Merge branch 'develop'

ied206 · ied206 · commit a38694e878f6 · 2025-02-16T23:41:27.000+09:00
diff --git a/Joveler.Compression.LZ4/CHANGELOG.md b/Joveler.Compression.LZ4/CHANGELOG.md
@@ -2,6 +2,10 @@
 
 ## v5.x
 
+### v5.0.2
+
+- Update NuGet README.
+
 ### v5.0.1
 
 Released on 2025-02-10
@@ -15,6 +19,7 @@ Released on 2025-01-31
 - (EXPERIMENTAL) Provides parallel lz4 compression.
 - (BREAKING CHANGE) Now targets .NET 8.0/.NET Framework 4.6.2/.NET Standard 2.0, to use TPL Dataflow library.
 - Upgrades packaged lz4 binaries to 1.10.0.
+- Adds support for nullable reference type information.
 
 ## v4.x
 
diff --git a/Joveler.Compression.LZ4/Joveler.Compression.LZ4.csproj b/Joveler.Compression.LZ4/Joveler.Compression.LZ4.csproj
@@ -10,7 +10,7 @@
     
     <PackageId>Joveler.Compression.LZ4</PackageId>
     <Title>Joveler.Compression.LZ4</Title>
-    <Version>5.0.1</Version>
+    <Version>5.0.2</Version>
     <Authors>Hajin Jang</Authors>
     <Company>Joveler</Company>
     <Description>lz4 pinvoke library for .NET.
@@ -21,7 +21,7 @@ Supports Windows, Linux and macOS.</Description>
     <RepositoryUrl>https://github.com/ied206/Joveler.Compression</RepositoryUrl>
     <PackageIcon>images\Logo.png</PackageIcon>
     <PackageProjectUrl>https://github.com/ied206/Joveler.Compression</PackageProjectUrl>
-    <PackageReleaseNotes>- Fix broken .NET Framework MSBuild script for copying native DLLs.</PackageReleaseNotes>
+    <PackageReleaseNotes>- Update NuGet README.</PackageReleaseNotes>
     <PackageTags>lz4 lz4-hc lz4hc lz4-frame compress compression decompress decompression archive native pinvoke interop</PackageTags>
     <PackageReadmeFile>README.md</PackageReadmeFile>
   </PropertyGroup>
diff --git a/Joveler.Compression.LZ4/NUGET_README.md b/Joveler.Compression.LZ4/NUGET_README.md
@@ -10,17 +10,19 @@ Joveler.Compression.LZ4 can be installed via [nuget](https://www.nuget.org/packa
 
 ## Performance
 
-**WARNING**: The library supports many customizable options, but performance is a bit slow due to pinvoke overhead. See [README.md](https://github.com/ied206/Joveler.Compression/blob/master/Joveler.Compression.LZ4/README.md) for details.
+**WARNING**: Due to LZ4's performant nature, P/Invoke overhead has a more negative effect on LZ4 than on conventional compression algorithms.
+
+The overhead becomes trivial enough if you handle big files and focus on multithreaded high-level compression.
+
+In decompression, pure managed implementation is much faster than this native wrapper.
+
+See project README for more details.
 
 ## Features
 
 - LZ4FrameStream, the stream for [lz4 frame format](https://github.com/lz4/lz4/blob/master/doc/lz4_Frame_format.md).
 - (EXPERIMENTAL) Parallel compression support on LZ4FrameStream.
 
-## Tested liblz4 versions
-
-- 1.9.4 (Included)
-
 ## Support
 
 ### Targeted .NET platforms
@@ -45,3 +47,7 @@ Joveler.Compression.LZ4 can be installed via [nuget](https://www.nuget.org/packa
 ### Supported LZ4 versions
 
 - 1.10.0 (Included)
+
+## Usage
+
+Please refer to the project homepage.
diff --git a/Joveler.Compression.LZ4/README.md b/Joveler.Compression.LZ4/README.md
@@ -10,9 +10,15 @@ Joveler.Compression.LZ4 can be installed via [nuget](https://www.nuget.org/packa
 
 ## Performance
 
-Please keep in mind that `Joveler.Compression.LZ4` have been uploaded to the nuget because it supports many customizable options, not best in performance.
+Due to LZ4's performant nature, P/Invoke overhead has a more negative effect on LZ4 than on conventional compression algorithms.
 
-Decompression of `Joveler.Compression.LZ4` is similar or slightly faster than the pure C# implementation, [K4os.Compression.LZ4](https://github.com/MiloszKrajewski/K4os.Compression.LZ4). In the meanwhile the compression is about 50% faster to 2000% slower. Please be careful about performance when using it on production.
+If you mostly handle big files and focus on multithreaded high-level compression, `Joveler.Compression.LZ4` is the right choice.
+
+If you mostly handle small files and are focused on performant single-threaded use, consider using [K4os.Compression.LZ4](https://github.com/MiloszKjewski/K4os.Compression.LZ4), the pure C# implementation. Its performance generally ties in with the native pinvoke method. 
+
+Suppose your use case involves mostly decompression, use [K4os.Compression.LZ4](https://github.com/MiloszKjewski/K4os.Compression.LZ4) as it is consistently faster.
+
+Read the [Benchmark](#Benchmark) section for more details.
 
 ## Features
 
@@ -56,7 +62,7 @@ See [CHANGELOG.md](./CHANGELOG.md).
 
 `Joveler.Compression.LZ4` is licensed under [BSD 2-Clause license](./LICENSE).
 
-## Performance
+## Benchmark
 
 ### Compression
 
@@ -90,3 +96,26 @@ In multithread compression, performance of `Joveler.Compression.LZ4` scales line
 | lz4    | reymont.pdf       | Fastest | 13,985 μs  |
 | lz4-T2 | reymont.pdf       | Fastest | 10,868 μs  |
 | K4os   | reymont.pdf       | Fastest | 15,343 μs  |
+
+### Decompression
+
+[K4os.Compression.LZ4](https://github.com/MiloszKrajewski/K4os.Compression.LZ4) is about twice faster than `Joveler.Compression.LZ4`.
+
+| Method | SrcFileName       | Level   | Mean       |
+|--------|-------------------|---------|------------|
+| lz4    | bible_en_utf8.txt  | Default | 4,598.377 μs |
+| K4os   | bible_en_utf8.txt  | Default | 2,728.262 μs |
+| lz4    | bible_en_utf8.txt  | Fastest | 4,966.349 μs |
+| K4os   | bible_en_utf8.txt  | Fastest | 2,807.782 μs |
+| lz4    | bible_kr_utf8.txt  | Default | 5,673.352 μs |
+| K4os   | bible_kr_utf8.txt  | Default | 3,845.430 μs |
+| lz4    | bible_kr_utf8.txt  | Fastest | 6,114.985 μs |
+| K4os   | bible_kr_utf8.txt  | Fastest | 3,753.987 μs |
+| lz4    | ooffice.dll       | Default | 7,170.132 μs |
+| K4os   | ooffice.dll       | Default | 4,635.963 μs | 
+| lz4    | ooffice.dll       | Fastest | 7,587.325 μs | 
+| K4os   | ooffice.dll       | Fastest | 4,743.326 μs | 
+| lz4    | reymont.pdf       | Default | 6,546.353 μs | 
+| K4os   | reymont.pdf       | Default | 4,529.416 μs | 
+| lz4    | reymont.pdf       | Fastest | 7,243.303 μs | 
+| K4os   | reymont.pdf       | Fastest | 4,760.949 μs | 
diff --git a/Joveler.Compression.XZ/CHANGELOG.md b/Joveler.Compression.XZ/CHANGELOG.md
@@ -20,6 +20,7 @@ Released on 2025-01-31
 
 - (BREAKING CHANGE) Now targets .NET 8.0/.NET Framework 4.6.2/.NET Standard 2.0.
 - Upgrades packaged xz4 binaries to 5.6.4.
+- Adds support for nullable reference type information.
 
 ## v4.x
 
diff --git a/Joveler.Compression.XZ/NUGET_README.md b/Joveler.Compression.XZ/NUGET_README.md
@@ -5,12 +5,9 @@
 ## Features
 
 - XZStream, the stream for [.xz file format](https://tukaani.org/xz/xz-file-format.txt).
+- Parallel compression and decompression of XZStream.
 - Fast native implementation of CRC32 and CRC64 checksums.
 
-## Usage
-
-Please refer the [project homepage](https://github.com/ied206/Joveler.Compression/blob/master/Joveler.Compression.XZ/USAGE.md).
-
 ## Support
 
 ### Targeted .NET platforms
@@ -36,3 +33,6 @@ Please refer the [project homepage](https://github.com/ied206/Joveler.Compressio
 
 - 5.6.4 (Included)
 
+## Usage
+
+Please refer to the project homepage.
diff --git a/Joveler.Compression.XZ/README.md b/Joveler.Compression.XZ/README.md
@@ -11,6 +11,7 @@ Joveler.Compression.XZ can be installed via [nuget](https://www.nuget.org/packag
 ## Features
 
 - XZStream, the stream for [.xz file format](https://tukaani.org/xz/xz-file-format.txt).
+- Parallel compression and decompression of XZStream.
 - Fast native implementation of CRC32 and CRC64 checksums.
 
 ## Support
diff --git a/Joveler.Compression.ZLib/CHANGELOG.md b/Joveler.Compression.ZLib/CHANGELOG.md
@@ -16,6 +16,7 @@ Released on 2025-01-31
 - (BREAKING CHANGE) Now targets .NET 8.0/.NET Framework 4.6.2/.NET Standard 2.0, to use TPL Dataflow library.
 - Provides APIs for combining two checksums.
 - Upgrades packaged zlib-ng binaries to 2.2.3.
+- Adds support for nullable reference type information.
 
 ## v5.x
 
