry reader is invoked.
+  /// </summary>
+  public virtual Stream OpenEntrySeekable(Stream archive, string entryName, string? password) {
+    ArgumentNullException.ThrowIfNull(archive);
+    ArgumentException.ThrowIfNullOrWhiteSpace(entryName);
+    if (!archive.CanRead)
+      throw new ArgumentException("Archive input must be readable.", nameof(archive));
+    if (!archive.CanSeek)
+      throw new ArgumentException("Seek-based archive input must support seeking.", nameof(archive));
+    archive.Position = 0;
+    return this.OpenEntry(archive, entryName, password);
+  }
+
+  /// <summary>
+  /// Opens one entry from an in-memory archive image. Because the returned stream may outlive
+  /// this call, the default bridge owns one copy of the supplied span until that stream is disposed.
+  /// Native span readers can override this method when they can return independently owned output.
+  /// </summary>
+  public virtual Stream OpenEntrySpan(ReadOnlySpan<byte> archive, string entryName, string? password) {
+    ArgumentException.ThrowIfNullOrWhiteSpace(entryName);
+    var source = new MemoryStream(archive.ToArray(), writable: false);
+    try {
+      var entry = this.OpenEntrySeekable(source, entryName, password);
+      return new OwnedArchiveEntryStream(entry, source);
+    } catch {
+      source.Dispose();
+      throw;
+    }
+  }
+
   /// <summary>
   /// Extracts a single entry to a byte array. This is the explicitly buffered
   /// convenience API; callers working with large entries should use
-  /// <see cref="OpenEntry"/> instead.
+  /// <see cref="OpenEntry(Stream,string,string?)"/> instead.
   /// </summary>
   /// <remarks>
-  /// The default routes through <see cref="OpenEntry"/>, so descriptor-specific
+  /// The default routes through <see cref="OpenEntry(Stream,string,string?)"/>, so descriptor-specific
   /// isolation/decoding semantics are preserved. A result past the runtime array
   /// limit naturally fails here rather than imposing that limit on the streaming
   /// API or filesystem-driver layer.
@@ -64,4 +193,20 @@ public virtual byte[] ExtractEntryToMemory(Stream archive, string entryName, str
     entry.CopyTo(memory);
     return memory.ToArray();
   }
+
+  /// <summary>Extracts one entry from a forward-only or seekable archive source into memory.</summary>
+  public virtual byte[] ExtractEntryToMemoryStreaming(Stream archive, string entryName, string? password) {
+    using var entry = this.OpenEntryStreaming(archive, entryName, password);
+    using var memory = new MemoryStream();
+    entry.CopyTo(memory);
+    return memory.ToArray();
+  }
+
+  /// <summary>Extracts one entry from an in-memory archive image into a new byte array.</summary>
+  public virtual byte[] ExtractEntryToMemorySpan(ReadOnlySpan<byte> archive, string entryName, string? password) {
+    using var entry = this.OpenEntrySpan(archive, entryName, password);
+    using var memory = new MemoryStream();
+    entry.CopyTo(memory);
+    return memory.ToArray();
+  }
 }