Started changes to Zig 0.16.0

Mainly adding functions to File

.github/workflows/release-build.yml

@@ -13,19 +13,19 @@ jobs:
         uses: actions/checkout@v6
       - uses: mlugg/setup-zig@v2
       - name: Install deps
-        run: sudo apt update && sudo apt install -y zlib1g-dev libzstd-dev liblzma-dev liblz4-dev liblzo2-dev
+        run: sudo apt update && sudo apt install -y liblzma-dev liblzo2-dev
       - name: Build normal version
-        run: zig build -Drelease=true -Dversion=${{ github.ref_name }}
-      - name: Move normal build out
-        run: mv zig-out/bin/unsquashfs ./
+        run: zig build --release=fast -Duse_zig_decomp=true -Dversion=${{ github.ref_name }}
+      - name: Move zig build out
+        run: mv zig-out/bin/unsquashfs ./unsquashfs-x86_64-zig-libs
       - name: Rebuild with C libraries
-        run: zig build -Drelease=true -Duse_c_libs=true -Dversion="${{ github.ref_name }}"
+        run: zig build --release=fast -Dversion="${{ github.ref_name }}"
       - name: Move C build out
-        run: mv zig-out/bin/unsquashfs ./unsquashfs-c-libs
+        run: mv zig-out/bin/unsquashfs ./unsquashfs-x86_64-c-libs
       - name: Release
         uses: softprops/action-gh-release@v2
         with:
           prerelease: true
           files: |
-            unsquashfs
-            unsquashfs-c-libs
+            unsquashfs-x86_64-zig-libs
+            unsquashfs-x86_64-c-libs

.gitignore

@@ -2,3 +2,4 @@ testing/
 
 .zig-cache/
 zig-out/
+zig-pkg/

.zed/debug.json

@@ -0,0 +1,25 @@
+// Project-local debug tasks
+//
+// For more documentation on how to configure debug tasks,
+// see: https://zed.dev/docs/debugger
+[
+	{
+		"label": "Build & Run",
+
+		"adapter": "CodeLLDB",
+		"request": "launch",
+
+		"build": {
+			"command": "zig",
+			"args": ["build", "-Duse_c_libs=true", "-Ddebug=true"],
+		},
+
+		"program": "zig-out/bin/unsquashfs",
+		"args": [
+			"--force",
+			"-d",
+			"testing/TestExtractUnsquashfs",
+			"testing/LinuxPATest.sfs",
+		],
+	},
+]

README.md

@@ -6,19 +6,23 @@ A library and application to decompress or view squashfs archives.
 
 ## Current State
 
-Overall works, but currently is missing some features (see below). Extraction is a bit slow compared to the normal `unsquashfs` (from my _very_ basic testing it's about ~3x slower). Only properly work on Linux, any other OSes probably won't work fully and are untested.
+Overall works, but currently is missing some features ([see below](#capabilities)) and has significantly slow performance compared to `unsquashfs` ([see below](#performance)).
 
 ## Build options
 
-> `-Duse_c_libs`
+> `-Duse_c_libs=true`
 
 Instead of using Zig's standard library for decompression, use the system's C libraries. Has the benefit of being much faster and enabling LZO and LZ4 decompression.
 
-> `-Dallow_lzo`
+> `-Dallow_lzo=true`
 
 Enable compiling with LZO decompression support. The LZO library currently has some issues with Zig when imported so it's easier to just disable it by default. Only has an effect when using `-Duse_c_libs=true`.
 
-> `-Dversion`
+> `-Ddebug=true`
+
+Sets various build options that make debugging easier. Specifically, debug optimization is forced, valgrind support is enabled, error tracing is enabled, stipping is disabled, and copmilation uses LLVM (this is due to some linking issues when on Debug optimization and is required for debugging tools such as `lldb`. In the future this may be removed from the debug flag).
+
+> `-Dversion=0.0.0`
 
 Sets the version of `unsquashfs` shown when `--version` is passed.
 
@@ -26,11 +30,29 @@ Sets the version of `unsquashfs` shown when `--version` is passed.
 
 Most features are present except for the following:
 
-* mod_time is not set on extraction
-* xattrs are not applied on extraction
 * When using Zig decompression libraries then lzo and lz4 compression types are unavailable. I don't _currently_ plan on spending the time to find and validate a library since neither is popular.
+* When using C decompression libraries, lzo is not supported by default due to [some issues](#build-considerations). If it's needed it's trivial to fix, but it's easiest to just leave it disabled.
 
-## Building considerations
+## Performance
+
+This is some basic observation's I've made about this library's performance when compared to `unsquashfs`. Unless otherwise stated, most observations were made when extracting my test archive (which is fairly small and uses zstd compression) and with `--release=fast`.
+
+Currently, my only performance checks are checking execution time, nothing deeper.
+
+* Under ideal circumstances, my library is ~70% slower (.12s vs .20s).
+* Using Zig decompression libraries *significantly* increases decompression time by ~600%. Under ideal circumstances.
+* Performance improvements/regressions will be common. I'm still learning Zig.
+
+Example Times:
+
+* *unsquashfs, multi-threaded*: .12s
+* *unsquashfs, single-threaded*: .13s
+* *C-libs, single-threaded*: .45s
+* *C-libs, multi-threaded*: .20s
+* *Zig-libs, single-threaded*: 5.78s
+* *Zig-libs, multi-threaded*: 1.08s
+
+## Build considerations
 
 Compilation without `use_c_libs` works completely fine, but Zig has issues with some symbols from the lzo library that needs to be manually fixed. In particular you need to fix the definitions for `lzo_bytep` and `lzo_voidp` to be `*u8` and `?*anyopaque` respectively. Due to this, you have to manually enable LZO decompression using `-Dallow_lzo=true` when building.
 

build.zig

@@ -1,30 +1,61 @@
 const std = @import("std");
 
 pub fn build(b: *std.Build) !void {
-    const use_c_libs_option = b.option(bool, "use_c_libs", "Use C versions of decompression libraries instead of the Zig standard library ones");
-    const allow_lzo = b.option(bool, "allow_lzo", "Compile with lzo support");
+    const use_zig_decomp = b.option(bool, "use_zig_decomp", "Use zig standard library for decompression.") orelse true;
+    // const allow_lzo = b.option(bool, "allow_lzo", "Compile with lzo support") orelse false;
+    const debug = b.option(bool, "debug", "Enable options to make debugging easier.") orelse false;
     const version_string_option = b.option([]const u8, "version", "Version of the library/binary");
 
     const zig_squashfs_options = b.addOptions();
-    zig_squashfs_options.addOption(bool, "use_c_libs", use_c_libs_option orelse false);
-    zig_squashfs_options.addOption(bool, "allow_lzo", allow_lzo orelse false);
+    zig_squashfs_options.addOption(bool, "use_zig_decomp", use_zig_decomp);
+    // zig_squashfs_options.addOption(bool, "allow_lzo", allow_lzo);
 
     const target = b.standardTargetOptions(.{});
-    const optimize = b.standardOptimizeOption(.{ .preferred_optimize_mode = .ReleaseFast });
+    const optimize = b.standardOptimizeOption(.{});
     const mod = b.addModule("zig_squashfs", .{
         .root_source_file = b.path("src/root.zig"),
         .target = target,
-        .optimize = optimize,
-        .link_libc = if (use_c_libs_option == true) true else false,
+        .optimize = if (debug == true) .Debug else optimize,
+        .valgrind = debug,
+        .error_tracing = debug,
+        .strip = if (debug == true) false else null,
     });
     mod.addOptions("config", zig_squashfs_options);
-    if (use_c_libs_option == true) {
-        mod.linkSystemLibrary("zlib", .{});
-        mod.linkSystemLibrary("lzma", .{});
-        if (allow_lzo == true)
-            mod.linkSystemLibrary("minilzo", .{});
-        mod.linkSystemLibrary("lz4", .{});
-        mod.linkSystemLibrary("zstd", .{});
+    if (!use_zig_decomp) {
+        mod.link_libc = true;
+
+        const c_imports = b.addTranslateC(.{
+            .optimize = optimize,
+            .target = target,
+            .root_source_file = b.path("src/imports.c"),
+        });
+        mod.addImport("c", c_imports.createModule());
+
+        var zlib_ng = b.dependency("zlib_ng", .{
+            .target = target,
+            .optimize = optimize,
+        });
+        mod.linkLibrary(zlib_ng.artifact("zng"));
+
+        mod.linkSystemLibrary("lzma", .{ .preferred_link_mode = .static });
+
+        var minilzo = b.dependency("minilzo", .{
+            .target = target,
+            .optimize = optimize,
+        });
+        mod.linkLibrary(minilzo.artifact("minilzo"));
+
+        var lz4 = b.dependency("lz4", .{
+            .target = target,
+            .optimize = optimize,
+        });
+        mod.linkLibrary(lz4.artifact("lz4"));
+
+        var zstd = b.dependency("zstd", .{
+            .target = target,
+            .optimize = optimize,
+        });
+        mod.linkLibrary(zstd.artifact("zstd"));
     }
 
     var version = version_string_option orelse "0.0.0-testing";
@@ -39,41 +70,52 @@ pub fn build(b: *std.Build) !void {
     var exe_mod = b.createModule(.{
         .root_source_file = b.path("src/bin/unsquashfs.zig"),
         .target = target,
-        .optimize = optimize,
-        .link_libc = if (use_c_libs_option == true) true else false,
+        .optimize = if (debug == true) .Debug else optimize,
+        .link_libc = !use_zig_decomp,
         .imports = &.{
             .{ .name = "zig_squashfs", .module = mod },
         },
+        .valgrind = debug,
+        .error_tracing = debug,
+        .strip = if (debug == true) false else null,
     });
     exe_mod.addOptions("config", unsquashfs_options);
     const exe = b.addExecutable(.{
         .name = "unsquashfs",
         .root_module = exe_mod,
+        .use_llvm = debug,
     });
 
     const lib = b.addLibrary(.{
         .name = "squashfs",
         .root_module = mod,
+        .use_llvm = debug,
     });
 
     b.installArtifact(lib);
     b.installArtifact(exe);
-    const run_step = b.step("run", "Run the app");
-    const run_cmd = b.addRunArtifact(exe);
-    run_step.dependOn(&run_cmd.step);
-    run_cmd.step.dependOn(b.getInstallStep());
-    if (b.args) |args| {
-        run_cmd.addArgs(args);
-    }
+
     const mod_tests = b.addTest(.{
         .root_module = mod,
+        .test_runner = .{
+            .mode = .simple,
+            .path = b.path("src/test.zig"),
+        },
     });
     const run_mod_tests = b.addRunArtifact(mod_tests);
-    const exe_tests = b.addTest(.{
-        .root_module = exe.root_module,
-    });
-    const run_exe_tests = b.addRunArtifact(exe_tests);
     const test_step = b.step("test", "Run tests");
     test_step.dependOn(&run_mod_tests.step);
-    test_step.dependOn(&run_exe_tests.step);
+
+    // zls build check steps
+    const lib_check = b.addLibrary(.{
+        .name = "squashfs",
+        .root_module = mod,
+    });
+    const exe_check = b.addExecutable(.{
+        .name = "unsquashfs",
+        .root_module = exe_mod,
+    });
+    const check = b.step("check", "Check if unsquashfs compiles");
+    check.dependOn(&exe_check.step);
+    check.dependOn(&lib_check.step);
 }

build.zig.zon

@@ -1,43 +1,31 @@
 .{
     .name = .squashfs,
-    .version = "0.0.1",
+    .version = "0.0.6",
     .fingerprint = 0x37ba29474b87f145, // Changing this has security and trust implications.
-    .minimum_zig_version = "0.15.2",
-    // This field is optional.
-    // Each dependency must either provide a `url` and `hash`, or a `path`.
-    // `zig build --fetch` can be used to fetch all dependencies of a package, recursively.
-    // Once all dependencies are fetched, `zig build` no longer requires
-    // internet connectivity.
+    .minimum_zig_version = "0.16.0",
     .dependencies = .{
-        // See `zig fetch --save <url>` for a command-line interface for adding dependencies.
-        //.example = .{
-        //    // When updating this field to a new URL, be sure to delete the corresponding
-        //    // `hash`, otherwise you are communicating that you expect to find the old hash at
-        //    // the new URL. If the contents of a URL change this will result in a hash mismatch
-        //    // which will prevent zig from using it.
-        //    .url = "https://example.com/foo.tar.gz",
-        //
-        //    // This is computed from the file contents of the directory of files that is
-        //    // obtained after fetching `url` and applying the inclusion rules given by
-        //    // `paths`.
-        //    //
-        //    // This field is the source of truth; packages do not come from a `url`; they
-        //    // come from a `hash`. `url` is just one of many possible mirrors for how to
-        //    // obtain a package matching this `hash`.
-        //    //
-        //    // Uses the [multihash](https://multiformats.io/multihash/) format.
-        //    .hash = "...",
-        //
-        //    // When this is provided, the package is found in a directory relative to the
-        //    // build root. In this case the package's hash is irrelevant and therefore not
-        //    // computed. This field and `url` are mutually exclusive.
-        //    .path = "foo",
-        //
-        //    // When this is set to `true`, a package is declared to be lazily
-        //    // fetched. This makes the dependency only get fetched if it is
-        //    // actually used.
-        //    .lazy = false,
-        //},
+        .zlib_ng = .{
+            // .url = "git+https://github.com/CalebQ42/zig-zlib-ng#5f2f02dfb28acca2517dacbbd09e9b987f57b133",
+            // .hash = "zlib_ng-2.3.3-pre1-2HYS4ClFAABW8KlHMyBHtlNKE3V7kCS8wqfxawG7xeaa",
+            .path = "../zig-zlib-ng",
+        },
+        .zstd = .{
+            .url = "git+https://github.com/allyourcodebase/zstd.git?ref=1.5.7-1#e1a501be57f42c541e8a5597e4b59a074dfd09a3",
+            .hash = "zstd-1.5.7-1-KEItkAMwAAD6OKY3m0OOmXG7aL-aLUfrDqbP5J5oYapU",
+        },
+        .lz4 = .{
+            .url = "git+https://github.com/allyourcodebase/lz4.git?ref=1.10.0-6#41f52ab227caf9d48cf88c89a4d2946caa12b102",
+            .hash = "lz4-1.10.0-6-ewyzw-4NAAAWDpY4xpiqr4LQhZQAC0x_rGnW2iPh6jk2",
+        },
+        .minilzo = .{
+            // .url = "git+https://github.com/CalebQ42/zig-minilzo.git#7cbae997b91a44d74b7cd6c073584dc9562a6c90",
+            // .hash = "minilzo-2.10.0-Ij7BO8wLAADeWI4Pe4jp8XTDsDaquZR14oZ7_9yKKDWP",
+            .path = "../zig-minilzo",
+        },
+        // .fastlzma2 = .{
+        //     .url = "git+https://github.com/allyourcodebase/fast-lzma2#d7615e0c957a62fcd6691b3fe9519a091885bfa2",
+        //     .hash = "fastlzma2-0.0.0-gNWHgVeLAAD0Tlak3xhNcgpPSYcjyJppq0tlGmPKCC_V",
+        // },
     },
     .paths = .{
         "build.zig",

run_tests.sh

@@ -1,10 +0,0 @@
-#!/bin/sh
-
-zig test \
-	-lc \
-	-lz \
-	-llzma \
-	-lminilzo \
-	-llz4 \
-	-lzstd \
-		src/test.zig

src/archive.zig

@@ -1,169 +1,184 @@
-//! A squashfs archive read from a file.
-//! Can be used to directly access File's contents or extract to the filesystem.
-
 const std = @import("std");
-const File = std.fs.File;
-const builtin = @import("builtin");
+const Io = std.Io;
 
-const Decomp = @import("decomp.zig");
+const DecompTypes = @import("decomp/types.zig");
+const Decompressor = @import("decomp.zig");
 const ExtractionOptions = @import("options.zig");
+const File = @import("file.zig");
 const Inode = @import("inode.zig");
-const InodeRef = Inode.Ref;
-const BlockSize = @import("inode_data/file.zig").BlockSize;
-const SfsFile = @import("file.zig");
-const Superblock = @import("super.zig").Superblock;
-const Table = @import("table.zig").Table;
+const BlockSize = @import("inode/file.zig").BlockSize;
+const LookupTable = @import("lookup_table.zig");
 const MetadataReader = @import("util/metadata.zig");
 const OffsetFile = @import("util/offset_file.zig");
+const Utils = @import("util/utils.zig");
 
-const config = if (builtin.is_test) .{
-    .use_c_libs = true,
-    .allow_lzo = false,
-} else @import("config");
-
-/// Information about a fragment section. Multiple fragments are contained in the block described by a single FragEntry.
-/// The offset into the block and fragment size is stored in the file's inode.
-pub const FragEntry = packed struct {
-    start: u64,
-    size: BlockSize,
-    _: u32,
+pub const Error = error{
+    BadMagic,
+    BadBlockLog,
+    BadVersion,
+    BadCheck,
 };
 
 const Archive = @This();
 
-// 4 Gigs
-const DEFAULT_MEM_SIZE = 4 * 1024 * 1024 * 1024;
-
-parent_alloc: std.mem.Allocator,
-alloc: std.heap.ThreadSafeAllocator,
-// alloc: std.heap.FixedBufferAllocator,
-// fixed_buf: []u8,
-thread_count: usize,
-
-fil: OffsetFile,
+file: OffsetFile,
 
 super: Superblock,
 
-setup: bool = false,
+stateless_decomp: Decompressor,
 
-decomp: Decomp.DecompFn,
-
-frag_table: Table(FragEntry) = undefined,
-id_table: Table(u16) = undefined,
-export_table: Table(InodeRef) = undefined,
-
-/// Default settings using std.Thread.getCpuCount() threads and the minimum of 4gb or half of system memory for memory usage.
-pub fn init(alloc: std.mem.Allocator, fil: File) !Archive {
-    return initAdvanced(
-        alloc,
-        fil,
-        0,
-        try std.Thread.getCpuCount(),
-    );
-}
-/// Create the Archive dictating the amount of threads & memory used.
-/// If trying to extract a full archive, a large memory size & thread count could help.
-/// If you're planning on only interacting with a small number of files, it should be fine to use few threads and a small memory size.
-pub fn initAdvanced(alloc: std.mem.Allocator, fil: File, offset: u64, threads: usize) !Archive {
+/// Create an Archive from a File.
+pub fn init(io: Io, fil: Io.File, offset: u64) !Archive {
     var super: Superblock = undefined;
-    const red = try fil.pread(@ptrCast(&super), offset);
-    std.debug.assert(red == @sizeOf(Superblock));
+    var fil_rdr = fil.reader(io, &[0]u8{});
+    if (offset > 0)
+        try fil_rdr.seekTo(offset);
+    try fil_rdr.interface.readSliceEndian(Superblock, @ptrCast(&super), .little);
     try super.validate();
-    // const fixed_buf = try alloc.alloc(u8, mem);
-    return .{
-        .parent_alloc = alloc,
-        .alloc = .{ .child_allocator = alloc },
-        // .fixed_buf = fixed_buf,
-        .thread_count = threads,
-        .fil = .init(fil, offset),
-        .decomp = switch (super.compression) {
-            .gzip => Decomp.gzipDecompress,
-            .lzma => Decomp.lzmaDecompress,
-            .xz => Decomp.xzDecompress,
-            .zstd => Decomp.zstdDecompress,
-            .lz4 => if (config.use_c_libs) Decomp.cLz4 else return error.Lz4Unsupported,
-            .lzo => if (config.use_c_libs and config.allow_lzo) Decomp.lzoDecompress else return error.LzoUnsupported,
-        },
 
+    return .{
+        .file = .{ .fil = fil, .offset = offset },
         .super = super,
+        .stateless_decomp = .{ .vtable = &.{ .stateless = try DecompTypes.getStatelessFn(super.compression) } },
     };
 }
-pub fn deinit(self: *Archive) void {
-    // self.parent_alloc.free(self.fixed_buf);
-    if (self.setup) {
-        self.frag_table.deinit();
-        self.export_table.deinit();
-        self.id_table.deinit();
+
+pub fn root(self: Archive, alloc: std.mem.Allocator) !File {
+    return .{
+        .file = self.file,
+        .super = self.super.toMinimal(),
+        .decomp = self.stateless_decomp.statelessCopy(alloc),
+
+        .inode = try Utils.readInode(
+            alloc,
+            &self.stateless_decomp,
+            self.file,
+            self.super.inode_start,
+            self.super.block_size,
+            self.super.root_ref.block_start,
+            self.super.root_ref.block_offset,
+        ),
+        .name = "",
+    };
+}
+pub fn open(self: Archive, alloc: std.mem.Allocator, path: []const u8) !File {
+    if (Utils.pathIsSelf(path)) return self.root(alloc);
+    var root_file = self.root(alloc);
+    defer root_file.deinit();
+    return root_file.open(alloc, path);
+}
+
+pub fn fragEntry(self: Archive, idx: u32) !FragEntry {
+    return LookupTable.stateless(FragEntry, self.fil, &self.stateless_decomp, self.super.frag_start, idx);
+}
+pub fn id(self: Archive, idx: u32) !u16 {
+    return LookupTable.stateless(u16, self.fil, &self.stateless_decomp, self.super.id_start, idx);
+}
+pub fn inode(self: Archive, alloc: std.mem.Allocator, inode_num: u32) !Inode {
+    const ref = try LookupTable.stateless(Inode.Ref, self.file, &self.stateless_decomp, self.super.export_start, inode_num - 1);
+    return Utils.readInode(
+        alloc,
+        &self.stateless_decomp,
+        self.file,
+        self.super.inode_start,
+        self.super.block_size,
+        ref.block_start,
+        ref.block_offset,
+    );
+}
+
+pub fn extract(self: Archive, alloc: std.mem.Allocator, path: []const u8, options: ExtractionOptions) !void {
+    _ = self;
+    _ = alloc;
+    _ = path;
+    _ = options;
+    return error.TODO;
+}
+
+// Superblock
+
+const SQUASHFS_MAGIC: u32 = std.mem.readInt(u32, "hsqs", .little);
+
+pub const Superblock = packed struct {
+    magic: u32,
+    inode_count: u32,
+    mod_time: u32,
+    block_size: u32,
+    frag_count: u32,
+    compression: DecompTypes.Enum,
+    block_log: u16,
+    flags: packed struct {
+        inode_uncompressed: bool,
+        data_uncompressed: bool,
+        check: bool,
+        frag_uncompressed: bool,
+        fragment_never: bool,
+        fragment_always: bool,
+        duplicates: bool,
+        exportable: bool,
+        xattr_uncompressed: bool,
+        xattr_never: bool,
+        compression_options: bool,
+        ids_uncompressed: bool,
+        _: u4,
+    },
+    id_count: u16,
+    ver_maj: u16,
+    ver_min: u16,
+    root_ref: Inode.Ref,
+    size: u64,
+    id_start: u64,
+    xattr_start: u64,
+    inode_start: u64,
+    dir_start: u64,
+    frag_start: u64,
+    export_start: u64,
+
+    /// Validate the Superblock. If an error is returned, it's likely the archive is corrupted or not a squashfs archive.
+    fn validate(self: Superblock) !void {
+        if (self.magic != SQUASHFS_MAGIC)
+            return Error.BadMagic;
+        if (self.flags.check)
+            return Error.BadCheck;
+        if (self.ver_maj != 4 or self.ver_min != 0)
+            return Error.BadVersion;
+        if (std.math.log2(self.block_size) != self.block_log)
+            return Error.BadBlockLog;
     }
-}
 
-pub fn allocator(self: *Archive) std.mem.Allocator {
-    return self.alloc.allocator();
-}
-
-fn setupValues(self: *Archive) !void {
-    const alloc = self.allocator();
-    self.frag_table = try .init(alloc, self.fil, self.decomp, self.super.frag_start, self.super.frag_count);
-    self.id_table = try .init(alloc, self.fil, self.decomp, self.super.id_start, self.super.id_count);
-    self.export_table = try .init(alloc, self.fil, self.decomp, self.super.export_start, self.super.inode_count);
-    self.setup = true;
-}
-
-pub fn id(self: *Archive, idx: u32) !u16 {
-    if (!self.setup) try self.setupValues();
-    return self.id_table.get(idx);
-}
-
-pub fn frag(self: *Archive, idx: u32) !FragEntry {
-    if (!self.setup) try self.setupValues();
-    return self.frag_table.get(idx);
-}
-
-pub fn inode(self: *Archive, num: u32) !Inode {
-    if (!self.setup) try self.setupValues();
-    const ref = try self.export_table.get(num - 1);
-    var rdr = try self.fil.readerAt(ref.block_start + self.super.inode_start, &[0]u8{});
-    var meta: MetadataReader = .init(self.allocator(), &rdr.interface, &self.decomp);
-    try meta.interface.discardAll(ref.block_offset);
-    return try .read(self.allocator(), &meta.interface, self.super.block_size);
-}
-
-pub fn root(self: *Archive) !SfsFile {
-    if (!self.setup) try self.setupValues();
-    var rdr = try self.fil.readerAt(self.super.root_ref.block_start + self.super.inode_start, &[0]u8{});
-    var meta: MetadataReader = .init(self.allocator(), &rdr.interface, self.decomp);
-    try meta.interface.discardAll(self.super.root_ref.block_offset);
-    const in: Inode = try .read(self.allocator(), &meta.interface, self.super.block_size);
-    return .init(self, in, "");
-}
-
-pub fn open(self: *Archive, path: []const u8) !SfsFile {
-    if (!self.setup) try self.setupValues();
-    var root_fil = try self.root();
-    defer if (!SfsFile.pathIsSelf(path)) root_fil.deinit();
-    return root_fil.open(path);
-}
-
-pub fn extract(self: *Archive, path: []const u8, options: ExtractionOptions) !void {
-    if (!self.setup) try self.setupValues();
-    var alloc = self.allocator();
-    var ext_path: []u8 = undefined;
-    if (std.fs.cwd().statFile(path)) |stat| {
-        if (stat.kind == .directory) {
-            ext_path = @constCast(path);
-        } else return error.ExtractionPathExists;
-    } else |err| {
-        if (err == error.FileNotFound) {
-            ext_path = @constCast(path);
-        } else {
-            std.log.err("Error stat-ing extraction path {s}: {}\n", .{ path, err });
-            return err;
+    pub fn toMinimal(self: Superblock) MinimalSuperblock {
+        return .{
+            .inode_count = self.inode_count,
+            .block_size = self.block_size,
+            .frag_count = self.frag_count,
+            .id_count = self.id_count,
+            .id_start = self.id_start,
+            .xattr_start = self.xattr_start,
+            .inode_start = self.inode_start,
+            .dir_start = self.dir_start,
+            .frag_start = self.frag_start,
+            .export_start = self.export_start,
+        };
     }
-    }
-    defer if (ext_path.len > path.len) alloc.free(ext_path);
-    var rdr = try self.fil.readerAt(self.super.root_ref.block_start + self.super.inode_start, &[0]u8{});
-    var meta: MetadataReader = .init(self.allocator(), &rdr.interface, self.decomp);
-    try meta.interface.discardAll(self.super.root_ref.block_offset);
-    const in: Inode = try .read(self.allocator(), &meta.interface, self.super.block_size);
-    try in.extractToThreaded(self, ext_path, options, self.thread_count);
-}
+};
+
+pub const MinimalSuperblock = struct {
+    inode_count: u32,
+    block_size: u32,
+    frag_count: u32,
+    id_count: u16,
+    id_start: u64,
+    xattr_start: u64,
+    inode_start: u64,
+    dir_start: u64,
+    frag_start: u64,
+    export_start: u64,
+};
+
+// Frag Entry
+
+pub const FragEntry = packed struct {
+    block_start: u64,
+    size: BlockSize,
+    _: u32,
+};

src/bin/unsquashfs.zig

@@ -1,5 +1,6 @@
 const std = @import("std");
-const Writer = std.Io.Writer;
+const Io = std.Io;
+const Writer = Io.Writer;
 const builtin = @import("builtin");
 
 const config = @import("config");
@@ -14,8 +15,13 @@ const help_mgs =
     \\  -d <location>   Extract to the given location instead of "squashfs-root"
     \\
     \\  -o <offset>     Start reading the archive at the given offset.
+    \\  -dx             Don't set xattr values
+    \\  -dp             Don't set permissions (includes setting uid & gid owner)
     \\
-    \\  -p <threads>    Specify how many threads to use. If no present, the system's logical cores count is used.
+    \\  -p <threads>    Specify how many threads to use. If no present or zero, the system's logical cores count is used.
+    \\  -v              Verbose
+    \\
+    \\  --force         Force extraction. If the destination already exists, it will be deleted.
     \\
     \\  --help          Display this messages
     \\  --version       Display the version
@@ -28,32 +34,46 @@ var archive: []const u8 = "";
 var extLoc: []const u8 = "squashfs-root";
 var offset: u64 = 0;
 var threads: u32 = 0;
+var verbose: bool = false;
+var ignore_xattrs: bool = false;
+var ignore_permissions: bool = false;
+var force: bool = false;
 
-pub fn main() !void {
-    const alloc = std.heap.smp_allocator;
-    var stdout = std.fs.File.stdout();
-    var out = stdout.writer(&[0]u8{});
+pub fn main(init: std.process.Init) !void {
+    const alloc = init.gpa;
+    const io = init.io;
+
+    var stdout = std.Io.File.stdout();
+    var out = stdout.writer(io, &[0]u8{});
     defer out.interface.flush() catch {};
-    try handleArgs(alloc, &out.interface);
+    try handleArgs(init.minimal.args, &out.interface);
     if (archive.len == 0) {
         try out.interface.print("You must provide a squashfs archive\n", .{});
         try out.interface.print(help_mgs, .{});
         return;
     }
-    var fil: std.fs.File = try std.fs.cwd().openFile(archive, .{}); //TODO: Handle error gracefully.
-    defer fil.close();
-    var arc: squashfs.Archive = try .initAdvanced(alloc, fil, offset, threads); //TODO: Update when memory size matters. //TODO: Handle error gracefully.
-    defer arc.deinit();
-    try arc.extract(extLoc, .Default); //TODO: Handle error gracefully.
+    var fil: Io.File = try Io.Dir.cwd().openFile(io, archive, .{}); //TODO: Handle error gracefully.
+    defer fil.close(io);
+    var arc: squashfs.Archive = try .init(io, fil, offset); //TODO: Handle error gracefully.
+    const options: squashfs.ExtractionOptions = .{
+        .threads = if (threads == 0) try std.Thread.getCpuCount() else threads,
+        .verbose = verbose,
+        .verbose_writer = if (verbose) &out.interface else null,
+        .ignore_xattr = ignore_xattrs,
+        .ignore_permissions = ignore_permissions,
+    };
+    if (force)
+        try Io.Dir.cwd().deleteTree(io, extLoc);
+    try arc.extract(alloc, extLoc, options); //TODO: Handle error gracefully.
 }
 
-fn handleArgs(alloc: std.mem.Allocator, out: *Writer) !void {
-    var args = try std.process.argsWithAllocator(alloc);
-    defer args.deinit();
-    _ = args.next(); // args[0] is the application launch command.
-    while (args.next()) |arg| {
+fn handleArgs(args: std.process.Args, out: *Writer) !void {
+    var arg_iter = args.iterate();
+    defer arg_iter.deinit();
+    _ = arg_iter.next(); // args[0] is the application launch command.
+    while (arg_iter.next()) |arg| {
         if (std.mem.eql(u8, arg, "-o")) {
-            const nxt = args.next();
+            const nxt = arg_iter.next();
             if (nxt == null or nxt.?.len == 0) {
                 try out.print("-o must be followed by a number\n", .{});
                 return errors.InvalidArguments;
@@ -64,7 +84,7 @@ fn handleArgs(alloc: std.mem.Allocator, out: *Writer) !void {
             };
             continue;
         } else if (std.mem.eql(u8, arg, "-d")) {
-            const nxt = args.next();
+            const nxt = arg_iter.next();
             if (nxt == null or nxt.?.len == 0) {
                 try out.print("-d must be followed by a location\n", .{});
                 return errors.InvalidArguments;
@@ -72,7 +92,7 @@ fn handleArgs(alloc: std.mem.Allocator, out: *Writer) !void {
             extLoc = nxt.?;
             continue;
         } else if (std.mem.eql(u8, arg, "-p")) {
-            const nxt = args.next();
+            const nxt = arg_iter.next();
             if (nxt == null or nxt.?.len == 0) {
                 try out.print("-p must be followed by a number\n", .{});
                 return errors.InvalidArguments;
@@ -82,10 +102,22 @@ fn handleArgs(alloc: std.mem.Allocator, out: *Writer) !void {
                 return errors.InvalidArguments;
             };
             continue;
+        } else if (std.mem.eql(u8, arg, "-v")) {
+            verbose = true;
+            continue;
+        } else if (std.mem.eql(u8, arg, "-dx")) {
+            ignore_xattrs = true;
+            continue;
+        } else if (std.mem.eql(u8, arg, "-dp")) {
+            ignore_permissions = true;
+            continue;
+        } else if (std.mem.eql(u8, arg, "--force")) {
+            force = true;
+            continue;
         } else if (std.mem.eql(u8, arg, "--version")) {
-            try out.print("zig-unsquashfs version ", .{});
+            try out.print("zig-unsquashfs v", .{});
             try config.version.format(out);
-            try out.print("\nBuilt using Zig {s} with {} backend in {} mode.\n", .{ builtin.zig_version_string, builtin.zig_backend, builtin.mode });
+            try out.print("\nBuilt using Zig {s} in {} mode\n", .{ builtin.zig_version_string, builtin.mode });
             std.process.exit(0);
             return;
         } else if (std.mem.eql(u8, arg, "--help")) {

src/decomp.zig

@@ -1,260 +1,31 @@
-//! Implementations for decompression.
-//! TODO: change to vtable interface to allow for shared decompressors for better performance/resource usage.
-
 const std = @import("std");
-const Reader = std.Io.Reader;
-const builtin = @import("builtin");
 
-const config = if (builtin.is_test) .{
-    .use_c_libs = builtin.link_libc == true,
-    .allow_lzo = false, // Change once LZO compilation is fixed
-} else @import("config");
+pub const StatelessDecomp = *const fn (std.mem.Allocator, in: []u8, out: []u8) Error!usize;
 
-const c = @cImport({
-    @cInclude("zlib.h");
-    @cInclude("lzma.h");
-    @cInclude("lz4.h");
-    @cInclude("zstd.h");
-    @cInclude("zstd_errors.h");
-    if (config.allow_lzo)
-        @cInclude("lzo/minilzo.h");
-});
-
-pub const CompressionType = enum(u16) {
-    gzip = 1,
-    lzma,
-    lzo,
-    xz,
-    lz4,
-    zstd,
+pub const Error = error{
+    OutOfMemory,
+    EndOfStream,
+    ReadFailed,
+    WriteFailed,
 };
 
-pub const DecompFn = *const fn (alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize; // TODO: replace anyerror to definitive error types.
+const Decompressor = @This();
 
-pub const gzipDecompress = if (config.use_c_libs) cGzip else zigGzip;
+alloc: std.mem.Allocator = std.heap.smp_allocator,
+vtable: *const struct {
+    decompress: *const fn (*const Decompressor, in: []u8, out: []u8) Error!usize = defaultDecompress,
+    stateless: StatelessDecomp,
+},
 
-fn zigGzip(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    var rdr: Reader = .fixed(in);
-    const buf = try alloc.alloc(u8, out.len);
-    defer alloc.free(buf);
-    var decomp = std.compress.flate.Decompress.init(&rdr, .zlib, buf);
-    return decomp.reader.readSliceShort(out);
-}
-fn cGzip(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    _ = alloc;
-    var out_len: usize = out.len;
-    const res = c.uncompress(out.ptr, &out_len, in.ptr, in.len);
-    return switch (res) {
-        c.Z_OK => out_len,
-        c.Z_MEM_ERROR => error.NotEnoughMemory,
-        c.Z_BUF_ERROR => error.OutBufferTooSmall,
-        c.Z_DATA_ERROR => error.BadData,
-        else => error.UnknownResult,
-    };
+/// Create a copy of the decompressor using it's stateless function and the new allocator.
+pub fn statelessCopy(self: Decompressor, alloc: std.mem.Allocator) Decompressor {
+    return &.{ .alloc = alloc, .vtable = &.{ .stateless = self.vtable.stateless } };
 }
 
-pub const lzmaDecompress = if (config.use_c_libs) cLzma else zigLzma;
-
-fn zigLzma(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    var rdr: Reader = .fixed(in);
-    var decomp = try std.compress.lzma.decompress(alloc, rdr.adaptToOldInterface());
-    return decomp.read(out);
-}
-fn cLzma(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    _ = alloc;
-    var stream: c.lzma_stream = .{
-        .next_in = in.ptr,
-        .avail_in = in.len,
-        .next_out = out.ptr,
-        .avail_out = out.len,
-    };
-    var res = c.lzma_alone_decoder(&stream, in.len * 2);
-    switch (res) {
-        c.LZMA_OK => {},
-        c.LZMA_MEM_ERROR => return error.LzmaMemoryError,
-        c.LZMA_PROG_ERROR => return error.LzmaProgramError,
-        else => return error.UnknownResult,
-    }
-    defer c.lzma_end(&stream);
-    while (res == c.LZMA_OK)
-        res = c.lzma_code(&stream, c.LZMA_RUN);
-    return switch (res) {
-        c.LZMA_STREAM_END => stream.total_out,
-        c.LZMA_MEM_ERROR => error.LzmaMemoryError,
-        c.LZMA_MEMLIMIT_ERROR => error.LzmaMemoryLimit,
-        c.LZMA_FORMAT_ERROR => error.LzmaBadFormat,
-        c.LZMA_DATA_ERROR => error.LzmaDataCorrupt,
-        c.LZMA_BUF_ERROR => error.LzmaCannotProgress,
-        c.LZMA_PROG_ERROR => error.LzmaProgramError,
-        else => error.UnknownResult,
-    };
+pub fn decompress(self: *const Decompressor, in: []u8, out: []u8) Error!usize {
+    return self.vtable.decompress(self, in, out);
 }
 
-// pub const lzoDecompress = if (config.use_c_libs) cLzo else zigLzo;
-
-// fn zigLzo(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-//     _ = alloc;
-//     _ = in;
-//     _ = out;
-//     return error.LzoUnsupported;
-// }
-pub fn cLzo(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    _ = alloc;
-    var res = c.lzo_init();
-    if (res != 0) return error.LzoInitFailed;
-    var out_len: usize = out.len;
-    res = c.lzo1x_decompress(in.ptr, in.len, out.ptr, &out_len, null);
-    return switch (res) {
-        c.LZO_E_OK => out_len,
-        c.LZO_E_ERROR => error.LzoError,
-        c.LZO_E_OUT_OF_MEMORY => error.LzoOutOfMemory,
-        c.LZO_E_NOT_COMPRESSIBLE => error.LzoNotCompressible,
-        c.LZO_E_INPUT_OVERRUN => error.LzoInputOverrun,
-        c.LZO_E_OUTPUT_OVERRUN => error.LzoOutputOverrun,
-        c.LZO_E_LOOKBEHIND_OVERRUN => error.LzoLookbehindOverrun,
-        c.LZO_E_EOF_NOT_FOUND => error.LzoEofNotFound,
-        c.LZO_E_INPUT_NOT_CONSUMED => error.LzoInputNotConsumed,
-        c.LZO_E_NOT_YET_IMPLEMENTED => error.LzoNotYetImplemented,
-        c.LZO_E_INVALID_ARGUMENT => error.LzoInvalidArgument,
-        c.LZO_E_INVALID_ALIGNMENT => error.LzoInvalidAlignment,
-        c.LZO_E_OUTPUT_NOT_CONSUMED => error.LzoOutputNotConsumed,
-        c.LZO_E_INTERNAL_ERROR => error.LzoInternalError,
-        else => error.UnknownResult,
-    };
+fn defaultDecompress(self: *const Decompressor, in: []u8, out: []u8) Error!usize {
+    return self.vtable.stateless(self.alloc, in, out);
 }
-
-pub const xzDecompress = if (config.use_c_libs) cXz else zigXz;
-
-fn zigXz(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    var rdr: Reader = .fixed(in);
-    var decomp = try std.compress.xz.decompress(alloc, rdr.adaptToOldInterface());
-    return decomp.read(out);
-}
-fn cXz(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    _ = alloc;
-    var stream: c.lzma_stream = .{
-        .next_in = in.ptr,
-        .avail_in = in.len,
-        .next_out = out.ptr,
-        .avail_out = out.len,
-    };
-    var res = c.lzma_stream_decoder(&stream, in.len * 2, 0);
-    switch (res) {
-        c.LZMA_OK => {},
-        c.LZMA_MEM_ERROR => return error.LzmaMemoryError,
-        c.LZMA_PROG_ERROR => return error.LzmaProgramError,
-        else => return error.UnknownResult,
-    }
-    defer c.lzma_end(&stream);
-    while (res == c.LZMA_OK)
-        res = c.lzma_code(&stream, c.LZMA_RUN);
-    return switch (res) {
-        c.LZMA_STREAM_END => stream.total_out,
-        c.LZMA_MEM_ERROR => error.LzmaMemoryError,
-        c.LZMA_MEMLIMIT_ERROR => error.LzmaMemoryLimit,
-        c.LZMA_FORMAT_ERROR => error.LzmaBadFormat,
-        c.LZMA_DATA_ERROR => error.LzmaDataCorrupt,
-        c.LZMA_BUF_ERROR => error.LzmaCannotProgress,
-        c.LZMA_PROG_ERROR => error.LzmaProgramError,
-        else => error.UnknownResult,
-    };
-}
-
-// pub const lz4Decompress = if (config.use_c_libs) cLz4 else zigLz4;
-
-// fn zigLz4(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-//     _ = alloc;
-//     _ = in;
-//     _ = out;
-//     return error.Lz4Unsupported;
-// }
-pub fn cLz4(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    _ = alloc;
-    const res = c.LZ4_decompress_safe(in.ptr, out.ptr, @intCast(in.len), @intCast(out.len));
-    if (res > 0) return @abs(res); // TODO: Find out what error values it can return.
-    return error.Lz4DecompressFailed;
-}
-
-pub const zstdDecompress = if (config.use_c_libs) cZstd else zigZstd;
-
-pub fn zigZstd(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    var rdr: Reader = .fixed(in);
-    const buf = try alloc.alloc(u8, 1024 * 1024);
-    defer alloc.free(buf);
-    var decomp = std.compress.zstd.Decompress.init(&rdr, buf, .{});
-    return decomp.reader.readSliceShort(out) catch |err| {
-        return decomp.err orelse err;
-    };
-}
-fn cZstd(alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize {
-    _ = alloc;
-    const res = c.ZSTD_decompress(out.ptr, out.len, in.ptr, in.len);
-    if (c.ZSTD_isError(res) == 0) return res;
-    return switch (c.ZSTD_getErrorCode(res)) {
-        c.ZSTD_error_prefix_unknown => cZstdError.PrefixUnknown,
-        c.ZSTD_error_version_unsupported => cZstdError.VersionUnsupported,
-        c.ZSTD_error_frameParameter_unsupported => cZstdError.FrameParameterUnsupported,
-        c.ZSTD_error_frameParameter_windowTooLarge => cZstdError.FrameParameterWindowTooLarge,
-        c.ZSTD_error_corruption_detected => cZstdError.CorruptionDetected,
-        c.ZSTD_error_checksum_wrong => cZstdError.ChecksumWrong,
-        c.ZSTD_error_literals_headerWrong => cZstdError.LiteralsHeaderWrong,
-        c.ZSTD_error_dictionary_corrupted => cZstdError.DictionaryCorrupted,
-        c.ZSTD_error_dictionary_wrong => cZstdError.DictionaryWrong,
-        c.ZSTD_error_dictionaryCreation_failed => cZstdError.DictionaryCreationFailed,
-        c.ZSTD_error_parameter_unsupported => cZstdError.ParameterUnsupported,
-        c.ZSTD_error_parameter_combination_unsupported => cZstdError.ParameterCombinationUnsupported,
-        c.ZSTD_error_parameter_outOfBound => cZstdError.ParameterOutOfBound,
-        c.ZSTD_error_tableLog_tooLarge => cZstdError.TableLogTooLarge,
-        c.ZSTD_error_maxSymbolValue_tooLarge => cZstdError.MaxSymbolValueTooLarge,
-        c.ZSTD_error_maxSymbolValue_tooSmall => cZstdError.MaxSymbolValueTooSmall,
-        c.ZSTD_error_stabilityCondition_notRespected => cZstdError.StabilityConditionNotRespected,
-        c.ZSTD_error_stage_wrong => cZstdError.StageWrong,
-        c.ZSTD_error_init_missing => cZstdError.InitMissing,
-        c.ZSTD_error_memory_allocation => cZstdError.MemoryAllocation,
-        c.ZSTD_error_workSpace_tooSmall => cZstdError.WorkSpaceTooSmall,
-        c.ZSTD_error_dstSize_tooSmall => cZstdError.DstSizeTooSmall,
-        c.ZSTD_error_srcSize_wrong => cZstdError.SrcSizeWrong,
-        c.ZSTD_error_dstBuffer_null => cZstdError.DstBufferNull,
-        c.ZSTD_error_noForwardProgress_destFull => cZstdError.NoForwardProgressDestFull,
-        c.ZSTD_error_noForwardProgress_inputEmpty => cZstdError.NoForwardProgressInputEmpty,
-        else => cZstdError.Generic,
-    };
-}
-
-pub const cZstdError = error{
-    Generic,
-    PrefixUnknown,
-    VersionUnsupported,
-    FrameParameterUnsupported,
-    FrameParameterWindowTooLarge,
-    CorruptionDetected,
-    ChecksumWrong,
-    LiteralsHeaderWrong,
-    DictionaryCorrupted,
-    DictionaryWrong,
-    DictionaryCreationFailed,
-    ParameterUnsupported,
-    ParameterCombinationUnsupported,
-    ParameterOutOfBound,
-    TableLogTooLarge,
-    MaxSymbolValueTooLarge,
-    MaxSymbolValueTooSmall,
-    CannotProduceUncompressedBlock,
-    StabilityConditionNotRespected,
-    StageWrong,
-    InitMissing,
-    MemoryAllocation,
-    WorkSpaceTooSmall,
-    DstSizeTooSmall,
-    SrcSizeWrong,
-    DstBufferNull,
-    NoForwardProgressDestFull,
-    NoForwardProgressInputEmpty,
-    FrameIndexTooLarge,
-    SeekableIo,
-    DstBufferWrong,
-    SrcBufferWrong,
-    SequenceProducerFailed,
-    ExternalSequencesInvalid,
-    MaxCode,
-};

src/decomp/c/lz4.zig

@@ -0,0 +1,12 @@
+const std = @import("std");
+
+const c = @import("c");
+const Decompressor = @import("../../decomp.zig");
+
+interface: Decompressor = .{ .vtable = &.{ .stateless = stateless } },
+
+pub fn stateless(_: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    const res = c.LZ4_decompress_safe(in.ptr, out.ptr, @intCast(in.len), @intCast(out.len));
+    if (res > 0) return @abs(res);
+    return Decompressor.Error.ReadFailed; // TOOD: Find out what errors can be returned.
+}

src/decomp/c/lzma.zig

@@ -0,0 +1,142 @@
+const std = @import("std");
+
+const c = @import("c");
+const Decompressor = @import("../../decomp.zig");
+
+const Lzma = @This();
+
+streams: std.AutoHashMap(std.Thread.Id, c.lzma_stream),
+
+interface: Decompressor,
+
+err: ?Error = null,
+
+pub fn init(alloc: std.mem.Allocator) !Lzma {
+    return .{
+        .streams = try .init(alloc),
+        .interface = &.{
+            .alloc = alloc,
+            .vtable = .{ .decompress = decompress, .stateless = stateless },
+        },
+    };
+}
+pub fn deinit(self: *Lzma) void {
+    var values = self.streams.valueIterator();
+    while (values.next()) |val| {
+        c.lzma_end(val);
+    }
+    self.streams.deinit();
+}
+
+fn getOrCreate(self: *Lzma) !*c.lzma_stream {
+    const res = try self.streams.getOrPut(std.Thread.getCurrentId());
+    if (res.found_existing) return res.value_ptr;
+    res.value_ptr.* = .{
+        .alloc = .{
+            .alloc = lzmaAlloc,
+            .free = lzmaFree,
+            .@"opaque" = &self.interface.alloc,
+        },
+    };
+    return res.value_ptr;
+}
+
+fn decompress(decomp: *const Decompressor, in: []u8, out: []u8) Decompressor.Error!usize {
+    var self: *Lzma = @fieldParentPtr("interface", decomp);
+
+    const stream = try self.getOrCreate();
+    stream.next_in = in.ptr;
+    stream.avail_in = in.len;
+    stream.next_out = out.ptr;
+    stream.avail_out = out.len;
+    var res = c.lzma_alone_decoder(stream, out.len);
+    decodeResult(res) catch |err| {
+        self.err = err;
+        return lzmaErrorToDecompError(err);
+    };
+    while (true) {
+        res = c.lzma_code(&stream, c.LZMA_RUN);
+        if (res == c.LZMA_OK) continue;
+        if (res == c.LZMA_STREAM_END) break;
+        decodeResult(res) catch |err| {
+            self.err = err;
+            return lzmaErrorToDecompError(err);
+        };
+    }
+    return stream.total_out;
+}
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var stream: c.lzma_stream = .{
+        .next_in = in.ptr,
+        .avail_in = in.len,
+        .next_out = out.ptr,
+        .avail_out = out.len,
+        .allocator = &.{
+            .alloc = lzmaAlloc,
+            .free = lzmaFree,
+            .@"opaque" = @ptrCast(@constCast(&alloc)),
+        },
+    };
+    var res = c.lzma_alone_decoder(&stream, out.len);
+    decodeResult(res) catch |err| return lzmaErrorToDecompError(err);
+    while (true) {
+        res = c.lzma_code(&stream, c.LZMA_RUN);
+        if (res == c.LZMA_OK) continue;
+        if (res == c.LZMA_STREAM_END) break;
+        decodeResult(res) catch |err| return lzmaErrorToDecompError(err);
+    }
+    return stream.total_out;
+}
+
+inline fn decodeResult(res: c_uint) Error!void {
+    return switch (res) {
+        c.LZMA_OK => {},
+        c.LZMA_STREAM_END => {},
+        c.LZMA_NO_CHECK => {},
+        c.LZMA_UNSUPPORTED_CHECK => Error.UnsupportedCheck,
+        c.LZMA_MEM_ERROR => Error.OutOfMemory,
+        c.LZMA_MEMLIMIT_ERROR => Error.OutOfMemory,
+        c.LZMA_FORMAT_ERROR => Error.Format,
+        c.LZMA_OPTIONS_ERROR => Error.Options,
+        c.LZMA_DATA_ERROR => Error.Data,
+        c.LZMA_BUF_ERROR => Error.BufferExhausted,
+        c.LZMA_PROG_ERROR => Error.Programming,
+        c.LZMA_SEEK_NEEDED => Error.SeekNeeded,
+        else => Error.Unknown,
+    };
+}
+fn lzmaErrorToDecompError(err: Error) Decompressor.Error {
+    switch (err) {
+        Error.OutOfMemory => return Decompressor.Error.OutOfMemory,
+        Error.UnsupportedCheck => return Decompressor.Error.ReadFailed,
+        Error.Format => return Decompressor.Error.ReadFailed,
+        Error.Options => return Decompressor.Error.ReadFailed,
+        Error.Data => return Decompressor.Error.ReadFailed,
+        Error.BufferExhausted => return Decompressor.Error.WriteFailed,
+        Error.Programming => return Decompressor.Error.ReadFailed,
+        Error.SeekNeeded => return Decompressor.Error.ReadFailed,
+        Error.Unknown => return Decompressor.Error.ReadFailed,
+    }
+}
+
+fn lzmaAlloc(ptr: ?*anyopaque, _: usize, size: usize) callconv(.c) ?*anyopaque {
+    var alloc: *std.mem.Allocator = @ptrCast(@alignCast(ptr));
+    return alloc.rawAlloc(size, .@"1", 0);
+}
+fn lzmaFree(ptr: ?*anyopaque, alloc_ptr: ?*anyopaque) callconv(.c) void {
+    if (alloc_ptr == null) return;
+    var alloc: *std.mem.Allocator = @ptrCast(@alignCast(ptr));
+    alloc.rawFree(@ptrCast(alloc_ptr), .@"1", 0);
+}
+
+pub const Error = error{
+    OutOfMemory,
+    UnsupportedCheck,
+    Format,
+    Options,
+    Data,
+    BufferExhausted,
+    Programming,
+    SeekNeeded,
+    Unknown,
+};

src/decomp/c/lzo.zig

@@ -0,0 +1,28 @@
+const std = @import("std");
+
+const c = @import("c");
+const Decompressor = @import("../../decomp.zig");
+
+interface: Decompressor = .{ .vtable = &.{ .stateless = stateless } },
+
+pub fn stateless(_: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var out_len = out.len;
+    const res = c.lzo1x_decompress(in.ptr, in.len, out.ptr, &out_len, null);
+    return switch (res) {
+        c.LZO_E_OK => out_len,
+        c.LZO_E_ERROR => Decompressor.Error.ReadFailed,
+        c.LZO_E_OUT_OF_MEMORY => Decompressor.Error.OutOfMemory,
+        c.LZO_E_NOT_COMPRESSIBLE => Decompressor.Error.ReadFailed,
+        c.LZO_E_INPUT_OVERRUN => Decompressor.Error.ReadFailed,
+        c.LZO_E_OUTPUT_OVERRUN => Decompressor.Error.WriteFailed,
+        c.LZO_E_LOOKBEHIND_OVERRUN => Decompressor.Error.ReadFailed,
+        c.LZO_E_EOF_NOT_FOUND => Decompressor.Error.ReadFailed,
+        c.LZO_E_INPUT_NOT_CONSUMED => Decompressor.Error.ReadFailed,
+        c.LZO_E_NOT_YET_IMPLEMENTED => Decompressor.Error.ReadFailed,
+        c.LZO_E_INVALID_ARGUMENT => Decompressor.Error.ReadFailed,
+        c.LZO_E_INVALID_ALIGNMENT => Decompressor.Error.ReadFailed,
+        c.LZO_E_OUTPUT_NOT_CONSUMED => Decompressor.Error.WriteFailed,
+        c.LZO_E_INTERNAL_ERROR => Decompressor.Error.ReadFailed,
+        else => Decompressor.Error.ReadFailed,
+    };
+}

src/decomp/c/xz.zig

@@ -0,0 +1,142 @@
+const std = @import("std");
+
+const c = @import("c");
+const Decompressor = @import("../../decomp.zig");
+
+const Xz = @This();
+
+streams: std.AutoHashMap(std.Thread.Id, c.lzma_stream),
+
+interface: Decompressor,
+
+err: ?Error = null,
+
+pub fn init(alloc: std.mem.Allocator) !Xz {
+    return .{
+        .streams = try .init(alloc),
+        .interface = &.{
+            .alloc = alloc,
+            .vtable = .{ .decompress = decompress, .stateless = stateless },
+        },
+    };
+}
+pub fn deinit(self: *Xz) void {
+    var values = self.streams.valueIterator();
+    while (values.next()) |val| {
+        c.xz_end(val);
+    }
+    self.streams.deinit();
+}
+
+fn getOrCreate(self: *Xz) !*c.xz_stream {
+    const res = try self.streams.getOrPut(std.Thread.getCurrentId());
+    if (res.found_existing) return res.value_ptr;
+    res.value_ptr.* = .{
+        .alloc = .{
+            .alloc = lzmaAlloc,
+            .free = lzmaFree,
+            .@"opaque" = &self.interface.alloc,
+        },
+    };
+    return res.value_ptr;
+}
+
+fn decompress(decomp: *const Decompressor, in: []u8, out: []u8) Decompressor.Error!usize {
+    var self: *Xz = @fieldParentPtr("interface", decomp);
+
+    const stream = try self.getOrCreate();
+    stream.next_in = in.ptr;
+    stream.avail_in = in.len;
+    stream.next_out = out.ptr;
+    stream.avail_out = out.len;
+    var res = c.lzma_alone_decoder(stream, out.len);
+    decodeResult(res) catch |err| {
+        self.err = err;
+        return xzErrorToDecompError(err);
+    };
+    while (true) {
+        res = c.lzma_code(&stream, c.LZMA_RUN);
+        if (res == c.LZMA_OK) continue;
+        if (res == c.LZMA_STREAM_END) break;
+        decodeResult(res) catch |err| {
+            self.err = err;
+            return xzErrorToDecompError(err);
+        };
+    }
+    return stream.total_out;
+}
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var stream: c.lzma_stream = .{
+        .next_in = in.ptr,
+        .avail_in = in.len,
+        .next_out = out.ptr,
+        .avail_out = out.len,
+        .allocator = &.{
+            .alloc = lzmaAlloc,
+            .free = lzmaFree,
+            .@"opaque" = @ptrCast(@constCast(&alloc)),
+        },
+    };
+    var res = c.lzma_alone_decoder(&stream, out.len);
+    decodeResult(res) catch |err| return xzErrorToDecompError(err);
+    while (true) {
+        res = c.lzma_code(&stream, c.LZMA_RUN);
+        if (res == c.LZMA_OK) continue;
+        if (res == c.LZMA_STREAM_END) break;
+        decodeResult(res) catch |err| return xzErrorToDecompError(err);
+    }
+    return stream.total_out;
+}
+
+inline fn decodeResult(res: c_uint) Error!void {
+    return switch (res) {
+        c.LZMA_OK => {},
+        c.LZMA_STREAM_END => {},
+        c.LZMA_NO_CHECK => {},
+        c.LZMA_UNSUPPORTED_CHECK => Error.UnsupportedCheck,
+        c.LZMA_MEM_ERROR => Error.OutOfMemory,
+        c.LZMA_MEMLIMIT_ERROR => Error.OutOfMemory,
+        c.LZMA_FORMAT_ERROR => Error.Format,
+        c.LZMA_OPTIONS_ERROR => Error.Options,
+        c.LZMA_DATA_ERROR => Error.Data,
+        c.LZMA_BUF_ERROR => Error.BufferExhausted,
+        c.LZMA_PROG_ERROR => Error.Programming,
+        c.LZMA_SEEK_NEEDED => Error.SeekNeeded,
+        else => Error.Unknown,
+    };
+}
+fn xzErrorToDecompError(err: Error) Decompressor.Error {
+    switch (err) {
+        Error.OutOfMemory => return Decompressor.Error.OutOfMemory,
+        Error.UnsupportedCheck => return Decompressor.Error.ReadFailed,
+        Error.Format => return Decompressor.Error.ReadFailed,
+        Error.Options => return Decompressor.Error.ReadFailed,
+        Error.Data => return Decompressor.Error.ReadFailed,
+        Error.BufferExhausted => return Decompressor.Error.WriteFailed,
+        Error.Programming => return Decompressor.Error.ReadFailed,
+        Error.SeekNeeded => return Decompressor.Error.ReadFailed,
+        Error.Unknown => return Decompressor.Error.ReadFailed,
+    }
+}
+
+fn lzmaAlloc(ptr: ?*anyopaque, _: usize, size: usize) callconv(.c) ?*anyopaque {
+    var alloc: *std.mem.Allocator = @ptrCast(@alignCast(ptr));
+    return alloc.rawAlloc(size, .@"1", 0);
+}
+fn lzmaFree(ptr: ?*anyopaque, alloc_ptr: ?*anyopaque) callconv(.c) void {
+    if (alloc_ptr == null) return;
+    var alloc: *std.mem.Allocator = @ptrCast(@alignCast(ptr));
+    alloc.rawFree(@ptrCast(alloc_ptr), .@"1", 0);
+}
+
+pub const Error = error{
+    OutOfMemory,
+    UnsupportedCheck,
+    Format,
+    Options,
+    Data,
+    BufferExhausted,
+    Programming,
+    SeekNeeded,
+    Unknown,
+};

src/decomp/c/zlib.zig

@@ -0,0 +1,106 @@
+const std = @import("std");
+
+const c = @import("c");
+const Decompressor = @import("../../decomp.zig");
+
+const Zlib = @This();
+
+streams: std.AutoHashMap(std.Thread.Id, c.zng_stream),
+
+interface: Decompressor,
+
+err: ?Error = null,
+
+pub fn init(alloc: std.mem.Allocator) !Zlib {
+    return .{
+        .streams = try .init(alloc),
+        .interface = &.{
+            .alloc = alloc,
+            .vtable = .{ .decompress = decompress, .stateless = stateless },
+        },
+    };
+}
+pub fn deinit(self: *Zlib) void {
+    var values = self.streams.valueIterator();
+    while (values.next()) |val| {
+        _ = c.zng_deflateEnd(val);
+    }
+    self.streams.deinit();
+}
+
+fn getOrCreate(self: *Zlib) !*c.zng_stream {
+    const res = try self.streams.getOrPut(std.Thread.getCurrentId());
+    if (res.found_existing) return res.value_ptr;
+    res.value_ptr.* = .{
+        .@"opaque" = self,
+        .zalloc = zalloc,
+        .zfree = zfree,
+    };
+    return res.value_ptr;
+}
+
+fn decompress(decomp: *const Decompressor, in: []u8, out: []u8) Decompressor.Error!usize {
+    var self: *Zlib = @fieldParentPtr("interface", decomp);
+
+    var stream = try self.getOrCreate();
+    stream.next_in = in.ptr;
+    stream.avail_in = in.len;
+    stream.next_out = out.ptr;
+    stream.avail_out = out.len;
+    var res = c.zng_inflateReset(stream);
+    decodeError(res) catch |err| {
+        self.err = err;
+        return Decompressor.Error.ReadFailed;
+    };
+    res = c.zng_inflate(stream, c.Z_FINISH);
+    decodeError(res) catch |err| {
+        self.err = err;
+        return switch (err) {
+            Error.OutOfMemory => err,
+            else => Decompressor.Error.ReadFailed,
+        };
+    };
+}
+
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    _ = alloc;
+    var out_len = out.len;
+    const res = c.zng_uncompress(out.ptr, &out_len, in.ptr, in.len);
+    return switch (res) {
+        c.Z_OK => out_len,
+        c.Z_MEM_ERROR => Decompressor.Error.OutOfMemory,
+        c.Z_BUF_ERROR => Decompressor.Error.WriteFailed,
+        else => Decompressor.Error.ReadFailed,
+    };
+}
+
+inline fn decodeError(res: i32) Error!void {
+    return switch (res) {
+        c.Z_OK => {},
+        c.Z_STREAM_ERROR => Error.Stream,
+        c.Z_BUF_ERROR => Error.Buffer,
+        c.Z_MEM_ERROR => Error.OutOfMemory,
+        c.Z_DATA_ERROR => Error.Data,
+        c.Z_VERSION_ERROR => Error.Version,
+        else => Error.Unknown,
+    };
+}
+
+fn zalloc(ptr: ?*anyopaque, items: c_uint, size: c_uint) callconv(.c) ?*anyopaque {
+    var self: *Zlib = @ptrCast(ptr);
+    return self.interface.alloc.rawAlloc(items * size, .@"1", 0);
+}
+
+fn zfree(ptr: ?*anyopaque, addr: ?*anyopaque) callconv(.c) void {
+    var self: *Zlib = @ptrCast(ptr);
+    self.interface.alloc.rawFree(@ptrCast(addr), .@"1", 0);
+}
+
+pub const Error = error{
+    OutOfMemory,
+    Stream,
+    Buffer,
+    Data,
+    Version,
+    Unknown,
+};

src/decomp/c/zstd.zig

@@ -0,0 +1,172 @@
+const std = @import("std");
+
+const c = @import("c");
+const Decompressor = @import("../../decomp.zig");
+
+const Zstd = @This();
+
+context: std.AutoHashMap(std.Thread.Id, ?*c.ZSTD_DCtx),
+
+interface: Decompressor,
+
+err: ?Error = null,
+
+pub fn init(alloc: std.mem.Allocator) !Zstd {
+    return .{
+        .streams = try .init(alloc),
+        .interface = &.{
+            .alloc = alloc,
+            .vtable = .{ .decompress = decompress, .stateless = stateless },
+        },
+    };
+}
+pub fn deinit(self: *Zstd) void {
+    var values = self.context.valueIterator();
+    while (values.next()) |val| {
+        _ = c.ZSTD_freeDCtx(val.*);
+    }
+    self.context.deinit();
+}
+
+fn getOrCreate(self: *Zstd) !*c.ZSTD_DCtx {
+    const res = try self.context.getOrPut(std.Thread.getCurrentId());
+    if (res.found_existing) return res.value_ptr;
+    res.value_ptr.* = c.ZSTD_createDCtx();
+    if (res.value_ptr.* == null) return Error.OutOfMemory;
+    return res.value_ptr;
+}
+
+fn decompress(decomp: *const Decompressor, in: []u8, out: []u8) Decompressor.Error!usize {
+    var self: *Zstd = @fieldParentPtr("interface", decomp);
+
+    const ctx = self.getOrCreate();
+    const res = c.ZSTD_decompressDCtx(ctx, out.ptr, out.len, in.ptr, in.len);
+    decodeError(res) catch |err| {
+        self.err = err;
+        return ZstdErrorToDecompError(err);
+    };
+    return res;
+}
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    _ = alloc;
+    const res = c.ZSTD_decompress(out.ptr, out.len, in.ptr, in.len);
+    decodeError(res) catch |err| return ZstdErrorToDecompError(err);
+    return res;
+}
+
+inline fn decodeError(res: usize) Error!void {
+    if (c.ZSTD_isError(res) == 0) return;
+    return switch (c.ZSTD_getErrorCode(res)) {
+        c.ZSTD_error_prefix_unknown => Error.PrefixUnknown,
+        c.ZSTD_error_version_unsupported => Error.VersionUnsupported,
+        c.ZSTD_error_frameParameter_unsupported => Error.FrameParameterUnsupported,
+        c.ZSTD_error_frameParameter_windowTooLarge => Error.FrameParameterWindowTooLarge,
+        c.ZSTD_error_corruption_detected => Error.CorruptionDetected,
+        c.ZSTD_error_checksum_wrong => Error.ChecksumWrong,
+        c.ZSTD_error_literals_headerWrong => Error.LiteralsHeaderWrong,
+        c.ZSTD_error_dictionary_corrupted => Error.DictionaryCorrupted,
+        c.ZSTD_error_dictionary_wrong => Error.DictionaryWrong,
+        c.ZSTD_error_dictionaryCreation_failed => Error.DictionaryCreationFailed,
+        c.ZSTD_error_parameter_unsupported => Error.ParameterUnsupported,
+        c.ZSTD_error_parameter_combination_unsupported => Error.ParameterCombinationUnsupported,
+        c.ZSTD_error_parameter_outOfBound => Error.ParameterOutOfBound,
+        c.ZSTD_error_tableLog_tooLarge => Error.TableLogTooLarge,
+        c.ZSTD_error_maxSymbolValue_tooLarge => Error.MaxSymbolValueTooLarge,
+        c.ZSTD_error_maxSymbolValue_tooSmall => Error.MaxSymbolValueTooSmall,
+        c.ZSTD_error_cannotProduce_uncompressedBlock => Error.CannotProduceUncompressedBlock,
+        c.ZSTD_error_stabilityCondition_notRespected => Error.StabilityConditionNotRespected,
+        c.ZSTD_error_stage_wrong => Error.StageWrong,
+        c.ZSTD_error_init_missing => Error.InitMissing,
+        c.ZSTD_error_memory_allocation => Error.MemoryAllocation,
+        c.ZSTD_error_workSpace_tooSmall => Error.WorkSpaceTooSmall,
+        c.ZSTD_error_dstSize_tooSmall => Error.DstSizeTooSmall,
+        c.ZSTD_error_srcSize_wrong => Error.SrcSizeWrong,
+        c.ZSTD_error_dstBuffer_null => Error.DstBufferNull,
+        c.ZSTD_error_noForwardProgress_destFull => Error.NoForwardProgressDestFull,
+        c.ZSTD_error_noForwardProgress_inputEmpty => Error.NoForwardProgressInputEmpty,
+        c.ZSTD_error_frameIndex_tooLarge => Error.FrameIndexTooLarge,
+        c.ZSTD_error_seekableIO => Error.SeekableIo,
+        c.ZSTD_error_dstBuffer_wrong => Error.DstBufferWrong,
+        c.ZSTD_error_srcBuffer_wrong => Error.SrcBufferWrong,
+        c.ZSTD_error_sequenceProducer_failed => Error.SequenceProducerFailed,
+        c.ZSTD_error_externalSequences_invalid => Error.ExternalSequencesInvalid,
+        else => Error.Generic,
+    };
+}
+inline fn ZstdErrorToDecompError(err: Error) Decompressor.Error {
+    return switch (err) {
+        Error.OutOfMemory => Decompressor.Error.OutOfMemory,
+        Error.Generic => Decompressor.Error.ReadFailed,
+        Error.PrefixUnknown => Decompressor.Error.ReadFailed,
+        Error.VersionUnsupported => Decompressor.Error.ReadFailed,
+        Error.FrameParameterUnsupported => Decompressor.Error.ReadFailed,
+        Error.FrameParameterWindowTooLarge => Decompressor.Error.ReadFailed,
+        Error.CorruptionDetected => Decompressor.Error.ReadFailed,
+        Error.ChecksumWrong => Decompressor.Error.ReadFailed,
+        Error.LiteralsHeaderWrong => Decompressor.Error.ReadFailed,
+        Error.DictionaryCorrupted => Decompressor.Error.ReadFailed,
+        Error.DictionaryWrong => Decompressor.Error.ReadFailed,
+        Error.DictionaryCreationFailed => Decompressor.Error.ReadFailed,
+        Error.ParameterUnsupported => Decompressor.Error.ReadFailed,
+        Error.ParameterCombinationUnsupported => Decompressor.Error.ReadFailed,
+        Error.ParameterOutOfBound => Decompressor.Error.ReadFailed,
+        Error.TableLogTooLarge => Decompressor.Error.ReadFailed,
+        Error.MaxSymbolValueTooLarge => Decompressor.Error.ReadFailed,
+        Error.MaxSymbolValueTooSmall => Decompressor.Error.ReadFailed,
+        Error.CannotProduceUncompressedBlock => Decompressor.Error.ReadFailed,
+        Error.StabilityConditionNotRespected => Decompressor.Error.ReadFailed,
+        Error.StageWrong => Decompressor.Error.ReadFailed,
+        Error.InitMissing => Decompressor.Error.ReadFailed,
+        Error.MemoryAllocation => Decompressor.Error.OutOfMemory,
+        Error.WorkSpaceTooSmall => Decompressor.Error.WriteFailed,
+        Error.DstSizeTooSmall => Decompressor.Error.WriteFailed,
+        Error.SrcSizeWrong => Decompressor.Error.ReadFailed,
+        Error.DstBufferNull => Decompressor.Error.WriteFailed,
+        Error.NoForwardProgressDestFull => Decompressor.Error.WriteFailed,
+        Error.NoForwardProgressInputEmpty => Decompressor.Error.ReadFailed,
+        Error.FrameIndexTooLarge => Decompressor.Error.ReadFailed,
+        Error.SeekableIo => Decompressor.Error.ReadFailed,
+        Error.DstBufferWrong => Decompressor.Error.WriteFailed,
+        Error.SrcBufferWrong => Decompressor.Error.ReadFailed,
+        Error.SequenceProducerFailed => Decompressor.Error.ReadFailed,
+        Error.ExternalSequencesInvalid => Decompressor.Error.ReadFailed,
+    };
+}
+
+pub const Error = error{
+    OutOfMemory,
+    Generic,
+    PrefixUnknown,
+    VersionUnsupported,
+    FrameParameterUnsupported,
+    FrameParameterWindowTooLarge,
+    CorruptionDetected,
+    ChecksumWrong,
+    LiteralsHeaderWrong,
+    DictionaryCorrupted,
+    DictionaryWrong,
+    DictionaryCreationFailed,
+    ParameterUnsupported,
+    ParameterCombinationUnsupported,
+    ParameterOutOfBound,
+    TableLogTooLarge,
+    MaxSymbolValueTooLarge,
+    MaxSymbolValueTooSmall,
+    CannotProduceUncompressedBlock,
+    StabilityConditionNotRespected,
+    StageWrong,
+    InitMissing,
+    MemoryAllocation,
+    WorkSpaceTooSmall,
+    DstSizeTooSmall,
+    SrcSizeWrong,
+    DstBufferNull,
+    NoForwardProgressDestFull,
+    NoForwardProgressInputEmpty,
+    FrameIndexTooLarge,
+    SeekableIo,
+    DstBufferWrong,
+    SrcBufferWrong,
+    SequenceProducerFailed,
+    ExternalSequencesInvalid,
+};

src/decomp/types.zig

@@ -0,0 +1,64 @@
+const config = @import("config");
+
+const Decompressor = @import("../decomp.zig");
+
+pub fn getStatelessFn(decomp: Enum) !Decompressor.StatelessDecomp {
+    if (config.use_zig_decomp) {
+        return switch (decomp) {
+            .gzip => @import("zig/zlib.zig").stateless,
+            .lzma => @import("zig/lzma.zig").stateless,
+            .xz => @import("zig/xz.zig").stateless,
+            .zstd => @import("zig/zstd.zig").stateless,
+            .lz4 => error.ZigLz4Unsupported,
+            .lzo => error.ZigLzoUnsupported,
+        };
+    }
+    return switch (decomp) {
+        .gzip => @import("c/zlib.zig").stateless,
+        .lzma => @import("c/lzma.zig").stateless,
+        .lzo => @import("c/lzo.zig").stateless,
+        .xz => @import("c/xz.zig").stateless,
+        .lz4 => @import("c/lz4.zig").stateless,
+        .zstd => @import("c/zstd.zig").stateless,
+    };
+}
+
+pub const Enum = enum(u16) {
+    gzip = 1, // Though officially named gzip, it actually uses zlib.
+    lzma,
+    lzo,
+    xz,
+    lz4,
+    zstd,
+};
+
+pub const Decomp = if (config.use_zig_decomp)
+    union(enum) {
+        gzip: @import("zig/zlib.zig"),
+        lzma: @import("zig/lzma.zig"),
+        xz: @import("zig/xz.zig"),
+        zstd: @import("zig/zstd.zig"),
+
+        pub fn deinit(_: *Decomp) void {
+            return;
+        }
+    }
+else
+    union(enum) {
+        gzip: @import("c/zlib.zig"),
+        lzma: @import("c/lzma.zig"),
+        lzo: @import("c/lzo.zig"),
+        xz: @import("c/xz.zig"),
+        lz4: @import("c/lz4.zig"),
+        zstd: @import("c/zstd.zig"),
+
+        pub fn deinit(self: *Decomp) void {
+            switch (self) {
+                .gzip => self.gzip.deinit(),
+                .lzma => self.lzma.deinit(),
+                .xz => self.xz.deinit(),
+                .zstd => self.zstd.deinit(),
+                else => {},
+            }
+        }
+    };

src/decomp/zig/lzma.zig

@@ -0,0 +1,23 @@
+const std = @import("std");
+const lzma = std.compress.lzma;
+const Reader = std.Io.Reader;
+
+const Decompressor = @import("../../decomp.zig");
+
+interface: Decompressor = .{ .vtable = &.{ .stateless = stateless } },
+
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var rdr: Reader = .fixed(in);
+    const buf = try alloc.alloc(u8, in.len);
+    defer alloc.free(buf);
+
+    var decomp = lzma.Decompress.initOptions(&rdr, alloc, buf, .{}, out.len) catch |err|
+        return switch (err) {
+            error.Overflow => Decompressor.Error.ReadFailed,
+            error.CorruptInput => Decompressor.Error.ReadFailed,
+            error.InvalidRangeCode => Decompressor.Error.ReadFailed,
+            else => @errorCast(err),
+        };
+    defer decomp.deinit();
+    return decomp.reader.readSliceShort(out);
+}

src/decomp/zig/xz.zig

@@ -0,0 +1,22 @@
+const std = @import("std");
+const xz = std.compress.xz;
+const Reader = std.Io.Reader;
+
+const Decompressor = @import("../../decomp.zig");
+
+interface: Decompressor = .{ .vtable = &.{ .stateless = stateless } },
+
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var rdr: Reader = .fixed(in);
+    const buf = try alloc.alloc(u8, in.len);
+    defer alloc.free(buf);
+
+    var decomp = xz.Decompress.init(&rdr, alloc, buf) catch |err|
+        return switch (err) {
+            error.WrongChecksum => Decompressor.Error.ReadFailed,
+            error.NotXzStream => Decompressor.Error.ReadFailed,
+            else => @errorCast(err),
+        };
+    defer decomp.deinit();
+    return decomp.reader.readSliceShort(out);
+}

src/decomp/zig/zlib.zig

@@ -0,0 +1,16 @@
+const std = @import("std");
+const flate = std.compress.flate;
+const Reader = std.Io.Reader;
+
+const Decompressor = @import("../../decomp.zig");
+
+interface: Decompressor = .{ .vtable = &.{ .stateless = stateless } },
+
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var rdr: Reader = .fixed(in);
+    const buf = try alloc.alloc(u8, out.len);
+    defer alloc.free(buf);
+
+    var decomp = flate.Decompress.init(&rdr, .zlib, buf);
+    return decomp.reader.readSliceShort(out);
+}

src/decomp/zig/zstd.zig

@@ -0,0 +1,16 @@
+const std = @import("std");
+const zstd = std.compress.zstd;
+const Reader = std.Io.Reader;
+
+const Decompressor = @import("../../decomp.zig");
+
+interface: Decompressor = .{ .vtable = &.{ .stateless = stateless } },
+
+pub fn stateless(alloc: std.mem.Allocator, in: []u8, out: []u8) Decompressor.Error!usize {
+    var rdr: Reader = .fixed(in);
+    const buf = try alloc.alloc(u8, out.len * 2);
+    defer alloc.free(buf);
+
+    var decomp = zstd.Decompress.init(&rdr, buf, .{ .window_len = @truncate(out.len) });
+    return decomp.reader.readSliceShort(out);
+}

src/dir_entry.zig

@@ -1,60 +0,0 @@
-//! Directory entry from the directory table.
-
-const std = @import("std");
-const Reader = std.Io.Reader;
-
-const InodeType = @import("inode.zig").InodeType;
-
-const Entry = @This();
-
-const Header = extern struct { // use extern due to bad alignment with packed.
-    count: u32,
-    block_start: u32,
-    num: u32,
-};
-
-const RawEntry = packed struct {
-    offset: u16,
-    inode_offset: i16,
-    inode_type: InodeType,
-    name_size: u16,
-};
-
-block_start: u32,
-block_offset: u16,
-num: u32,
-inode_type: InodeType,
-name: []const u8,
-
-pub fn readDir(alloc: std.mem.Allocator, rdr: *Reader, size: u32) ![]Entry {
-    var cur_red: u32 = 3; // start at 3 due to "." & ".." being counted in the dir size.
-    var hdr: Header = undefined;
-    var raw: RawEntry = undefined;
-    var out: std.ArrayList(Entry) = try .initCapacity(alloc, 100); // Start out with a decent capacity instead of needing to allocate per header.
-    errdefer out.deinit(alloc);
-    while (cur_red < size) {
-        try rdr.readSliceEndian(Header, @ptrCast(&hdr), .little);
-        cur_red += @sizeOf(Header);
-        try out.ensureUnusedCapacity(alloc, hdr.count + 1);
-        for (0..hdr.count + 1) |_| {
-            try rdr.readSliceEndian(RawEntry, @ptrCast(&raw), .little);
-            const name = try alloc.alloc(u8, raw.name_size + 1);
-            errdefer alloc.free(name);
-            try rdr.readSliceEndian(u8, name, .little);
-            const val = out.addOneAssumeCapacity();
-            val.* = .{
-                .block_start = hdr.block_start,
-                .block_offset = raw.offset,
-                .num = @abs(hdr.num + raw.offset),
-                .inode_type = raw.inode_type,
-                .name = name,
-            };
-            cur_red += @sizeOf(RawEntry) + raw.name_size + 1;
-        }
-    }
-    return out.toOwnedSlice(alloc);
-}
-
-pub fn deinit(self: Entry, alloc: std.mem.Allocator) void {
-    alloc.free(self.name);
-}

src/directory.zig

@@ -0,0 +1,63 @@
+const std = @import("std");
+const Reader = std.Io.Reader;
+
+const InodeType = @import("inode.zig").Type;
+
+pub fn readDirectory(alloc: std.mem.Allocator, rdr: *Reader, size: u32) []Entry {
+    var read: u32 = 3;
+    var hdr: Header = undefined;
+    var raw: RawEntry = undefined;
+    var out: std.ArrayList(Entry) = .initCapacity(alloc, 50);
+    errdefer {
+        for (out.items) |i|
+            alloc.free(i.name);
+        out.deinit(alloc);
+    }
+    while (read < size) {
+        try rdr.readSliceEndian(Header, @ptrCast(&hdr), .little);
+        try out.ensureUnusedCapacity(alloc, hdr.count + 1);
+        read += @sizeOf(Header);
+        for (0..hdr.count + 1) |_| {
+            try rdr.readSliceEndian(RawEntry, @ptrCast(&raw), .little);
+            read += @sizeOf(RawEntry) + raw.name_size + 1;
+            const new = out.addOneAssumeCapacity();
+            new.* = .{
+                .block_start = hdr.block_start,
+                .block_offset = raw.block_offset,
+                .num = @abs(hdr.num + raw.num_offset),
+                .inode_type = raw.inode_type,
+                .name = try alloc.alloc(u8, raw.name_size + 1),
+            };
+            try rdr.readSliceEndian(u8, new.name, .little);
+        }
+    }
+    return out.toOwnedSlice(alloc);
+}
+
+// Types
+
+pub const Entry = struct {
+    block_start: u32,
+    block_offset: u16,
+    num: u32,
+    inode_type: InodeType,
+    name: []u8,
+
+    pub fn deinit(self: Entry, alloc: std.mem.Allocator) void {
+        alloc.free(self.name);
+    }
+};
+
+// extern instead of packed due to alignment issues (packed will read it as 16 bytes instead of 12).
+const Header = extern struct {
+    count: u32,
+    block_start: u32,
+    num: u32,
+};
+
+const RawEntry = packed struct {
+    block_offset: u16,
+    num_offset: i16,
+    inode_type: InodeType,
+    name_size: u16,
+};

src/file.zig

@@ -1,230 +1,118 @@
-//! A file/directory within the squashfs archive.
-
 const std = @import("std");
-const File = std.fs.File;
-const WaitGroup = std.Thread.WaitGroup;
-const Mutex = std.Thread.Mutex;
+const Io = std.Io;
 
 const Archive = @import("archive.zig");
-const DirEntry = @import("dir_entry.zig");
+const Decompressor = @import("decomp.zig");
+const Directory = @import("directory.zig");
 const ExtractionOptions = @import("options.zig");
 const Inode = @import("inode.zig");
-const BlockSize = @import("inode_data/file.zig").BlockSize;
-const DataReader = @import("util/data.zig");
+const DataReader = @import("util/data_reader.zig");
+const FileIter = @import("util/iter.zig");
 const MetadataReader = @import("util/metadata.zig");
+const OffsetFile = @import("util/offset_file.zig");
+const Utils = @import("util/utils.zig");
 
-const FileError = error{
+pub const Error = error{
     NotDirectory,
     NotRegularFile,
-    NotSymlink,
-    NotDevice,
-    NotFound,
-    ExtractionPathExists,
 };
 
-const SfsFile = @This();
+const File = @This();
 
-archive: *Archive,
+file: OffsetFile,
+super: Archive.MinimalSuperblock,
+decomp: Decompressor,
 
-inode: Inode,
 name: []const u8,
+inode: Inode,
 
-/// Initialize a new File.
-/// name is copied to the File so can be safely freed afterwards.
-pub fn init(archive: *Archive, inode: Inode, name: []const u8) !SfsFile {
-    const new_name = try archive.allocator().alloc(u8, name.len);
-    @memcpy(new_name, name);
+pub fn init(alloc: std.mem.Allocator, archive: Archive, entry: Directory.Entry) !File {
+    const new_name = try alloc.alloc(u8, entry.name.len);
+    errdefer alloc.free(new_name);
+    @memcpy(new_name, entry.name);
     return .{
-        .archive = archive,
-        .inode = inode,
+        .file = archive.file,
+        .super = archive.super,
+        .decomp = archive.stateless_decomp.statelessCopy(alloc),
         .name = new_name,
+        .inode = try Utils.readInode(
+            alloc,
+            &archive.decomp,
+            archive.file,
+            archive.super.inode_start,
+            archive.super.block_size,
+            entry.block_start,
+            entry.block_offset,
+        ),
     };
 }
-pub fn fromEntry(archive: *Archive, entry: DirEntry) !SfsFile {
-    var rdr = try archive.fil.readerAt(entry.block_start + archive.super.inode_start, &[0]u8{});
-    var meta: MetadataReader = .init(archive.allocator(), &rdr.interface, archive.decomp);
-    try meta.interface.discardAll(entry.block_offset);
-    const inode: Inode = try .read(archive.allocator(), &meta.interface, archive.super.block_size);
-    errdefer inode.deinit(archive.allocator());
-    return .init(archive, inode, entry.name);
+pub fn deinit(self: File) void {
+    self.decomp.alloc.free(self.name);
+    self.inode.deinit(self.decomp.alloc);
 }
 
-pub fn deinit(self: SfsFile) void {
-    var alloc = self.archive.allocator();
-    alloc.free(self.name);
-    self.inode.deinit(alloc);
-}
+// Directory functions
 
-fn getEntries(self: SfsFile) ![]DirEntry {
-    return self.inode.dirEntries(self.archive.allocator(), self.archive.*);
-}
-
-pub fn ownerUid(self: SfsFile) !u16 {
-    return self.archive.id(self.inode.hdr.uid_idx);
-}
-pub fn ownerGid(self: SfsFile) !u16 {
-    return self.archive.id(self.inode.hdr.gid_idx);
-}
-pub fn permissions(self: SfsFile) u16 {
-    return self.inode.hdr.permissions;
-}
-
-pub fn isRegular(self: SfsFile) bool {
-    return switch (self.inode.hdr.inode_type) {
-        .file, .ext_file => true,
-        else => false,
-    };
-}
-/// The returned DataReader will no longer work if the File's deinit function is called
-/// or, more specifically, it's inode's deinit function is called.
-pub fn dataReader(self: SfsFile) !DataReader {
-    return self.inode.dataReader(self.archive);
-}
-
-pub fn isDir(self: SfsFile) bool {
+pub fn isDir(self: File) bool {
     return switch (self.inode.hdr.inode_type) {
         .dir, .ext_dir => true,
         else => false,
     };
 }
-pub fn iterate(self: SfsFile) !Iterator {
-    if (!self.isDir()) return FileError.NotDirectory;
+/// Opens a sub-file. If the given path is "" or "." (after trimming /) a copy of the File is returned.
+pub fn open(self: File, alloc: std.mem.Allocator, filepath: []const u8) !File {
+    var res = try self.inode.findInode(
+        alloc,
+        &self.decomp,
+        self.file,
+        self.super.dir_start,
+        self.super.inode_start,
+        self.super.block_size,
+        filepath,
+    );
+    if (res.name.len == 0) {
+        res.name = try alloc.alloc(u8, self.name.len);
+        @memcpy(res.name, self.name);
+    }
     return .{
-        .entries = try self.getEntries(),
-        .archive = self.archive,
+        .file = self.file,
+        .super = self.super,
+        .decomp = self.decomp.statelessCopy(alloc),
+        .name = res.name,
+        .inode = res.inode,
     };
 }
-/// Open a sub-file/folder within a directory at the given path.
-/// If path is "", ".", "/", or "./", this File is returned.
-pub fn open(self: SfsFile, path: []const u8) !SfsFile {
-    if (!self.isDir()) return FileError.NotDirectory;
-    if (pathIsSelf(path)) return self;
-
-    // Recursively stip ending & leading path separators.
-    if (path[0] == '/') return self.open(path[1..]);
-    if (path[path.len - 1] == '/') return self.open(path[0 .. path.len - 1]);
-
-    const idx = std.mem.indexOf(u8, path, "/") orelse path.len;
-    const first_element = path[0..idx];
-    if (std.mem.eql(u8, first_element, ".")) return self.open(path[idx + 1 ..]);
-    const entries = try self.getEntries();
-    defer {
-        var alloc = self.archive.allocator();
-        for (entries) |e| {
-            e.deinit(alloc);
-        }
-        alloc.free(entries);
-    }
-    var cur_slice = entries;
-    var split = cur_slice.len / 2;
-    while (cur_slice.len > 0) {
-        split = cur_slice.len / 2;
-        const comp = std.mem.order(u8, first_element, cur_slice[split].name);
-        switch (comp) {
-            .eq => {
-                var fil: SfsFile = try .fromEntry(self.archive, cur_slice[split]);
-                if (idx == path.len) {
-                    return fil;
-                }
-                defer fil.deinit();
-                return fil.open(path[idx + 1 ..]);
-            },
-            .lt => cur_slice = cur_slice[0..split],
-            .gt => cur_slice = cur_slice[split + 1 ..],
-        }
-    }
-    return FileError.NotFound;
+pub fn iter(self: File, alloc: std.mem.Allocator) !FileIter {
+    return .{
+        .alloc = alloc,
+        .entries = try self.inode.readDirectory(alloc, &self.decomp, self.file, self.super.dir_start),
+    };
 }
 
-pub fn isSymlink(self: SfsFile) bool {
+// Regular file functions
+
+pub fn isRegularFile(self: File) bool {
     return switch (self.inode.hdr.inode_type) {
-        .symlink, .ext_symlink => true,
+        .file, .ext_file => true,
         else => false,
     };
 }
-pub fn symlinkPath(self: SfsFile) ![]const u8 {
-    if (!self.isSymlink()) return FileError.NotSymlink;
-    return switch (self.inode.data) {
-        .symlink => |s| s.target,
-        .ext_symlink => |s| s.target,
-        else => unreachable,
-    };
+// a std.Io.Reader compatible reader that reads a regular file's data.
+pub fn dataReader(self: File, alloc: std.mem.Allocator) !DataReader {
+    return self.inode.dataReader(
+        &self.decomp.statelessCopy(alloc),
+        self.file,
+        self.super.frag_start,
+        self.super.block_size,
+    );
 }
 
-/// Check if the File is a block or character device.
-pub fn isDevice(self: SfsFile) bool {
-    return switch (self.inode.hdr.inode_type) {
-        .block_dev, .char_dev, .ext_block_dev, .ext_char_dev => true,
-        else => false,
-    };
+// Universal functions
+
+pub fn extract(self: File, alloc: std.mem.Allocator, path: []const u8, options: ExtractionOptions) !void {
+    _ = self;
+    _ = alloc;
+    _ = path;
+    _ = options;
+    return error.TODO;
 }
-/// If the File is a block or character device, get's it's device number.
-pub fn devNum(self: SfsFile) !u32 {
-    if (!self.isDevice()) return FileError.NotDevice;
-    return switch (self.inode.data) {
-        .block_dev, .char_dev => |d| d.dev,
-        .ext_block_dev, .ext_char_dev => |d| d.dev,
-        else => unreachable,
-    };
-}
-
-/// Extract the given File to the path. If File is a regular file, the path must be a directory or not exist.
-/// If the gievn path is a folder, the File's contents will be extracted within.
-pub fn extract(self: *SfsFile, path: []const u8, options: ExtractionOptions) !void {
-    var alloc = self.archive.allocator();
-    var ext_path: []u8 = undefined;
-    if (std.fs.cwd().statFile(path)) |stat| {
-        if (stat.kind == .directory) {
-            if (!self.isDir()) {
-                const has_end_sep = path[path.len - 1] == '/';
-                const alloc_size = if (has_end_sep)
-                    path.len + self.name.len
-                else
-                    path.len + self.name.len + 1;
-                ext_path = try alloc.alloc(u8, alloc_size);
-                @memcpy(ext_path[0..path.len], path);
-                @memcpy(ext_path[ext_path.len - self.name.len ..], self.name);
-                if (!has_end_sep) ext_path[path.len] = '/';
-            } else {
-                ext_path = @constCast(path);
-            }
-        } else return FileError.ExtractionPathExists;
-    } else |err| {
-        if (err == error.FileNotFound) {
-            ext_path = @constCast(path);
-        } else {
-            std.log.err("Error stat-ing extraction path {s}: {}\n", .{ path, err });
-            return err;
-        }
-    }
-    defer if (ext_path.len > path.len) alloc.free(ext_path);
-    return self.inode.extractToThreaded(self.archive, path, options, self.archive.thread_count);
-}
-
-/// Utility function.
-pub fn pathIsSelf(path: []const u8) bool {
-    if (path.len == 0) return true;
-    if (path.len == 1 and (path[0] == '/' or path[0] == '.')) return true;
-    if (path.len == 2 and (path[0] == '.' and path[1] == '/')) return true;
-    return false;
-}
-
-pub const Iterator = struct {
-    entries: []DirEntry,
-    archive: *Archive,
-
-    idx: u32 = 0,
-
-    pub fn next(self: *Iterator) !?SfsFile {
-        if (self.idx >= self.entries.len) return null;
-        defer self.idx += 1;
-        return try SfsFile.fromEntry(self.archive, self.entries[self.idx]);
-    }
-    pub fn deinit(self: Iterator) void {
-        var alloc = self.archive.allocator();
-        for (self.entries) |e| {
-            e.deinit(alloc);
-        }
-        alloc.free(self.entries);
-    }
-};

src/imports.c

@@ -0,0 +1,5 @@
+#include <zlib-ng.h>
+#include <lzo/minilzo.h>
+#include <lz4.h>
+#include <zstd.h>
+#include <lzma.h>

src/inode.zig

@@ -1,73 +1,27 @@
-//! A file-system object. Represents a File or directory.
+//! This is the raw squashfs representation of a file/directory.
+//! Most of the time using File is a better experience and using Inodes directory
+//! is only required for more technical use cases.
 
 const std = @import("std");
 const Reader = std.Io.Reader;
-const WaitGroup = std.Thread.WaitGroup;
-const Pool = std.Thread.Pool;
 
-const Archive = @import("archive.zig");
-const DirEntry = @import("dir_entry.zig");
-const ExtractionOptions = @import("options.zig");
-const dir = @import("inode_data/dir.zig");
-const file = @import("inode_data/file.zig");
-const misc = @import("inode_data/misc.zig");
-const DataReader = @import("util/data.zig");
-const ThreadedDataReader = @import("util/data_threaded.zig");
+const Decompressor = @import("decomp.zig");
+const Directory = @import("directory.zig");
+const FragEntry = @import("archive.zig").FragEntry;
+const Dir = @import("inode/dir.zig");
+const File = @import("inode/file.zig");
+const Misc = @import("inode/misc.zig");
+const Sym = @import("inode/sym.zig");
+const LookupTable = @import("lookup_table.zig");
+const MinimalSuperblock = @import("archive.zig").MinimalSuperblock;
+const DataReader = @import("util/data_reader.zig");
 const MetadataReader = @import("util/metadata.zig");
-
-pub const Ref = packed struct {
-    block_offset: u16,
-    block_start: u32,
-    _: u16,
-};
-
-pub const InodeType = enum(u16) {
-    dir = 1,
-    file,
-    symlink,
-    block_dev,
-    char_dev,
-    fifo,
-    socket,
-    ext_dir,
-    ext_file,
-    ext_symlink,
-    ext_block_dev,
-    ext_char_dev,
-    ext_fifo,
-    ext_socket,
-};
-
-pub const InodeData = union(InodeType) {
-    dir: dir.Dir,
-    file: file.File,
-    symlink: misc.Symlink,
-    block_dev: misc.Dev,
-    char_dev: misc.Dev,
-    fifo: misc.IPC,
-    socket: misc.IPC,
-    ext_dir: dir.ExtDir,
-    ext_file: file.ExtFile,
-    ext_symlink: misc.ExtSymlink,
-    ext_block_dev: misc.ExtDev,
-    ext_char_dev: misc.ExtDev,
-    ext_fifo: misc.ExtIPC,
-    ext_socket: misc.ExtIPC,
-};
-
-pub const Header = packed struct {
-    inode_type: InodeType,
-    permissions: u16,
-    uid_idx: u16,
-    gid_idx: u16,
-    mod_time: u32,
-    num: u32,
-};
+const OffsetFile = @import("util/offset_file.zig");
 
 const Inode = @This();
 
 hdr: Header,
-data: InodeData,
+data: Data,
 
 pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !Inode {
     var hdr: Header = undefined;
@@ -75,30 +29,23 @@ pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !Inode {
     return .{
         .hdr = hdr,
         .data = switch (hdr.inode_type) {
-            .dir => .{ .dir = try .read(rdr) },
-            .file => .{ .file = try .read(alloc, rdr, block_size) },
-            .symlink => .{ .symlink = try .read(alloc, rdr) },
-            .block_dev => .{ .block_dev = try .read(rdr) },
-            .char_dev => .{ .char_dev = try .read(rdr) },
-            .fifo => .{ .fifo = try .read(rdr) },
-            .socket => .{ .socket = try .read(rdr) },
-            .ext_dir => .{ .ext_dir = try .read(rdr) },
-            .ext_file => .{ .ext_file = try .read(alloc, rdr, block_size) },
-            .ext_symlink => .{ .ext_symlink = try .read(alloc, rdr) },
-            .ext_block_dev => .{ .ext_block_dev = try .read(rdr) },
-            .ext_char_dev => .{ .ext_char_dev = try .read(rdr) },
-            .ext_fifo => .{ .ext_fifo = try .read(rdr) },
-            .ext_socket => .{ .ext_socket = try .read(rdr) },
+            .dir => .{ .dir = .read(rdr) },
+            .file => .{ .file = .read(alloc, rdr, block_size) },
+            .symlink => .{ .symlink = .read(alloc, rdr) },
+            .block => .{ .block = .read(rdr) },
+            .char => .{ .char = .read(rdr) },
+            .fifo => .{ .fifo = .read(rdr) },
+            .sock => .{ .sock = .read(rdr) },
+            .ext_dir => .{ .ext_dir = .read(rdr) },
+            .ext_file => .{ .ext_file = .read(alloc, rdr, block_size) },
+            .ext_symlink => .{ .ext_symlink = .read(alloc, rdr) },
+            .ext_block => .{ .ext_block = .read(rdr) },
+            .ext_char => .{ .ext_char = .read(rdr) },
+            .ext_fifo => .{ .ext_fifo = .read(rdr) },
+            .ext_sock => .{ .ext_sock = .read(rdr) },
         },
     };
 }
-pub fn readFromEntry(alloc: std.mem.Allocator, archive: *Archive, entry: DirEntry) !Inode {
-    var rdr = try archive.fil.readerAt(archive.super.inode_start + entry.block_start, &[0]u8{});
-    var meta: MetadataReader = .init(alloc, &rdr.interface, archive.decomp);
-    try meta.interface.discardAll(entry.block_offset);
-    return read(alloc, &meta.interface, archive.super.block_size);
-}
-
 pub fn deinit(self: Inode, alloc: std.mem.Allocator) void {
     switch (self.data) {
         .file => |f| alloc.free(f.block_sizes),
@@ -108,371 +55,301 @@ pub fn deinit(self: Inode, alloc: std.mem.Allocator) void {
         else => {},
     }
 }
+pub fn copy(self: Inode, alloc: std.mem.Allocator) !Inode {
+    switch (self.data) {
+        .dir,
+        .ext_dir,
+        .block,
+        .ext_block,
+        .char,
+        .ext_char,
+        .fifo,
+        .ext_fifo,
+        .sock,
+        .ext_sock,
+        => return self,
+        .file => |f| {
+            const new_sizes = try alloc.alloc(File.BlockSize, f.block_sizes.len);
+            @memcpy(new_sizes, f.block_sizes);
+            return .{
+                .hdr = self.hdr,
+                .data = .{ .file = .{
+                    .block_start = f.block_start,
+                    .frag_idx = f.frag_idx,
+                    .block_offset = f.block_offset,
+                    .size = f.size,
+                    .block_sizes = new_sizes,
+                } },
+            };
+        },
+        .ext_file => |f| {
+            const new_sizes = try alloc.alloc(File.BlockSize, f.block_sizes.len);
+            @memcpy(new_sizes, f.block_sizes);
+            return .{
+                .hdr = self.hdr,
+                .data = .{ .ext_file = .{
+                    .block_start = self.block_start,
+                    .size = self.size,
+                    .sparse = self.sparse,
+                    .hard_links = self.hard_links,
+                    .frag_idx = self.frag_idx,
+                    .block_offset = self.block_offset,
+                    .xattr_idx = self.xattr_idx,
+                    .block_sizes = new_sizes,
+                } },
+            };
+        },
+        .symlink => |s| {
+            const new_target = try alloc.alloc(u8, s.target.len);
+            @memcpy(new_target, s.target);
+            return .{
+                .hdr = self.hdr,
+                .data = .{ .symlink = .{
+                    .hard_links = s.hard_links,
+                    .target = new_target,
+                } },
+            };
+        },
+        .ext_symlink => |s| {
+            const new_target = try alloc.alloc(u8, s.target.len);
+            @memcpy(new_target, s.target);
+            return .{
+                .hdr = self.hdr,
+                .data = .{ .ext_symlink = .{
+                    .hard_links = s.hard_links,
+                    .xattr_idx = s.xattr_idx,
+                    .target = new_target,
+                } },
+            };
+        },
+    }
+}
 
-/// Get the data reader for a file inode.
-pub fn dataReader(self: Inode, alloc: std.mem.Allocator, archive: *Archive) !DataReader {
-    return switch (self.hdr.inode_type) {
-        .file => readerFromData(alloc, archive, self.data.file),
-        .ext_file => readerFromData(alloc, archive, self.data.ext_file),
-        else => error.NotRegularFile,
+// Types
+
+pub const Ref = packed struct {
+    block_offset: u16,
+    block_start: u32,
+    _: u16,
+};
+
+pub const Type = enum(u16) {
+    dir = 1,
+    file,
+    symlink,
+    block,
+    char,
+    fifo,
+    sock,
+    ext_dir,
+    ext_file,
+    ext_symlink,
+    ext_block,
+    ext_char,
+    ext_fifo,
+    ext_sock,
+};
+
+const Header = packed struct {
+    inode_type: Type,
+    permission: u16,
+    uid_idx: u16,
+    gid_idx: u16,
+    mod_time: u32,
+    num: u32,
+};
+
+pub const Data = union(Type) {
+    dir: Dir.Dir,
+    file: File.File,
+    symlink: Sym.Symlink,
+    block: Misc.Device,
+    char: Misc.Device,
+    fifo: Misc.Ipc,
+    sock: Misc.Ipc,
+    ext_dir: Dir.ExtDir,
+    ext_file: File.ExtFile,
+    ext_symlink: Sym.ExtSymlink,
+    ext_block: Misc.ExtDevice,
+    ext_char: Misc.ExtDevice,
+    ext_fifo: Misc.ExtIpc,
+    ext_sock: Misc.ExtIpc,
+};
+
+// Errors
+
+pub const Error = error{
+    NotDirectory,
+    NotFound,
+    NotRegularFile,
+};
+
+// Utils functions
+
+// Universal
+
+pub fn uid(self: Inode, decomp: *const Decompressor, fil: OffsetFile, id_start: u64) !u16 {
+    return LookupTable.stateless(u16, fil, decomp, id_start, self.hdr.uid_idx);
+}
+pub fn uidCached(self: Inode, table: LookupTable.CachedTable(u16)) !u16 {
+    return table.get(self.hdr.uid_idx);
+}
+pub fn gid(self: Inode, decomp: *const Decompressor, fil: OffsetFile, id_start: u64) !u16 {
+    return LookupTable.stateless(u16, fil, decomp, id_start, self.hdr.gid_idx);
+}
+pub fn gidCached(self: Inode, table: LookupTable.CachedTable(u16)) !u16 {
+    return table.get(self.hdr.gid_idx);
+}
+pub fn xattr(self: Inode, alloc: std.mem.Allocator, decomp: *const Decompressor, fil: OffsetFile, xattr_start: u64) !?LookupTable.XattrValues {
+    if (@intFromEnum(self.hdr.inode_type) < 8) return null;
+    const idx: u32 = switch (self.data) {
+        .ext_dir => |d| d.xattr_idx,
+        .ext_file => |f| f.xattr_idx,
+        .ext_symlink => |s| s.xattr_idx,
+        .ext_block, .ext_char => |d| d.xattr_idx,
+        .ext_fifo, .ext_sock => |d| d.xattr_idx,
+        else => unreachable,
     };
+    if (idx == 0xFFFFFFFF) return null;
+    return LookupTable.statelessXattr(alloc, fil, decomp, xattr_start, idx);
 }
-fn readerFromData(alloc: std.mem.Allocator, archive: *Archive, data: anytype) !DataReader {
-    var out: DataReader = .init(alloc, archive.*, data.block_sizes, data.block_start, data.size);
-    if (data.frag_idx != 0xFFFFFFFF)
-        out.addFragment(try archive.frag(data.frag_idx), data.frag_block_offset);
-    return out;
+
+// Dir inodes
+
+/// For directory inodes, tries to find the inode at the given path. Returns both the inode, and it's file name.
+/// If the path is empty or "." then a copy of this inode is returned with no name ("").
+pub fn findInode(
+    inode: Inode,
+    alloc: std.mem.Allocator,
+    decomp: *const Decompressor,
+    fil: OffsetFile,
+    dir_start: u64,
+    inode_start: u64,
+    block_size: u32,
+    filepath: []const u8,
+) !struct { inode: Inode, name: []const u8 } {
+    switch (inode.data) {
+        .dir => |d| {
+            const path: []const u8 = std.mem.trim(u8, filepath, "/");
+            if (path.len == 0 or (path.len == 1 and path[0] == '.'))
+                return .{ .inode = inode.copy(alloc), .name = "" };
+            return findInodeRaw(
+                alloc,
+                decomp,
+                fil,
+                dir_start,
+                inode_start,
+                block_size,
+                path,
+                d,
+            );
+        },
+        .ext_dir => |d| {
+            const path: []const u8 = std.mem.trim(u8, filepath, "/");
+            if (path.len == 0 or (path.len == 1 and path[0] == '.'))
+                return .{ .inode = inode.copy(alloc), .name = "" };
+            return findInodeRaw(
+                alloc,
+                decomp,
+                fil,
+                dir_start,
+                inode_start,
+                block_size,
+                path,
+                d,
+            );
+        },
+        else => return Error.NotDirectory,
+    }
 }
-/// Get a threaded data reader for a file inode.
-pub fn threadedDataReader(self: Inode, alloc: std.mem.Allocator, archive: *Archive) !ThreadedDataReader {
-    return switch (self.hdr.inode_type) {
-        .file => threadedReaderFromData(alloc, archive, self.data.file),
-        .ext_file => threadedReaderFromData(alloc, archive, self.data.ext_file),
-        else => error.NotRegularFile,
-    };
-}
-fn threadedReaderFromData(alloc: std.mem.Allocator, archive: *Archive, data: anytype) !ThreadedDataReader {
-    var out: ThreadedDataReader = .init(alloc, archive.*, data.block_sizes, data.block_start, data.size);
-    if (data.frag_idx != 0xFFFFFFFF)
-        out.addFragment(try archive.frag(data.frag_idx), data.frag_block_offset);
-    return out;
+inline fn findInodeRaw(
+    inode: Inode,
+    alloc: std.mem.Allocator,
+    decomp: *const Decompressor,
+    fil: OffsetFile,
+    dir_start: u64,
+    inode_start: u64,
+    block_size: u32,
+    path: []const u8,
+    dat: anytype,
+) !struct { inode: Inode, name: []const u8 } {
+    const first_element: []const u8 = std.mem.sliceTo(path, '/');
+
+    const dirs = try readDirRaw(alloc, decomp, fil, dir_start, dat);
+    defer {
+        for (dirs) |dir|
+            dir.deinit(alloc);
+        alloc.free(dirs);
+    }
+
+    // Directories are stored ASCIIbetically, so we can use binary search.
+    var cur_slice = dirs;
+    var idx: usize = 0;
+    while (cur_slice.len > 0) {
+        idx = cur_slice.len / 2;
+        const mid_name = cur_slice[idx].name;
+        switch (std.mem.order(u8, first_element, mid_name)) {
+            .gt => {
+                cur_slice = cur_slice[idx + 1 ..];
+                continue;
+            },
+            .lt => {
+                cur_slice = cur_slice[0..idx];
+                continue;
+            },
+            .eq => break,
+        }
+    } else return Error.NotFound;
+    const entry = cur_slice[idx];
+    var rdr = try fil.readerAt(inode_start + entry.block_start, &[0]u8{});
+    var meta_rdr: MetadataReader = .init(&rdr.interface, decomp);
+    try meta_rdr.interface.discardAll(entry.block_offset);
+    const ret_inode: Inode = try .read(alloc, &meta_rdr.interface, block_size);
+    if (first_element.len == path.len) {
+        const name_copy = try alloc.alloc(u8, entry.name.len);
+        @memcpy(name_copy, entry.name.len);
+        return .{ .inode = ret_inode, .name = name_copy };
+    }
+    return inode.findInode(alloc, decomp, fil, dir_start, inode_start, block_size, path[first_element.len..]);
 }
 
 /// Get the directory entries for a directory inode.
-pub fn dirEntries(self: Inode, alloc: std.mem.Allocator, archive: Archive) ![]DirEntry {
-    return switch (self.hdr.inode_type) {
-        .dir => entriesFromData(alloc, archive, self.data.dir),
-        .ext_dir => entriesFromData(alloc, archive, self.data.ext_dir),
-        else => error.NotDirectory,
+pub fn readDirectory(inode: Inode, alloc: std.mem.Allocator, decomp: *const Decompressor, fil: OffsetFile, dir_start: u64) ![]Directory.Entry {
+    return switch (inode.data) {
+        .dir => |d| readDirRaw(alloc, decomp, fil, dir_start, d),
+        .ext_dir => |d| readDirRaw(alloc, decomp, fil, dir_start, d),
+        else => Error.NotDirectory,
     };
 }
-fn entriesFromData(alloc: std.mem.Allocator, archive: Archive, data: anytype) ![]DirEntry {
-    var rdr = try archive.fil.readerAt(archive.super.dir_start + data.block_start, &[0]u8{});
-    var meta: MetadataReader = .init(alloc, &rdr.interface, archive.decomp);
-    try meta.interface.discardAll(data.block_offset);
-    return DirEntry.readDir(alloc, &meta.interface, data.size);
+inline fn readDirRaw(alloc: std.mem.Allocator, decomp: *const Decompressor, fil: OffsetFile, dir_start: u64, dat: anytype) ![]Directory.Entry {
+    var rdr = try fil.readerAt(dir_start + dat.block_start, &[0]u8{});
+    var meta_rdr: MetadataReader = .init(&rdr.interface, decomp);
+    try meta_rdr.interface.discardAll(dat.block_offset);
+    return Directory.readDirectory(alloc, meta_rdr, dat.size);
 }
 
-/// Extract the inode to the given path. Single threaded.
-pub fn extractTo(self: Inode, archive: *Archive, path: []const u8, options: ExtractionOptions) !void {
-    switch (self.hdr.inode_type) {
-        .dir, .ext_dir => {
-            // Removing any trailing separators since that's the easiest path forward.
-            if (path[path.len - 1] == '/') return self.extractTo(archive, path[0 .. path.len - 1], options);
-            std.fs.cwd().makeDir(path) catch |err| {
-                if (err != std.fs.Dir.MakeError.PathAlreadyExists) return err;
-            };
-            var alloc = archive.allocator();
-            const entries = try self.dirEntries(alloc, archive.*);
-            defer {
-                for (entries) |entry| entry.deinit(alloc);
-                alloc.free(entries);
-            }
-            for (entries) |entry| {
-                var new_path = try alloc.alloc(u8, path.len + 1 + entry.name.len);
-                @memcpy(new_path[0..path.len], path);
-                @memcpy(new_path[path.len + 1 ..], entry.name);
-                new_path[path.len] = '/';
-                defer alloc.free(new_path);
+// file inodes
 
-                var inode: Inode = try readFromEntry(alloc, archive, entry);
-                defer inode.deinit(alloc);
-                try inode.extractTo(archive, new_path, options);
-            }
-        },
-        .file, .ext_file => try self.extractRegFile(archive.allocator(), archive, path, options),
-        .symlink, .ext_symlink => try self.extractSymlink(path),
-        else => try self.extractDevice(archive, path, options),
-    }
+/// Gets the data reader for a file inode.
+pub fn dataReader(inode: Inode, decomp: *const Decompressor, fil: OffsetFile, frag_start: u64, block_size: u32) !DataReader {
+    return switch (inode.data) {
+        .file => |f| dataReaderRaw(decomp, fil, frag_start, block_size, f),
+        .ext_file => |f| dataReaderRaw(decomp, fil, frag_start, block_size, f),
+        else => Error.NotRegularFile,
+    };
 }
-
-const Perms = struct {
-    path: []const u8,
-    uid: u16,
-    gid: u16,
-    perm: u16,
-};
-
-/// Extract the inode to the given path. Multi-threaded.
-/// Functions identically to extractTo on all but regular files and directories.
-///
-/// If threads <= 1, then this just calls extractTo.
-pub fn extractToThreaded(self: Inode, archive: *Archive, path: []const u8, options: ExtractionOptions, threads: usize) !void {
-    if (threads <= 1) return self.extractTo(archive, path, options);
-    switch (self.hdr.inode_type) {
-        .dir, .ext_dir => {
-            // Removing any trailing separators since that's the easiest path forward.
-            if (path[path.len - 1] == '/') return self.extractToThreaded(archive, path[0 .. path.len - 1], options, threads);
-
-            // Fixed Allocator
-            // const mem_buf = archive.allocator().alloc(u8, 2 * 1024 * 1024 * 1024);
-            // defer archive.allocator().free(mem_buf);
-            // var fixed_alloc: std.heap.FixedBufferAllocator = .init(mem_buf);
-            // const alloc = fixed_alloc.threadSafeAllocator();
-
-            // Arena Allocator
-            var arena_alloc: std.heap.ArenaAllocator = .init(archive.allocator());
-            defer arena_alloc.deinit();
-            var thread_alloc: std.heap.ThreadSafeAllocator = .{ .child_allocator = arena_alloc.allocator() };
-            const alloc = thread_alloc.allocator();
-
-            var wg: WaitGroup = .{};
-            var perms: ?std.ArrayList(Perms) = if (options.ignore_permissions) null else try .initCapacity(alloc, 100);
-            // defer if(!options.ignore_permissions) perms.?.deinit(alloc); We don't need to do this due to ArenaAllocator
-            var pool: Pool = undefined;
-            try pool.init(.{ .allocator = alloc, .n_jobs = threads - 1 });
-            defer pool.deinit();
-            var out_err: ?anyerror = null;
-
-            wg.start();
-            self.extractThread(alloc, archive, path, options, &wg, &pool, &out_err, &perms);
-            pool.waitAndWork(&wg);
-            if (out_err != null) return out_err.?;
-
-            if (perms != null) {
-                for (perms.?.items) |p| {
-                    var fil = try std.fs.cwd().openFile(p.path, .{});
-                    try fil.chmod(p.perm);
-                    try fil.chown(p.uid, p.gid);
-                }
-            }
-        },
-        .file, .ext_file => {
-            const alloc = archive.allocator();
-
-            var pool: Pool = undefined;
-            try pool.init(.{ .allocator = alloc, .n_jobs = threads });
-            defer pool.deinit();
-
-            try self.extractRegFileThreaded(alloc, archive, path, options, &pool);
-
-            if (!options.ignore_permissions) {
-                var fil = try std.fs.cwd().openFile(path, .{});
-                try fil.chmod(self.hdr.permissions);
-                try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
-            }
-        },
-        .symlink, .ext_symlink => try self.extractSymlink(path),
-        else => try self.extractDevice(archive, path, options),
-    }
-}
-
-fn extractThreadEntry(
-    entry: DirEntry,
-    alloc: std.mem.Allocator,
-    archive: *Archive,
-    path: []const u8,
-    options: ExtractionOptions,
-    wg: *WaitGroup,
-    pool: *Pool,
-    out_err: *?anyerror,
-    perms: *?std.ArrayList(Perms),
-) void {
-    var new_path = alloc.alloc(u8, path.len + entry.name.len + 1) catch |err| {
-        wg.finish();
-        out_err.* = err;
-        return;
-    };
-    @memcpy(new_path[0..path.len], path);
-    @memcpy(new_path[path.len + 1 ..], entry.name);
-    new_path[path.len] = '/';
-    var inode = readFromEntry(alloc, archive, entry) catch |err| {
-        out_err.* = err;
-        wg.finish();
-        return;
-    };
-    inode.extractThread(alloc, archive, new_path, options, wg, pool, out_err, perms);
-}
-
-/// Extract threadedly the inode to the path.
-fn extractThread(
-    self: Inode,
-    alloc: std.mem.Allocator,
-    archive: *Archive,
-    path: []const u8,
-    options: ExtractionOptions,
-    wg: *WaitGroup,
-    pool: *Pool,
-    out_err: *?anyerror,
-    perms: *?std.ArrayList(Perms),
-) void {
-    defer wg.finish();
-    if (out_err.* != null) return;
-    switch (self.hdr.inode_type) {
-        .dir, .ext_dir => {
-            std.fs.cwd().makeDir(path) catch |err| {
-                if (err != std.fs.Dir.MakeError.PathAlreadyExists) {
-                    out_err.* = err;
-                    return;
-                }
-            };
-
-            const entries = self.dirEntries(alloc, archive.*) catch |err| {
-                out_err.* = err;
-                return;
-            };
-            wg.startMany(entries.len);
-            // defer files.deinit(alloc); We don't need to do this due to ArenaAllocator
-            for (entries) |entry| {
-                if (entry.inode_type == .dir) {
-                    extractThreadEntry(entry, alloc, archive, path, options, wg, pool, out_err, perms);
-                    continue;
-                }
-                pool.spawn(
-                    extractThreadEntry,
-                    .{
-                        entry,
-                        alloc,
-                        archive,
-                        path,
-                        options,
-                        wg,
-                        pool,
-                        out_err,
-                        perms,
-                    },
-                ) catch |err| {
-                    wg.finish();
-                    out_err.* = err;
-                    continue;
-                };
-            }
-            if (!options.ignore_permissions) {
-                const new_val = perms.*.?.addOne(alloc) catch |err| {
-                    out_err.* = err;
-                    return;
-                };
-                new_val.* = .{
-                    .path = path,
-                    .uid = archive.id(self.hdr.uid_idx) catch |err| {
-                        out_err.* = err;
-                        return;
-                    },
-                    .gid = archive.id(self.hdr.gid_idx) catch |err| {
-                        out_err.* = err;
-                        return;
-                    },
-                    .perm = self.hdr.permissions,
-                };
-            }
-        },
-        .file, .ext_file => {
-            self.extractRegFileThreaded(alloc, archive, path, options, pool) catch |err| {
-                out_err.* = err;
-            };
-        },
-        .symlink, .ext_symlink => {
-            self.extractSymlink(path) catch |err| {
-                out_err.* = err;
-            };
-        },
-        else => {
-            self.extractDevice(archive, path, options) catch |err| {
-                out_err.* = err;
-            };
-        },
-    }
-}
-/// Creates and writes the inode file contents to the given path.
-/// Optionally set owner & permissions.
-///
-/// Assumes the inode is a file or ext_file type.
-fn extractRegFile(self: Inode, alloc: std.mem.Allocator, archive: *Archive, path: []const u8, options: ExtractionOptions) !void {
-    var fil = try std.fs.cwd().createFile(path, .{});
-    defer fil.close();
-    var wrt = fil.writer(&[0]u8{});
-    var dat_rdr = try self.dataReader(alloc, archive);
-    defer dat_rdr.deinit();
-    _ = try dat_rdr.interface.streamRemaining(&wrt.interface);
-    try wrt.interface.flush();
-    // updateTime is in nanoseconds (a billionth of a second). mod_time is in seconds.
-    // TODO: fix
-    // try fil.updateTimes(self.hdr.mod_time, self.hdr.mod_time);
-    if (!options.ignore_permissions) {
-        try fil.chmod(self.hdr.permissions);
-        try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
-    }
-}
-/// Extract the inode file contents to the given path threadedly.
-/// pool is used to spawn threads.
-///
-/// Assumes the inode is a file or ext_file type.
-fn extractRegFileThreaded(self: Inode, alloc: std.mem.Allocator, archive: *Archive, path: []const u8, options: ExtractionOptions, pool: *Pool) !void {
-    var fil = try std.fs.cwd().createFile(path, .{});
-    var data = try self.threadedDataReader(alloc, archive);
-    try data.extractThreaded(fil, pool);
-    if (!options.ignore_permissions) {
-        try fil.chmod(self.hdr.permissions);
-        try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
-    }
-}
-/// Creates the symlink described by the inode.
-///
-/// Assumes the inode is a symlink or ext_symlink type.
-fn extractSymlink(self: Inode, path: []const u8) !void {
-    const target = switch (self.data) {
-        .symlink => |s| s.target,
-        .ext_symlink => |s| s.target,
-        else => unreachable,
-    };
-    try std.fs.cwd().symLink(target, path, .{});
-}
-/// Creates the device described by the inode.
-///
-/// Optionally set owner & permissions.
-/// Assumes the inode is a char_dev, block_dev, fifo, socket, or their extended counterparts.
-fn extractDevice(self: Inode, archive: *Archive, path: []const u8, options: ExtractionOptions) !void {
-    var mode: u32 = undefined;
-    var dev: u32 = 0;
-    switch (self.data) {
-        .char_dev => |d| {
-            mode = std.posix.S.IFCHR;
-            dev = d.dev;
-        },
-        .ext_char_dev => |d| {
-            mode = std.posix.S.IFCHR;
-            dev = d.dev;
-        },
-        .block_dev => |d| {
-            mode = std.posix.S.IFBLK;
-            dev = d.dev;
-        },
-        .ext_block_dev => |d| {
-            mode = std.posix.S.IFBLK;
-            dev = d.dev;
-        },
-        .fifo, .ext_fifo => mode = std.posix.S.IFIFO,
-        .socket, .ext_socket => mode = std.posix.S.IFSOCK,
-        else => unreachable,
-    }
-    const res: std.os.linux.E = @enumFromInt(std.os.linux.mknod(@ptrCast(path), mode, dev));
-    switch (res) {
-        .SUCCESS => {},
-        .ACCES => return std.fs.Dir.MakeError.AccessDenied,
-        .DQUOT => return std.fs.Dir.MakeError.DiskQuota,
-        .EXIST => return std.fs.Dir.MakeError.PathAlreadyExists,
-        .FAULT, .NOENT => return std.fs.Dir.MakeError.BadPathName,
-        .LOOP => return std.fs.Dir.MakeError.SymLinkLoop,
-        .NAMETOOLONG => return std.fs.Dir.MakeError.NameTooLong,
-        .NOMEM => return std.fs.Dir.MakeError.SystemResources,
-        .NOSPC => return std.fs.Dir.MakeError.NoSpaceLeft,
-        .NOTDIR => return std.fs.Dir.MakeError.NotDir,
-        .PERM => return std.fs.Dir.MakeError.PermissionDenied,
-        .ROFS => return std.fs.Dir.MakeError.ReadOnlyFileSystem,
-        else => return blk: {
-            std.debug.print("unhandled mknod result: {}\n", .{res});
-            break :blk std.fs.Dir.MakeError.Unexpected;
-        },
-    }
-    var fil = try std.fs.cwd().openFile(path, .{});
-    // updateTime is in nanoseconds (a billionth of a second). mod_time is in seconds.
-    // TODO: fix
-    // try fil.updateTimes(self.hdr.mod_time, self.hdr.mod_time);
-    if (!options.ignore_permissions) {
-        try fil.chmod(self.hdr.permissions);
-        try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
-    }
-    if (!options.ignore_xattr) {
-        // TODO
-    }
+inline fn dataReaderRaw(decomp: *const Decompressor, fil: OffsetFile, frag_start: u64, block_size: u32, dat: anytype) !DataReader {
+    return .init(
+        decomp,
+        fil,
+        block_size,
+        dat.block_sizes,
+        dat.size,
+        dat.block_start,
+        if (dat.frag_idx != 0xFFFFFFFF)
+            try LookupTable.stateless(FragEntry, fil, decomp, frag_start, dat.frag_idx)
+        else
+            null,
+        dat.frag_offset,
+    );
 }

src/inode/dir.zig

@@ -0,0 +1,35 @@
+const Reader = @import("std").Io.Reader;
+
+pub const Dir = packed struct {
+    block_start: u32,
+    hard_links: u32,
+    size: u16,
+    block_offset: u16,
+    parent_num: u32,
+
+    const Self = @This();
+
+    pub fn read(rdr: *Reader) !Self {
+        var new: Self = undefined;
+        try rdr.readSliceEndian(Self, @ptrCast(&new), .little);
+        return new;
+    }
+};
+
+pub const ExtDir = packed struct {
+    hard_links: u32,
+    size: u32,
+    block_start: u32,
+    parent_num: u32,
+    idx_count: u16,
+    block_offset: u16,
+    xattr_idx: u32,
+
+    const Self = @This();
+
+    pub fn read(rdr: *Reader) !Self {
+        var new: Self = undefined;
+        try rdr.readSliceEndian(Self, @ptrCast(&new), .little);
+        return new;
+    }
+};

src/inode/file.zig

@@ -0,0 +1,69 @@
+const std = @import("std");
+const Reader = std.Io.Reader;
+
+pub const BlockSize = packed struct {
+    size: u31,
+    uncompressed: bool,
+};
+
+pub const File = struct {
+    block_start: u32,
+    frag_idx: u32,
+    frag_offset: u32,
+    size: u32,
+    block_sizes: []BlockSize,
+
+    pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !File {
+        var buf: [16]u8 = undefined;
+        try rdr.readSliceAll(&buf);
+        const frag_idx = std.mem.readVarInt(u32, buf[4..8], .little);
+        const size = std.mem.readVarInt(u32, buf[12..], .little);
+        const sizes_len = size / block_size;
+        if (frag_idx != 0xFFFFFFFF and size % block_size > 0)
+            sizes_len += 1;
+        const sizes = try alloc.alloc(BlockSize, sizes_len);
+        errdefer alloc.free(sizes);
+        try rdr.readSliceEndian(BlockSize, sizes, .little);
+        return .{
+            .block_start = std.mem.readVarInt(u32, buf[0..4], .little),
+            .frag_idx = frag_idx,
+            .frag_offset = std.mem.readVarInt(u32, buf[8..12], .little),
+            .size = size,
+            .block_sizes = sizes,
+        };
+    }
+};
+
+pub const ExtFile = struct {
+    block_start: u64,
+    size: u64,
+    sparse: u64,
+    hard_links: u32,
+    frag_idx: u32,
+    frag_offset: u32,
+    xattr_idx: u32,
+    block_sizes: []BlockSize,
+
+    pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !File {
+        var buf: [40]u8 = undefined;
+        try rdr.readSliceAll(&buf);
+        const frag_idx = std.mem.readVarInt(u32, buf[28..32], .little);
+        const size = std.mem.readVarInt(u64, buf[8..16], .little);
+        const sizes_len = size / block_size;
+        if (frag_idx != 0xFFFFFFFF and size % block_size > 0)
+            sizes_len += 1;
+        const sizes = try alloc.alloc(BlockSize, sizes_len);
+        errdefer alloc.free(sizes);
+        try rdr.readSliceEndian(BlockSize, sizes, .little);
+        return .{
+            .block_start = std.mem.readVarInt(u64, buf[0..8], .little),
+            .size = size,
+            .sparse = std.mem.readVarInt(u64, buf[16..24], .little),
+            .hard_links = std.mem.readVarInt(u32, buf[24..28], .little),
+            .frag_idx = frag_idx,
+            .frag_offset = std.mem.readVarInt(u32, buf[32..36], .little),
+            .xattr_idx = std.mem.readVarInt(u32, buf[36..40], .little),
+            .block_sizes = sizes,
+        };
+    }
+};

src/inode/misc.zig

@@ -0,0 +1,53 @@
+const Reader = @import("std").Io.Reader;
+
+pub const Device = packed struct {
+    hard_links: u32,
+    device: u32,
+
+    const Self = @This();
+
+    pub fn read(rdr: *Reader) !Self {
+        var new: Self = undefined;
+        try rdr.readSliceEndian(Self, @ptrCast(&new), .little);
+        return new;
+    }
+};
+
+pub const ExtDevice = packed struct {
+    hard_links: u32,
+    device: u32,
+    xattr_idx: u32,
+
+    const Self = @This();
+
+    pub fn read(rdr: *Reader) !Self {
+        var new: Self = undefined;
+        try rdr.readSliceEndian(Self, @ptrCast(&new), .little);
+        return new;
+    }
+};
+
+pub const Ipc = packed struct {
+    hard_links: u32,
+
+    const Self = @This();
+
+    pub fn read(rdr: *Reader) !Self {
+        var new: Self = undefined;
+        try rdr.readSliceEndian(Self, @ptrCast(&new), .little);
+        return new;
+    }
+};
+
+pub const ExtIpc = packed struct {
+    hard_links: u32,
+    xattr_idx: u32,
+
+    const Self = @This();
+
+    pub fn read(rdr: *Reader) !Self {
+        var new: Self = undefined;
+        try rdr.readSliceEndian(Self, @ptrCast(&new), .little);
+        return new;
+    }
+};

src/inode/sym.zig

@@ -0,0 +1,42 @@
+const std = @import("std");
+const Reader = std.Io.Reader;
+
+pub const Symlink = struct {
+    hard_links: u32,
+    target: []const u8,
+
+    pub fn read(alloc: std.mem.Allocator, rdr: *Reader) !Symlink {
+        var buf: [8]u8 = undefined;
+        try rdr.readSliceAll(&buf);
+        const size = std.mem.readVarInt(u32, buf[4..], .little);
+        const target = try alloc.alloc(u8, size);
+        errdefer alloc.free(target);
+        try rdr.readSliceEndian(u8, target, .little);
+        return .{
+            .hard_links = std.mem.readVarInt(u32, buf[0..4], .little),
+            .target = target,
+        };
+    }
+};
+
+pub const ExtSymlink = struct {
+    hard_links: u32,
+    xattr_idx: u32,
+    target: []const u8,
+
+    pub fn read(alloc: std.mem.Allocator, rdr: *Reader) !ExtSymlink {
+        var buf: [8]u8 = undefined;
+        try rdr.readSliceAll(&buf);
+        const size = std.mem.readVarInt(u32, buf[4..], .little);
+        const target = try alloc.alloc(u8, size);
+        errdefer alloc.free(target);
+        try rdr.readSliceEndian(u8, target, .little);
+        var xattr_idx: u32 = undefined;
+        try rdr.readSliceEndian(u32, @ptrCast(&xattr_idx), .little);
+        return .{
+            .hard_links = std.mem.readVarInt(u32, buf[0..4], .little),
+            .target = target,
+            .xattr_idx = xattr_idx,
+        };
+    }
+};

src/inode_data/dir.zig

@@ -1,32 +0,0 @@
-const Reader = @import("std").Io.Reader;
-
-pub const Dir = packed struct {
-    block_start: u32,
-    hard_links: u32,
-    size: u16,
-    block_offset: u16,
-    parent_num: u32,
-
-    pub fn read(rdr: *Reader) !Dir {
-        var d: Dir = undefined;
-        try rdr.readSliceEndian(Dir, @ptrCast(&d), .little);
-        return d;
-    }
-};
-
-pub const ExtDir = packed struct {
-    hard_links: u32,
-    size: u32,
-    block_start: u32,
-    parent_num: u32,
-    idx_count: u16,
-    block_offset: u16,
-    xattr_id: u32,
-    // index: []DirIndex
-
-    pub fn read(rdr: *Reader) !ExtDir {
-        var d: ExtDir = undefined;
-        try rdr.readSliceEndian(Dir, @ptrCast(&d), .little);
-        return d;
-    }
-};

src/inode_data/file.zig

@@ -1,76 +0,0 @@
-const std = @import("std");
-const Reader = std.Io.Reader;
-
-pub const BlockSize = packed struct {
-    size: u24,
-    uncompressed: bool,
-    _: u7,
-};
-
-pub const File = struct {
-    block_start: u32, // bytes 0-3
-    frag_idx: u32, // bytes 4-7
-    frag_block_offset: u32, // bytes 8-11
-    size: u32, // bytes 12-15
-    block_sizes: []BlockSize,
-
-    pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !File {
-        var start: [16]u8 = undefined;
-        try rdr.readSliceAll(&start);
-        const frag_idx: u32 = std.mem.readInt(u32, start[4..8], .little);
-        const size: u32 = std.mem.readInt(u32, start[12..16], .little);
-        var num_blocks: u32 = size / block_size;
-        if (size % block_size != 0 and frag_idx == 0xFFFFFFFF) num_blocks += 1;
-        const sizes = try alloc.alloc(BlockSize, num_blocks);
-        errdefer alloc.free(sizes);
-        try rdr.readSliceEndian(BlockSize, sizes, .little);
-        return .{
-            .block_start = std.mem.readInt(u32, start[0..4], .little),
-            .frag_idx = frag_idx,
-            .frag_block_offset = std.mem.readInt(u32, start[8..12], .little),
-            .size = size,
-            .block_sizes = sizes,
-        };
-    }
-
-    pub fn deinit(self: File, alloc: std.mem.Allocator) void {
-        alloc.free(self.block_sizes);
-    }
-};
-
-pub const ExtFile = struct {
-    block_start: u64, // bytes 0-7
-    size: u64, // bytes 8-15
-    sparse: u64, // bytes 16-23
-    hard_links: u32, // bytes 24-27
-    frag_idx: u32, // bytes 28-31
-    frag_block_offset: u32, // bytes 32-35
-    xattr_idx: u32, // bytes 36-39
-    block_sizes: []BlockSize,
-
-    pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !ExtFile {
-        var start: [40]u8 = undefined;
-        try rdr.readSliceAll(&start);
-        const frag_idx: u32 = std.mem.readInt(u32, start[28..32], .little);
-        const size: u64 = std.mem.readInt(u64, start[8..16], .little);
-        var num_blocks: u32 = @truncate(size / block_size);
-        if (size % block_size != 0 and frag_idx == 0xFFFFFFFF) num_blocks += 1;
-        const sizes = try alloc.alloc(BlockSize, num_blocks);
-        errdefer alloc.free(sizes);
-        try rdr.readSliceEndian(BlockSize, sizes, .little);
-        return .{
-            .block_start = std.mem.readInt(u64, start[0..8], .little),
-            .size = size,
-            .sparse = std.mem.readInt(u64, start[16..24], .little),
-            .hard_links = std.mem.readInt(u32, start[24..28], .little),
-            .frag_idx = frag_idx,
-            .frag_block_offset = std.mem.readInt(u32, start[32..36], .little),
-            .xattr_idx = std.mem.readInt(u32, start[36..40], .little),
-            .block_sizes = sizes,
-        };
-    }
-
-    pub fn deinit(self: ExtFile, alloc: std.mem.Allocator) void {
-        alloc.free(self.block_sizes);
-    }
-};

src/inode_data/misc.zig

@@ -1,98 +0,0 @@
-const std = @import("std");
-const Reader = std.Io.Reader;
-
-pub const Symlink = struct {
-    hard_links: u32,
-    target: []const u8,
-
-    pub fn read(alloc: std.mem.Allocator, rdr: *Reader) !Symlink {
-        var start: [8]u8 = undefined;
-        try rdr.readSliceAll(&start);
-        const target_size = std.mem.readInt(u32, start[4..8], .little);
-        const target = try alloc.alloc(u8, target_size + 1);
-        errdefer alloc.free(target);
-        try rdr.readSliceEndian(u8, target, .little);
-        return .{
-            .hard_links = std.mem.readInt(u32, start[0..4], .little),
-            .target = target,
-        };
-    }
-
-    pub fn deinit(self: Symlink, alloc: std.mem.Allocator) void {
-        alloc.free(self.target);
-    }
-};
-
-pub const ExtSymlink = struct {
-    hard_links: u32,
-    target: []const u8,
-    xattr_idx: u32,
-
-    pub fn read(alloc: std.mem.Allocator, rdr: *Reader) !ExtSymlink {
-        var start: [8]u8 = undefined;
-        try rdr.readSliceAll(&start);
-        const target_size = std.mem.readInt(u32, start[4..8], .little);
-        const target = try alloc.alloc(u8, target_size + 1);
-        errdefer alloc.free(target);
-        try rdr.readSliceEndian(u8, target, .little);
-        var xattr_idx: u32 = undefined;
-        try rdr.readSliceEndian(u32, @ptrCast(&xattr_idx), .little);
-        return .{
-            .hard_links = std.mem.readInt(u32, start[0..4], .little),
-            .target = target,
-            .xattr_idx = xattr_idx,
-        };
-    }
-
-    pub fn deinit(self: ExtSymlink, alloc: std.mem.Allocator) void {
-        alloc.free(self.target);
-    }
-};
-
-/// A block or character device.
-pub const Dev = packed struct {
-    hard_links: u32,
-    dev: u32,
-
-    pub fn read(rdr: *Reader) !Dev {
-        var d: Dev = undefined;
-        try rdr.readSliceEndian(Dev, @ptrCast(&d), .little);
-        return d;
-    }
-};
-
-/// An extended block or character device.
-pub const ExtDev = packed struct {
-    hard_links: u32,
-    dev: u32,
-    xattr_idx: u32,
-
-    pub fn read(rdr: *Reader) !ExtDev {
-        var d: ExtDev = undefined;
-        try rdr.readSliceEndian(ExtDev, @ptrCast(&d), .little);
-        return d;
-    }
-};
-
-/// A socket or FIFO file.
-pub const IPC = packed struct {
-    hard_links: u32,
-
-    pub fn read(rdr: *Reader) !IPC {
-        var d: IPC = undefined;
-        try rdr.readSliceEndian(IPC, @ptrCast(&d), .little);
-        return d;
-    }
-};
-
-/// An extended socket or FIFO file.
-pub const ExtIPC = packed struct {
-    hard_links: u32,
-    xattr_idx: u32,
-
-    pub fn read(rdr: *Reader) !ExtIPC {
-        var d: ExtIPC = undefined;
-        try rdr.readSliceEndian(ExtIPC, @ptrCast(&d), .little);
-        return d;
-    }
-};

src/lookup_table.zig

@@ -0,0 +1,168 @@
+const std = @import("std");
+
+const Decompressor = @import("decomp.zig");
+const MetadataReader = @import("util/metadata.zig");
+const OffsetFile = @import("util/offset_file.zig");
+
+pub fn stateless(comptime T: anytype, fil: OffsetFile, decomp: *const Decompressor, table_start: u64, idx: u32) !T {
+    const VALS_PER_BLOCK = 8192 / @sizeOf(T);
+    const block = idx / VALS_PER_BLOCK;
+    const block_idx = idx % VALS_PER_BLOCK;
+
+    const offset = try fil.valueAt(u64, table_start + (8 * block));
+    var buf: [8192]u8 = undefined;
+    var rdr = try fil.readerAt(offset, &buf);
+    var meta_rdr: MetadataReader = .init(&rdr.interface, decomp);
+    try meta_rdr.interface.discardAll(@sizeOf(T) * block_idx);
+
+    var out: T = undefined;
+    try meta_rdr.interface.readSliceEndian(T, @ptrCast(&out), .little);
+    return out;
+}
+
+const InodeRef = @import("inode.zig").Ref;
+
+const XattrLookup = packed struct {
+    // This isn't actuall an inode ref, but is stored that exact same way.
+    ref: InodeRef,
+    kv_count: u32,
+    size: u32,
+};
+
+const XattrKey = packed struct {
+    type: enum(u2) {
+        user,
+        trusted,
+        security,
+    },
+    out_of_line: bool,
+    _: u13,
+    name_size: u16,
+};
+
+pub const XattrValues = std.AutoHashMap([:0]u8, []u8);
+
+pub fn statelessXattr(alloc: std.mem.Allocator, fil: OffsetFile, decomp: *const Decompressor, table_start: u64, idx: u32) !XattrValues {
+    const xattr_start = try fil.valueAt(u64, table_start);
+    const block = idx / 512;
+    const block_idx = idx % 512;
+
+    const block_start = try fil.valueAt(u64, table_start + 8 + (block * 8));
+    var rdr = try fil.readerAt(block_start, &[0]u8{});
+    var meta_rdr: MetadataReader = .init(&rdr.interface, decomp);
+    try meta_rdr.interface.discardAll(16 * block_idx);
+
+    var lookup: XattrLookup = undefined;
+    try meta_rdr.interface.readSliceEndian(XattrLookup, @ptrCast(&lookup), .little);
+
+    rdr = try fil.readerAt(xattr_start + lookup.ref.block_start, &[0]u8{});
+    meta_rdr = .init(&rdr.interface, decomp);
+    try meta_rdr.interface.discardAll(lookup.ref.block_offset);
+
+    var out: XattrValues = try .init(alloc);
+    for (0..lookup.kv_count) |_| {
+        var key: XattrKey = undefined;
+        try meta_rdr.interface.readSliceEndian(XattrKey, @ptrCast(&key), .little);
+        const prefix_size = switch (key.type) {
+            .user => 4,
+            .trusted => 7,
+            .security => 8,
+        };
+        const name: [:0]u8 = try alloc.alloc(u8, prefix_size + key.name_size + 1);
+        name[prefix_size + key.name_size] = 0;
+        try meta_rdr.interface.readSliceEndian(u8, name[prefix_size .. prefix_size + key.name_size], .little);
+        switch (key.type) {
+            .user => @memcpy(name[0..4], "user"),
+            .trusted => @memcpy(name[0..7], "trusted"),
+            .security => @memcpy(name[0..8], "security"),
+        }
+        if (key.out_of_line) {
+            try meta_rdr.interface.discardAll(4);
+            var value_offset: InodeRef = undefined;
+            try meta_rdr.interface.readSliceEndian(InodeRef, @ptrCast(&value_offset), .little);
+
+            var value_rdr = try fil.readerAt(xattr_start + value_offset.block_start, &[0]u8{});
+            var value_meta: MetadataReader = .init(&value_rdr.interface, decomp);
+            try value_meta.interface.discardAll(value_offset.block_offset);
+
+            var val_size: u32 = undefined;
+            try value_meta.interface.readSliceEndian(u32, @ptrCast(&val_size), .little);
+            const value = try alloc.alloc(u8, val_size);
+            try value_meta.interface.readSliceEndian(u8, value, .little);
+            try out.put(name, value);
+        } else {
+            var val_size: u32 = undefined;
+            try meta_rdr.interface.readSliceEndian(u32, @ptrCast(&val_size), .little);
+            const value = try alloc.alloc(u8, val_size);
+            try meta_rdr.interface.readSliceEndian(u8, value, .little);
+            try out.put(name, value);
+        }
+    }
+    return out;
+}
+
+pub fn CachedTable(comptime T: anytype) type {
+    return struct {
+        const Self = @This();
+
+        const VALS_PER_BLOCK = 8192 / @sizeOf(T);
+
+        alloc: std.mem.Allocator,
+        decomp: *const Decompressor,
+
+        fil: OffsetFile,
+        table_start: u64,
+        num: u32,
+
+        cache: std.AutoHashMap(u32, []T),
+        cache_mut: std.Thread.Mutex = .{},
+
+        pub fn init(alloc: std.mem.Allocator, decomp: *const Decompressor, fil: OffsetFile, table_offset: u64, num: u32) !Self {
+            return .{
+                .alloc = alloc,
+                .decomp = decomp,
+
+                .fil = fil,
+                .table_start = table_offset,
+                .num = num,
+
+                .cache = .init(alloc),
+            };
+        }
+        pub fn deinit(self: *Self) void {
+            var values = self.cache.valueIterator();
+            while (values.next()) |val|
+                self.alloc.free(val);
+            self.cache.deinit();
+        }
+
+        pub fn get(self: *Self, idx: u32) !T {
+            const block = idx / VALS_PER_BLOCK;
+            const block_idx = idx % VALS_PER_BLOCK;
+
+            if (self.cache.get(block)) |val|
+                return val[block_idx];
+
+            self.cache_mut.lock();
+            defer self.cache_mut.unlock();
+
+            // Double check in case another thread was doing your work.
+            if (self.cache.get(block)) |val|
+                return val[block_idx];
+
+            const offset = try self.fil.valueAt(u64, self.table_start + (8 * block));
+            var buf: [8192]u8 = undefined;
+            var rdr = try self.fil.readerAt(offset, &buf);
+            var meta_rdr: MetadataReader = .init(&rdr.interface, self.decomp);
+            const block_size = if (block == (self.num - 1) / VALS_PER_BLOCK)
+                self.num % VALS_PER_BLOCK
+            else
+                VALS_PER_BLOCK;
+            const new_block = try self.alloc.alloc(T, block_size);
+            errdefer self.alloc.free(new_block);
+            try meta_rdr.interface.readSliceEndian(T, new_block, .little);
+            try self.cache.put(block, new_block);
+            return new_block[block_idx];
+        }
+    };
+}

src/options.zig

@@ -5,6 +5,8 @@ const Writer = std.Io.Writer;
 
 const ExtractionOptions = @This();
 
+/// The number of threads used for extraction. 0 implies single threaded.
+threads: usize = 1,
 /// Don't set the file's owner & permissions after extraction
 ignore_permissions: bool = false,
 /// Don't set xattr values. Currently xattrs are never set anyway.
@@ -16,10 +18,16 @@ verbose: bool = false,
 /// Where to print verbose log.
 verbose_writer: ?*Writer = null,
 
-pub const Default: ExtractionOptions = .{};
-pub fn VerboseDefault(wrt: *Writer) ExtractionOptions {
+pub const SingleThreadedDefault: ExtractionOptions = .{};
+pub fn Default() !ExtractionOptions {
+    return .{
+        .threads = try std.Thread.getCpuCount(),
+    };
+}
+pub fn VerboseDefault(wrt: *Writer) !ExtractionOptions {
     return .{
         .verbose = true,
         .verbose_writer = wrt,
+        .threads = try std.Thread.getCpuCount(),
     };
 }

src/root.zig

@@ -1 +1,3 @@
 pub const Archive = @import("archive.zig");
+
+pub const ExtractionOptions = @import("options.zig");

src/super.zig

@@ -1,63 +0,0 @@
-const std = @import("std");
-const math = std.math;
-
-const CompressionType = @import("decomp.zig").CompressionType;
-const InodeRef = @import("inode.zig").Ref;
-
-const SQUASHFS_MAGIC: u32 = std.mem.readInt(u32, "hsqs", .little);
-
-const SuperblockError = error{
-    InvalidMagic,
-    InvalidBlockLog,
-    InvalidVersion,
-    InvalidCheck,
-};
-
-/// A squashfs Superblock
-pub const Superblock = packed struct {
-    magic: u32,
-    inode_count: u32,
-    mod_time: u32,
-    block_size: u32,
-    frag_count: u32,
-    compression: CompressionType,
-    block_log: u16,
-    flags: packed struct {
-        inode_uncompressed: bool,
-        data_uncompressed: bool,
-        check: bool,
-        frag_uncompressed: bool,
-        fragment_never: bool,
-        fragment_always: bool,
-        duplicates: bool,
-        exportable: bool,
-        xattr_uncompressed: bool,
-        xattr_never: bool,
-        compression_options: bool,
-        ids_uncompressed: bool,
-        _: u4,
-    },
-    id_count: u16,
-    ver_maj: u16,
-    ver_min: u16,
-    root_ref: InodeRef,
-    size: u64,
-    id_start: u64,
-    xattr_start: u64,
-    inode_start: u64,
-    dir_start: u64,
-    frag_start: u64,
-    export_start: u64,
-
-    /// Validate the Superblock. If an error is returned, it's likely the archive is corrupted or not a squashfs archive.
-    pub fn validate(self: Superblock) !void {
-        if (self.magic != SQUASHFS_MAGIC)
-            return SuperblockError.InvalidMagic;
-        if (self.flags.check)
-            return SuperblockError.InvalidCheck;
-        if (self.ver_maj != 4 or self.ver_min != 0)
-            return SuperblockError.InvalidVersion;
-        if (math.log2(self.block_size) != self.block_log)
-            return SuperblockError.InvalidBlockLog;
-    }
-};

src/table.zig

@@ -1,77 +0,0 @@
-const std = @import("std");
-const Mutex = std.Thread.Mutex;
-
-const DecompFn = @import("decomp.zig").DecompFn;
-const MetadataReader = @import("util/metadata.zig");
-const OffsetFile = @import("util/offset_file.zig");
-
-const TableError = error{
-    InvalidIndex,
-};
-
-/// A two-layer metadata table.
-pub fn Table(T: anytype) type {
-    return struct {
-        const Self = @This();
-
-        const VALS_PER_BLOCK = 8192 / @sizeOf(T);
-
-        alloc: std.mem.Allocator,
-        fil: OffsetFile,
-        decomp: DecompFn,
-        tab_start: u64,
-
-        tab: std.AutoHashMap(u32, []T),
-        values: u32,
-
-        mut: Mutex = .{},
-
-        pub fn init(alloc: std.mem.Allocator, fil: OffsetFile, decomp: DecompFn, tab_start: u64, values: u32) !Self {
-            return .{
-                .alloc = alloc,
-                .fil = fil,
-                .decomp = decomp,
-                .tab_start = tab_start,
-
-                .tab = .init(alloc),
-                .values = values,
-            };
-        }
-
-        pub fn deinit(self: *Self) void {
-            var iter = self.tab.valueIterator();
-            while (iter.next()) |s| {
-                self.alloc.free(s.*);
-            }
-            self.tab.deinit();
-        }
-
-        pub fn get(self: *Self, idx: u32) !T {
-            if (idx >= self.values) return TableError.InvalidIndex;
-            const block_num = idx / VALS_PER_BLOCK;
-            const idx_offset = idx - (block_num * VALS_PER_BLOCK);
-            if (self.tab.contains(block_num)) {
-                const block = self.tab.get(block_num).?;
-                return block[idx_offset];
-            }
-            self.mut.lock();
-            defer self.mut.unlock();
-            // Double check in case of race condition..
-            if (self.tab.contains(block_num)) {
-                const block = self.tab.get(block_num).?;
-                return block[idx_offset];
-            }
-            const is_last = (self.values - 1) / VALS_PER_BLOCK == block_num;
-            const slice_size = if (is_last) self.values - (block_num * VALS_PER_BLOCK) else VALS_PER_BLOCK;
-            const slice = try self.alloc.alloc(T, slice_size);
-            var rdr = try self.fil.readerAt(self.tab_start + (8 * block_num), &[0]u8{});
-            var offset: u64 = 0;
-            try rdr.interface.readSliceEndian(u64, @ptrCast(&offset), .little);
-            rdr = try self.fil.readerAt(offset, &[0]u8{});
-            var meta: MetadataReader = .init(self.alloc, &rdr.interface, self.decomp);
-            try meta.interface.readSliceEndian(T, @ptrCast(slice), .little);
-            try self.tab.put(block_num, slice);
-            return slice[idx_offset];
-        }
-    };
-}

src/test.zig

@@ -1,84 +0,0 @@
-const std = @import("std");
-const stuff = @import("builtin");
-
-const Archive = @import("archive.zig");
-const Superblock = @import("super.zig").Superblock;
-
-const TestArchive = "testing/LinuxPATest.sfs";
-
-test "Basics" {
-    var fil = try std.fs.cwd().openFile(TestArchive, .{});
-    defer fil.close();
-    var sfs: Archive = try .init(std.testing.allocator, fil);
-    defer sfs.deinit();
-    if (sfs.super != LinuxPATestCorrectSuperblock) {
-        std.debug.print("Superblock wrong\nShould be: {}\n\nis: {}\n", .{ LinuxPATestCorrectSuperblock, sfs.super });
-        return error.BadSuperblock;
-    }
-}
-
-const TestFile = "Start.exe";
-const TestFileExtractLocation = "testing/Start.exe";
-
-test "ExtractSingleFile" {
-    std.fs.cwd().deleteFile(TestFileExtractLocation) catch {};
-    var fil = try std.fs.cwd().openFile(TestArchive, .{});
-    defer fil.close();
-    var sfs: Archive = try .init(std.testing.allocator, fil);
-    defer sfs.deinit();
-    var test_fil = try sfs.open(TestFile);
-    defer test_fil.deinit();
-    try test_fil.extract(TestFileExtractLocation, .Default);
-    //TODO: validate extracted file.
-}
-
-const TestFullExtractLocation = "testing/TestExtract";
-
-test "ExtractCompleteArchive" {
-    std.fs.cwd().deleteTree(TestFullExtractLocation) catch {};
-    var fil = try std.fs.cwd().openFile(TestArchive, .{});
-    defer fil.close();
-    var sfs: Archive = try .init(std.testing.allocator, fil);
-    defer sfs.deinit();
-    try sfs.extract(TestFullExtractLocation, .Default);
-}
-
-const LinuxPATestCorrectSuperblock: Superblock = .{
-    .magic = std.mem.readInt(u32, "hsqs", .little),
-    .inode_count = 2974,
-    .mod_time = 1632696724,
-    .block_size = 131072,
-    .frag_count = 264,
-    .compression = .zstd,
-    .block_log = 17,
-    .flags = .{
-        .inode_uncompressed = false,
-        .data_uncompressed = false,
-        .check = false,
-        .frag_uncompressed = false,
-        .fragment_never = false,
-        .fragment_always = false,
-        .duplicates = true,
-        .exportable = true,
-        .xattr_uncompressed = false,
-        .xattr_never = false,
-        .compression_options = false,
-        .ids_uncompressed = false,
-        ._ = 0,
-    },
-    .id_count = 1,
-    .ver_maj = 4,
-    .ver_min = 0,
-    .root_ref = .{
-        .block_offset = 1363,
-        .block_start = 29237,
-        ._ = 0,
-    },
-    .size = 106841744,
-    .id_start = 106841632,
-    .xattr_start = 106841720,
-    .inode_start = 106778274,
-    .dir_start = 106807998,
-    .frag_start = 106837747,
-    .export_start = 106841602,
-};

src/util/data.zig

@@ -1,174 +0,0 @@
-//! A reader for a regular file.
-
-const std = @import("std");
-const Reader = std.Io.Reader;
-const Writer = std.Io.Writer;
-const Limit = std.Io.Limit;
-
-const Archive = @import("../archive.zig");
-const FragEntry = Archive.FragEntry;
-const DecompFn = @import("../decomp.zig").DecompFn;
-const BlockSize = @import("../inode_data/file.zig").BlockSize;
-const OffsetFile = @import("offset_file.zig");
-
-const DataReader = @This();
-
-alloc: std.mem.Allocator,
-fil: OffsetFile,
-decomp: DecompFn,
-block_size: u32,
-
-blocks: []BlockSize,
-
-frag: ?FragEntry = null, // TODO: do something better?
-frag_offset: u32 = 0,
-size: u64,
-
-interface: Reader,
-
-cur_offset: u64,
-block_idx: u32 = 0,
-
-pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, start: u64, size: u64) DataReader {
-    return .{
-        .alloc = alloc,
-        .fil = archive.fil,
-        .decomp = archive.decomp,
-        .block_size = archive.super.block_size,
-        .blocks = blocks,
-        .size = size,
-        .cur_offset = start,
-        .interface = .{
-            .end = 0,
-            .seek = 0,
-            .buffer = &[0]u8{},
-            .vtable = &.{
-                .stream = stream,
-                .discard = discard,
-                .readVec = readVec,
-            },
-        },
-    };
-}
-pub fn deinit(self: *DataReader) void {
-    self.alloc.free(self.interface.buffer);
-    self.interface.end = 0;
-    self.interface.seek = 0;
-}
-
-pub fn addFragment(self: *DataReader, entry: FragEntry, frag_offset: u32) void {
-    self.frag = entry;
-    self.frag_offset = frag_offset;
-}
-
-fn numBlocks(self: DataReader) usize {
-    var res = self.blocks.len;
-    if (self.frag != null) res += 1;
-    return res;
-}
-
-fn advance(self: *DataReader) !void {
-    if (self.block_idx > self.blocks.len or (self.block_idx == self.blocks.len and self.frag == null)) {
-        if (self.interface.buffer.len > 0) {
-            self.alloc.free(self.interface.buffer);
-            self.interface.buffer = &[0]u8{};
-            self.interface.end = 0;
-            self.interface.seek = 0;
-        }
-        return Reader.Error.EndOfStream;
-    }
-    defer self.block_idx += 1;
-    const cur_block_size = if (self.block_idx == self.numBlocks() - 1) self.size % self.block_size else self.block_size;
-    try self.resizeBuffer(cur_block_size);
-    self.interface.seek = 0;
-    self.interface.end = cur_block_size;
-    if (self.block_idx == self.blocks.len) { // fragment
-        var rdr = try self.fil.readerAt(self.frag.?.start, &[0]u8{});
-        if (self.frag.?.size.uncompressed) {
-            try rdr.interface.discardAll(self.frag_offset);
-            try rdr.interface.readSliceAll(self.interface.buffer);
-            return;
-        }
-        const tmp_buf = try self.alloc.alloc(u8, self.frag.?.size.size);
-        defer self.alloc.free(tmp_buf);
-        try rdr.interface.readSliceAll(tmp_buf);
-        const needed_block = try self.alloc.alloc(u8, self.block_size);
-        defer self.alloc.free(needed_block);
-        _ = try self.decomp(self.alloc, tmp_buf, needed_block);
-        @memcpy(self.interface.buffer, needed_block[self.frag_offset .. self.frag_offset + cur_block_size]);
-        return;
-    }
-    const block = self.blocks[self.block_idx];
-    if (block.size == 0) {
-        @memset(self.interface.buffer, 0);
-        return;
-    }
-    var rdr = try self.fil.readerAt(self.cur_offset, &[0]u8{});
-    self.cur_offset += block.size;
-    if (block.uncompressed) {
-        try rdr.interface.readSliceAll(self.interface.buffer);
-        return;
-    }
-    const tmp_buf = try self.alloc.alloc(u8, block.size);
-    defer self.alloc.free(tmp_buf);
-    try rdr.interface.readSliceAll(tmp_buf);
-    _ = try self.decomp(self.alloc, tmp_buf, self.interface.buffer);
-}
-/// Does not guarentee that data currently in the buffer is retained.
-fn resizeBuffer(self: *DataReader, size: usize) !void {
-    if (self.interface.buffer.len == size) return;
-    if (!self.alloc.resize(self.interface.buffer, size)) {
-        self.alloc.free(self.interface.buffer);
-        self.interface.buffer = self.alloc.alloc(u8, size) catch |err| {
-            self.interface.buffer = &[0]u8{};
-            return err;
-        };
-    } else {
-        self.interface.buffer.len = size;
-    }
-}
-
-fn stream(rdr: *Reader, wrt: *Writer, limit: Limit) Reader.StreamError!usize {
-    var self: *DataReader = @alignCast(@fieldParentPtr("interface", rdr));
-    if (rdr.seek >= rdr.end) self.advance() catch |err| {
-        if (err == error.EndOfStream) return error.EndOfStream;
-        std.log.err("Error advancing data reader: {}\n", .{err});
-        return Reader.Error.ReadFailed;
-    };
-    if (limit == .nothing) return 0;
-    const to_read = @min(rdr.end - rdr.seek, @intFromEnum(limit));
-    const res = try wrt.write(rdr.buffer[rdr.seek .. rdr.seek + to_read]);
-    rdr.seek += res;
-    return res;
-}
-
-fn discard(rdr: *Reader, limit: Limit) Reader.Error!usize {
-    var self: *DataReader = @alignCast(@fieldParentPtr("interface", rdr));
-    if (rdr.seek >= rdr.end) self.advance() catch |err| {
-        if (err == error.EndOfStream) return error.EndOfStream;
-        std.log.err("Error advancing data reader: {}\n", .{err});
-        return Reader.Error.ReadFailed;
-    };
-    if (limit == .nothing) return 0;
-    const to_adv = @min(rdr.end - rdr.seek, @intFromEnum(limit));
-    rdr.seek += to_adv;
-    return to_adv;
-}
-
-fn readVec(rdr: *Reader, vec: [][]u8) Reader.Error!usize {
-    var self: *DataReader = @alignCast(@fieldParentPtr("interface", rdr));
-    if (rdr.seek >= rdr.end) self.advance() catch |err| {
-        if (err == error.EndOfStream) return error.EndOfStream;
-        std.log.err("Error advancing data reader: {}\n", .{err});
-        return Reader.Error.ReadFailed;
-    };
-    var cur_red: usize = 0;
-    for (vec) |s| {
-        const to_copy: usize = @min(rdr.end - rdr.seek, s.len);
-        @memcpy(s[0..to_copy], rdr.buffer[rdr.seek .. rdr.seek + to_copy]);
-        rdr.seek += to_copy;
-        cur_red += to_copy;
-        if (rdr.end == rdr.seek) break;
-    }
-    return cur_red;
-}

src/util/data_reader.zig

@@ -0,0 +1,138 @@
+const std = @import("std");
+const Reader = std.Io.Reader;
+const Writer = std.Io.Writer;
+const Limit = std.Io.Limit;
+
+const FragEntry = @import("../archive.zig").FragEntry;
+const Decompressor = @import("../decomp.zig");
+const BlockSize = @import("../inode/file.zig").BlockSize;
+const OffsetFile = @import("offset_file.zig");
+
+const DataReader = @This();
+
+decomp: *const Decompressor,
+file: OffsetFile,
+block_size: u32,
+blocks: []BlockSize,
+size: u64,
+frag: ?FragEntry,
+frag_offset: u32,
+
+offset: u64,
+idx: usize = 0,
+sparse: bool = false,
+
+interface: Reader,
+
+pub fn init(decomp: *const Decompressor, file: OffsetFile, block_size: u32, blocks: []BlockSize, size: u64, init_offset: u64, frag: ?FragEntry, frag_offset: u32) DataReader {
+    return .{
+        .decomp = decomp,
+        .file = file,
+        .block_size = block_size,
+        .blocks = blocks,
+        .size = size,
+        .frag = frag,
+        .frag_offset = frag_offset,
+
+        .offset = init_offset,
+
+        .interface = .{
+            .buffer = &[1]u8{undefined} ** (1024 * 1024),
+            .end = 0,
+            .seek = 0,
+            .vtable = &.{ .stream = stream, .discard = discard, .readVec = readVec },
+        },
+    };
+}
+
+fn numBlocks(self: *DataReader) usize {
+    return if (self.frag == null)
+        self.blocks.len
+    else
+        self.blocks.len + 1;
+}
+fn advanceBuffer(self: *DataReader) Reader.Error!void {
+    if (self.idx >= self.numBlocks()) return Reader.Error.EndOfStream;
+    defer self.idx += 1;
+    self.sparse = false;
+    self.interface.end = 0; // If we error out and the error is ignored, we'll stil end up back here to error again.
+    self.interface.seek = 0;
+    if (self.idx == self.blocks.len) { // Fragment
+        var rdr = self.file.readerAt(self.frag.?.block_start, &[0]u8{}) catch return Reader.Error.ReadFailed;
+        const size = self.size % self.block_size;
+        if (self.frag.?.size.uncompressed) {
+            try rdr.interface.discardAll(self.frag_offset);
+            try rdr.interface.readSliceAll(self.interface.buffer[0..size]);
+            self.interface.end = size;
+            return;
+        }
+        const raw_loc = self.interface.buffer.len - self.frag.?.size.size;
+        try rdr.interface.readSliceAll(self.interface.buffer[raw_loc..]);
+        _ = self.decomp.decompress(self.interface.buffer[raw_loc..], self.interface.buffer) catch
+            return Reader.Error.ReadFailed;
+        @memmove(self.interface.buffer[0..size], self.interface.buffer[self.frag_offset .. self.frag_offset + size]);
+        self.interface.end = size;
+        return;
+    }
+    const block = self.blocks[self.idx];
+    if (block.size == 0) {
+        self.interface.end = if (self.idx == self.numBlocks() - 1)
+            self.size % self.block_size
+        else
+            self.block_size;
+        self.sparse = true;
+        return;
+    }
+    defer self.offset += block.size;
+    var rdr = try self.file.readerAt(self.offset, &[0]u8{});
+    if (block.uncompressed) {
+        try rdr.interface.readSliceAll(self.interface.buffer[0..block.size]);
+        self.interface.end = block.size;
+        return;
+    }
+    const raw_loc = self.interface.buffer.len - block.size;
+    try rdr.interface.readSliceAll(self.interface.buffer[raw_loc..]);
+    self.interface.end = self.decomp.decompress(self.interface.buffer[raw_loc..], self.interface.buffer) catch
+        return Reader.Error.ReadFailed;
+}
+
+fn stream(r: *Reader, wrt: *Writer, limit: Limit) Reader.StreamError!usize {
+    var self: *DataReader = @fieldParentPtr("interface", r);
+    if (r.seek == r.end) try self.advanceBuffer();
+    if (limit == .nothing) return 0;
+
+    const to_write = @min(r.end - r.seek, @intFromEnum(limit));
+    const wrote = if (self.sparse)
+        try wrt.splatByte(0, to_write)
+    else
+        try wrt.write(r.buffer[r.seek .. r.seek + to_write]);
+    r.seek += wrote;
+    return wrote;
+}
+fn discard(r: *Reader, limit: Limit) Reader.Error!usize {
+    var self: *DataReader = @fieldParentPtr("interface", r);
+    if (r.seek == r.end) try self.advanceBuffer();
+    if (limit == .nothing) return 0;
+
+    const adv = @min(r.end - r.seek, @intFromEnum(limit));
+    r.seek += adv;
+    return adv;
+}
+fn readVec(r: *Reader, vec: [][]u8) Reader.Error!usize {
+    var self: *DataReader = @fieldParentPtr("interface", r);
+    if (r.seek == r.end) try self.advanceBuffer();
+
+    var wrote: usize = 0;
+    for (vec) |slice| {
+        if (r.seek == r.end) break;
+        const to_copy = @min(r.end - r.seek, slice.len);
+        if (self.sparse) {
+            @memset(slice[0..to_copy], 0);
+        } else {
+            @memcpy(slice[0..to_copy], r.buffer[r.seek .. r.seek + to_copy]);
+        }
+        r.seek += to_copy;
+        wrote += to_copy;
+    }
+    return wrote;
+}

src/util/data_threaded.zig

@@ -1,183 +0,0 @@
-//! Similiar to DataReader, but set-up for threaded writing to files.
-
-const std = @import("std");
-const Reader = std.Io.Reader;
-const Writer = std.Io.Writer;
-const Limit = std.Io.Limit;
-const WaitGroup = std.Thread.WaitGroup;
-const Pool = std.Thread.Pool;
-
-const Archive = @import("../archive.zig");
-const FragEntry = Archive.FragEntry;
-const DecompFn = @import("../decomp.zig").DecompFn;
-const BlockSize = @import("../inode_data/file.zig").BlockSize;
-const OffsetFile = @import("offset_file.zig");
-
-const ThreadedDataReader = @This();
-
-alloc: std.mem.Allocator,
-fil: OffsetFile,
-decomp: DecompFn,
-block_size: u32,
-
-blocks: []BlockSize,
-
-frag: ?FragEntry = null, // TODO: do something better?
-frag_offset: u32 = 0,
-size: u64,
-
-start_offset: u64,
-
-pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, start: u64, size: u64) ThreadedDataReader {
-    return .{
-        .alloc = alloc,
-        .fil = archive.fil,
-        .decomp = archive.decomp,
-        .block_size = archive.super.block_size,
-        .blocks = blocks,
-        .size = size,
-        .start_offset = start,
-    };
-}
-
-pub fn addFragment(self: *ThreadedDataReader, entry: FragEntry, frag_offset: u32) void {
-    self.frag = entry;
-    self.frag_offset = frag_offset;
-}
-
-fn numBlocks(self: ThreadedDataReader) usize {
-    var res = self.blocks.len;
-    if (self.frag != null) res += 1;
-    return res;
-}
-
-/// Extract the data to the file threadedly, using pool to spawn threads.
-/// If multiple errors occur, thread spawning errors will have, then the last decompression error that occurs;
-///
-/// The function must be called from an unused DataReader. The DataReader is still usable afterwards.
-/// If only extractThreaded is used, there is no need to call deinit() afterwards.
-///
-/// The file will always be written to starting at 0.
-pub fn extractThreaded(self: ThreadedDataReader, file: std.fs.File, pool: *Pool) !void {
-    var wg: WaitGroup = .{};
-    wg.startMany(self.numBlocks());
-    var out_err: ?anyerror = null;
-
-    var cur_write_offset: u64 = 0;
-    var cur_read_offset: u64 = self.start_offset;
-    for (0..self.blocks.len) |i| {
-        const cur_block_size = if (i == self.numBlocks() - 1) self.size % self.block_size else self.block_size;
-        try pool.spawn(workThreadBlocks, .{ self, file, cur_write_offset, cur_read_offset, self.blocks[i], cur_block_size, &wg, &out_err });
-        cur_write_offset += cur_block_size;
-        cur_read_offset += self.blocks[i].size;
-    }
-    if (self.frag != null) {
-        try pool.spawn(workThreadFragment, .{ self, file, cur_write_offset, &wg, &out_err });
-    }
-    pool.waitAndWork(&wg);
-    if (out_err != null) return out_err.?;
-}
-
-fn workThreadBlocks(self: ThreadedDataReader, fil: std.fs.File, write_offset: u64, read_offset: u64, block: BlockSize, cur_block_size: u64, wg: *WaitGroup, out_err: *?anyerror) void {
-    defer wg.finish();
-    var wrt = fil.writer(&[0]u8{});
-    wrt.seekTo(write_offset) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    defer wrt.interface.flush() catch |err| {
-        out_err.* = err;
-    };
-    if (block.size == 0) {
-        wrt.interface.splatByteAll(0, cur_block_size) catch |err| {
-            out_err.* = err;
-            return;
-        };
-        return;
-    }
-    var rdr = self.fil.readerAt(read_offset, &[0]u8{}) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    if (block.uncompressed) {
-        rdr.interface.streamExact(&wrt.interface, block.size) catch |err| {
-            out_err.* = err;
-            return;
-        };
-        return;
-    }
-    // TODO: shared buffers
-    const read_buf = self.alloc.alloc(u8, block.size) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    defer self.alloc.free(read_buf);
-    rdr.interface.readSliceAll(read_buf) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    // TODO: shared buffers
-    const res_buf = self.alloc.alloc(u8, cur_block_size) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    defer self.alloc.free(res_buf);
-    _ = self.decomp(self.alloc, read_buf, res_buf) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    wrt.interface.writeAll(res_buf) catch |err| {
-        out_err.* = err;
-        return;
-    };
-}
-fn workThreadFragment(self: ThreadedDataReader, fil: std.fs.File, write_offset: u64, wg: *WaitGroup, out_err: *?anyerror) void {
-    defer wg.finish();
-
-    var wrt = fil.writer(&[0]u8{});
-    wrt.seekTo(write_offset) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    defer wrt.interface.flush() catch |err| {
-        out_err.* = err;
-    };
-
-    var rdr = self.fil.readerAt(self.frag.?.start, &[0]u8{}) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    if (self.frag.?.size.uncompressed) {
-        rdr.interface.discardAll(self.frag_offset) catch |err| {
-            out_err.* = err;
-            return;
-        };
-        rdr.interface.streamExact(&wrt.interface, self.size % self.block_size) catch |err| {
-            out_err.* = err;
-            return;
-        };
-        return;
-    }
-    const tmp_buf = self.alloc.alloc(u8, self.frag.?.size.size) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    defer self.alloc.free(tmp_buf);
-    rdr.interface.readSliceAll(tmp_buf) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    const needed_block = self.alloc.alloc(u8, self.block_size) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    defer self.alloc.free(needed_block);
-    _ = self.decomp(self.alloc, tmp_buf, needed_block) catch |err| {
-        out_err.* = err;
-        return;
-    };
-    wrt.interface.writeAll(needed_block[self.frag_offset .. self.frag_offset + (self.size % self.block_size)]) catch |err| {
-        out_err.* = err;
-        return;
-    };
-}

src/util/iter.zig

@@ -0,0 +1,54 @@
+const std = @import("std");
+
+const MinimalSuperblock = @import("../archive.zig").MinimalSuperblock;
+const Decompressor = @import("../decomp.zig");
+const DirEntry = @import("../directory.zig").Entry;
+const File = @import("../file.zig");
+const Inode = @import("../inode.zig");
+const MetadataReader = @import("metadata.zig");
+const OffsetFile = @import("offset_file.zig");
+const Utils = @import("utils.zig");
+
+const Iter = @This();
+
+file: OffsetFile,
+super: MinimalSuperblock,
+decomp: Decompressor,
+
+entries: []DirEntry,
+idx: usize = 0,
+
+pub fn deinit(self: Iter) void {
+    for (self.entries) |ent|
+        ent.deinit(self.decomp.alloc);
+    self.decomp.alloc.free(self.entries);
+}
+
+pub fn next(self: *Iter) !?File {
+    if (self.idx >= self.entries.len) return null;
+    defer self.idx += 1;
+
+    const entry = self.entries[self.idx];
+
+    const new_name = try self.decomp.alloc.alloc(u8, entry.name.len);
+    @memcpy(new_name, entry.name);
+    return .{
+        .file = self.file,
+        .super = self.super,
+        .decomp = self.decomp,
+
+        .name = new_name,
+        .inode = Utils.readInode(
+            self.decomp.alloc,
+            &self.decomp,
+            self.file,
+            self.super.inode_start,
+            self.super.block_size,
+            entry.block_start,
+            entry.block_offset,
+        ),
+    };
+}
+pub fn reset(self: *Iter) void {
+    self.idx = 0;
+}

src/util/metadata.zig

@@ -2,33 +2,22 @@ const std = @import("std");
 const Reader = std.Io.Reader;
 const Writer = std.Io.Writer;
 const Limit = std.Io.Limit;
-const StreamError = std.Io.Reader.StreamError;
 
-const DecompFn = @import("../decomp.zig").DecompFn;
+const Decompressor = @import("../decomp.zig");
 
-const BlockHeader = packed struct {
+const Header = packed struct {
     size: u15,
     uncompressed: bool,
 };
 
-const This = @This();
+const MetadataReader = @This();
 
-alloc: std.mem.Allocator,
 rdr: *Reader,
-decomp: DecompFn,
+decomp: *const Decompressor,
 
-buf: [8192]u8 = undefined,
-
-interface: Reader,
-err: ?anyerror = null,
-
-pub fn init(alloc: std.mem.Allocator, rdr: *Reader, decomp: DecompFn) This {
-    return .{
-        .alloc = alloc,
-        .rdr = rdr,
-        .decomp = decomp,
-        .interface = .{
-            .buffer = &[0]u8{},
+read_buf: [8192]u8 = undefined,
+interface: Reader = .{
+    .buffer = &([1]u8{undefined} ** 8192),
     .end = 0,
     .seek = 0,
     .vtable = &.{
@@ -36,62 +25,53 @@ pub fn init(alloc: std.mem.Allocator, rdr: *Reader, decomp: DecompFn) This {
         .discard = discard,
         .readVec = readVec,
     },
-        },
-    };
-}
+},
 
-fn advance(self: *This) !void {
+pub fn init(rdr: *Reader, decomp: *const Decompressor) MetadataReader {
+    return .{ .rdr = rdr, .decomp = decomp };
+}
+fn advanceBuffer(self: *MetadataReader) Reader.Error!void {
     self.interface.seek = 0;
-    var hdr: BlockHeader = undefined;
-    try self.rdr.readSliceEndian(BlockHeader, @ptrCast(&hdr), .little);
+    var hdr: Header = undefined;
+    try self.rdr.readSliceEndian(Header, @ptrCast(&hdr), .little);
+    try self.rdr.readSliceAll(self.read_buf[0..hdr.size]);
     if (hdr.uncompressed) {
-        try self.rdr.readSliceEndian(u8, self.buf[0..hdr.size], .little);
+        @memcpy(self.interface.buffer[0..hdr.size], self.read_buf[0..hdr.size]);
         self.interface.end = hdr.size;
-        self.interface.buffer = self.buf[0..hdr.size];
         return;
     }
-    var tmp_buf: [8192]u8 = undefined;
-    try self.rdr.readSliceAll(tmp_buf[0..hdr.size]);
-    self.interface.end = try self.decomp(self.alloc, tmp_buf[0..hdr.size], &self.buf);
-    self.interface.buffer = self.buf[0..self.interface.end];
+    self.interface.end = self.decomp.decompress(self.read_buf[0..hdr.size], self.interface.buffer) catch |err|
+        return switch (err) {
+            error.OutOfMemory => error.ReadFailed,
+            else => err,
+        };
 }
 
-fn stream(rdr: *Reader, wrt: *Writer, limit: Limit) StreamError!usize {
-    const self: *This = @fieldParentPtr("interface", rdr);
-    if (rdr.end == rdr.seek) self.advance() catch |err| {
-        self.err = err;
-        return StreamError.ReadFailed;
-    };
-    if (@intFromEnum(limit) == 0) return 0;
-    const to_write = @min(rdr.end - rdr.seek, @intFromEnum(limit));
-    const wrote = try wrt.write(self.buf[rdr.seek .. rdr.seek + to_write]);
-    self.interface.seek += wrote;
+fn stream(rdr: *Reader, wrt: *Writer, limit: Limit) Reader.StreamError!usize {
+    var self: *MetadataReader = @fieldParentPtr("interface", rdr);
+    if (rdr.seek == rdr.end) try self.advanceBuffer();
+    const to_write = @min(@intFromEnum(limit), rdr.end - rdr.seek);
+    const wrote = try wrt.write(rdr.buffer[rdr.seek .. rdr.seek + to_write]);
+    rdr.seek += wrote;
     return wrote;
 }
-fn discard(rdr: *Reader, limit: Limit) error{ EndOfStream, ReadFailed }!usize {
-    const self: *This = @fieldParentPtr("interface", rdr);
-    if (rdr.end == rdr.seek) self.advance() catch |err| {
-        self.err = err;
-        return StreamError.ReadFailed;
-    };
-    if (@intFromEnum(limit) == 0) return 0;
-    const to_skip = @min(rdr.end - rdr.seek, @intFromEnum(limit));
-    rdr.seek += to_skip;
-    return to_skip;
+fn discard(rdr: *Reader, limit: Limit) Reader.Error!usize {
+    var self: *MetadataReader = @fieldParentPtr("interface", rdr);
+    if (rdr.seek == rdr.end) try self.advanceBuffer();
+    const to_adv = @min(@intFromEnum(limit), rdr.end - rdr.seek);
+    rdr.seek += to_adv;
+    return to_adv;
 }
-fn readVec(rdr: *Reader, vec: [][]u8) error{ EndOfStream, ReadFailed }!usize {
-    const self: *This = @fieldParentPtr("interface", rdr);
-    if (rdr.end == rdr.seek) self.advance() catch |err| {
-        self.err = err;
-        return StreamError.ReadFailed;
-    };
-    var cur_red: usize = 0;
-    for (vec) |s| {
-        const to_copy: usize = @min(rdr.end - rdr.seek, s.len);
-        @memcpy(s[0..to_copy], self.buf[rdr.seek .. rdr.seek + to_copy]);
-        rdr.seek += to_copy;
-        cur_red += to_copy;
-        if (rdr.end == rdr.seek) break;
+fn readVec(rdr: *Reader, vec: [][]u8) Reader.Error!usize {
+    var self: *MetadataReader = @fieldParentPtr("interface", rdr);
+    if (rdr.seek == rdr.end) try self.advanceBuffer();
+    var wrote = 0;
+    for (vec) |v| {
+        if (rdr.seek == rdr.end) break;
+        const to_write = @min(v.len, rdr.end - rdr.seek);
+        @memcpy(v[0..to_write], rdr.buffer[rdr.seek .. rdr.seek + to_write]);
+        wrote += to_write;
+        rdr.seek += to_write;
     }
-    return cur_red;
+    return wrote;
 }

src/util/offset_file.zig

@@ -1,20 +1,21 @@
-//! A File where it's meaningful (to us) content starts at a given offset.
-
 const std = @import("std");
-const File = std.fs.File;
-const Reader = std.fs.File.Reader;
+const Io = std.Io;
+const FileReader = Io.File.Reader;
 
 const OffsetFile = @This();
 
-fil: File,
-offset: u64,
+fil: Io.File,
+offset: u64 = 0,
 
-pub fn init(fil: File, init_offset: u64) OffsetFile {
-    return .{ .fil = fil, .offset = init_offset };
-}
-
-pub fn readerAt(self: OffsetFile, offset: u64, buffer: []u8) !Reader {
-    var rdr = self.fil.reader(buffer);
+pub fn readerAt(self: OffsetFile, io: Io, offset: u64, buf: []u8) !FileReader {
+    var rdr = self.fil.reader(io, buf);
     try rdr.seekTo(self.offset + offset);
     return rdr;
 }
+pub fn valueAt(self: OffsetFile, comptime T: type, io: Io, offset: u64) !T {
+    var rdr = self.fil.reader(io, &[0]u8{});
+    try rdr.seekTo(self.offset + offset);
+    var new: T = undefined;
+    try rdr.interface.readSliceEndian(T, @ptrCast(&new), .little);
+    return new;
+}

src/util/utils.zig

@@ -0,0 +1,25 @@
+const std = @import("std");
+
+const Decompressor = @import("../decomp.zig");
+const DirEntry = @import("../directory.zig").Entry;
+const Inode = @import("../inode.zig");
+const MetadataReader = @import("metadata.zig");
+const OffsetFile = @import("offset_file.zig");
+
+pub fn pathIsSelf(path: []const u8) bool {
+    if (path.len == 0) return true;
+    if (path.len == 1) {
+        return switch (path[0]) {
+            '.', '/' => true,
+            else => false,
+        };
+    }
+    return std.mem.eql(u8, path, "./");
+}
+
+pub fn readInode(alloc: std.mem.Allocator, decomp: *const Decompressor, fil: OffsetFile, inode_start: u64, block_size: u32, block_start: u32, block_offset: u16) !Inode {
+    var rdr = try fil.readerAt(inode_start + block_start, &[0]u8{});
+    var meta: MetadataReader = .init(&rdr.interface, decomp);
+    try meta.interface.discardAll(block_offset);
+    return .read(alloc, &meta.interface, block_size);
+}