5 files changed, 193 insertions, 57 deletions

diff --git a/codegen/ccc.zig b/codegen/ccc.zig
index 4e470ae..e76222f 100644
--- a/codegen/ccc.zig
+++ b/codegen/ccc.zig
@@ -112,12 +112,24 @@ pub fn main() anyerror!void {
     defer out_file.close();
     var writer = out_file.writer(&write_buf);
 
-    const endian = builtin.cpu.arch.endian();
-    try writer.interface.writeInt(u16, @intCast(stage1.items.len), endian);
-    for (stage1.items) |i| try writer.interface.writeInt(u16, i, endian);
-
-    try writer.interface.writeInt(u16, @intCast(stage2.items.len), endian);
-    try writer.interface.writeAll(stage2.items);
+    try writer.interface.print(
+        \\//! This file is auto-generated. Do not edit.
+        \\
+        \\pub const s1: [{}]u16 = .{{
+    , .{stage1.items.len});
+    for (stage1.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.print(
+        \\
+        \\}};
+        \\
+        \\pub const s2: [{}]u8 = .{{
+    , .{stage2.items.len});
+    for (stage2.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.writeAll(
+        \\};
+    );
 
     try writer.interface.flush();
 }
diff --git a/codegen/compat.zig b/codegen/compat.zig
index debb83d..a9d1f92 100644
--- a/codegen/compat.zig
+++ b/codegen/compat.zig
@@ -1,58 +1,82 @@
 const std = @import("std");
 const builtin = @import("builtin");
 
+const block_size = 256;
+const Block = [block_size][]const u21;
+
+const BlockMap = std.HashMap(
+    Block,
+    u16,
+    struct {
+        pub fn hash(_: @This(), k: Block) u64 {
+            var hasher = std.hash.Wyhash.init(0);
+            std.hash.autoHashStrat(&hasher, k, .DeepRecursive);
+            return hasher.final();
+        }
+
+        pub fn eql(_: @This(), aBlock: Block, bBlock: Block) bool {
+            return for (aBlock, bBlock) |a, b| {
+                if (a.len != b.len) return false;
+                for (a, b) |a_cp, b_cp| {
+                    if (a_cp != b_cp) return false;
+                }
+            } else true;
+        }
+    },
+    std.hash_map.default_max_load_percentage,
+);
+
 pub fn main() anyerror!void {
     var arena = std.heap.ArenaAllocator.init(std.heap.page_allocator);
     defer arena.deinit();
     const allocator = arena.allocator();
 
     // Process UnicodeData.txt
-    var write_buf: [4096]u8 = undefined;
-
     var in_reader = std.io.Reader.fixed(@embedFile("UnicodeData.txt"));
     var args_iter = try std.process.argsWithAllocator(allocator);
     defer args_iter.deinit();
     _ = args_iter.skip();
     const output_path = args_iter.next() orelse @panic("No output file arg!");
 
-    var out_file = try std.fs.cwd().createFile(output_path, .{});
-    defer out_file.close();
-    var writer = out_file.writer(&write_buf);
+    var compat_map = std.AutoHashMap(u21, []u21).init(allocator);
+    defer compat_map.deinit();
 
-    const endian = builtin.cpu.arch.endian();
-
-    lines: while (in_reader.takeDelimiterInclusive('\n')) |took| {
-        const line = std.mem.trimRight(u8, took, "\n");
+    while (in_reader.takeDelimiterInclusive('\n')) |line| {
         if (line.len == 0) continue;
 
         var field_iter = std.mem.splitScalar(u8, line, ';');
-        var cps: [19]u24 = undefined;
-        var len: u8 = 1;
+        var cp: u21 = undefined;
 
         var i: usize = 0;
         while (field_iter.next()) |field| : (i += 1) {
+            if (field.len == 0) continue;
+
             switch (i) {
-                0 => cps[0] = try std.fmt.parseInt(u24, field, 16),
+                0 => {
+                    cp = try std.fmt.parseInt(u21, field, 16);
+                },
 
                 5 => {
                     // Not compatibility.
-                    if (field.len == 0 or field[0] != '<') continue :lines;
+                    if (field[0] != '<') continue;
+
                     var cp_iter = std.mem.tokenizeScalar(u8, field, ' ');
                     _ = cp_iter.next(); // <compat type>
 
+                    var cps: [18]u21 = undefined;
+                    var len: u8 = 0;
+
                     while (cp_iter.next()) |cp_str| : (len += 1) {
-                        cps[len] = try std.fmt.parseInt(u24, cp_str, 16);
+                        cps[len] = try std.fmt.parseInt(u21, cp_str, 16);
                     }
-                },
 
-                2 => if (line[0] == '<') continue :lines,
+                    const slice = try allocator.dupe(u21, cps[0..len]);
+                    try compat_map.put(cp, slice);
+                },
 
                 else => {},
             }
         }
-
-        try writer.interface.writeInt(u8, @intCast(len), endian);
-        for (cps[0..len]) |cp| try writer.interface.writeInt(u24, cp, endian);
     } else |err| switch (err) {
         error.EndOfStream => {},
         else => {
@@ -60,6 +84,63 @@ pub fn main() anyerror!void {
         },
     }
 
-    try writer.interface.writeInt(u16, 0, endian);
+    // Build multi-tiered lookup tables for compatibility decompositions
+    var blocks_map = BlockMap.init(allocator);
+    defer blocks_map.deinit();
+
+    var stage1 = std.array_list.Managed(u16).init(allocator);
+    defer stage1.deinit();
+
+    var stage2 = std.array_list.Managed([]const u21).init(allocator);
+    defer stage2.deinit();
+
+    var block: Block = [_][]const u21{&[_]u21{}} ** block_size;
+    var block_len: u16 = 0;
+
+    for (0..0x110000) |i| {
+        const cp: u21 = @intCast(i);
+        const compat: []const u21 = compat_map.get(cp) orelse &[_]u21{};
+
+        block[block_len] = compat;
+        block_len += 1;
+
+        if (block_len < block_size and cp != 0x10ffff) continue;
+
+        const gop = try blocks_map.getOrPut(block);
+        if (!gop.found_existing) {
+            gop.value_ptr.* = @intCast(stage2.items.len);
+            try stage2.appendSlice(&block);
+        }
+
+        try stage1.append(gop.value_ptr.*);
+        block_len = 0;
+    }
+    // Write out
+    var write_buf: [4096]u8 = undefined;
+    var out_file = try std.fs.cwd().createFile(output_path, .{});
+    defer out_file.close();
+    var writer = out_file.writer(&write_buf);
+
+    try writer.interface.print(
+        \\//! This file is auto-generated. Do not edit.
+        \\
+        \\pub const s1: [{}]u16 = .{{
+    , .{stage1.items.len});
+    for (stage1.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.print(
+        \\
+        \\}};
+        \\
+        \\pub const s2: [{}][]const u21 = .{{
+    , .{stage2.items.len});
+    for (stage2.items) |entry| {
+        try writer.interface.print("&.{any}, ", .{entry});
+    }
+
+    try writer.interface.writeAll(
+        \\};
+    );
+
     try writer.interface.flush();
 }
diff --git a/codegen/fold.zig b/codegen/fold.zig
index 366ed79..c5f54eb 100644
--- a/codegen/fold.zig
+++ b/codegen/fold.zig
@@ -228,26 +228,45 @@ pub fn main() anyerror!void {
         var out_file = try std.fs.cwd().createFile(output_path, .{});
         defer out_file.close();
         var writer = out_file.writer(&write_buf);
-
-        const endian = builtin.cpu.arch.endian();
         // Table metadata.
-        try writer.interface.writeInt(u24, @intCast(codepoint_cutoff), endian);
-        try writer.interface.writeInt(u24, @intCast(multiple_codepoint_start), endian);
-        // Stage 1
-        try writer.interface.writeInt(u16, @intCast(meaningful_stage1.len), endian);
-        try writer.interface.writeAll(meaningful_stage1);
-        // Stage 2
-        try writer.interface.writeInt(u16, @intCast(stage2.len), endian);
-        try writer.interface.writeAll(stage2);
-        // Stage 3
-        try writer.interface.writeInt(u16, @intCast(stage3.len), endian);
-        for (stage3) |offset| try writer.interface.writeInt(i24, offset, endian);
-        // Changes when case folded
-        // Min and max
-        try writer.interface.writeInt(u24, std.mem.min(u21, changes_when_casefolded_exceptions.items), endian);
-        try writer.interface.writeInt(u24, std.mem.max(u21, changes_when_casefolded_exceptions.items), endian);
-        try writer.interface.writeInt(u16, @intCast(changes_when_casefolded_exceptions.items.len), endian);
-        for (changes_when_casefolded_exceptions.items) |cp| try writer.interface.writeInt(u24, cp, endian);
+        try writer.interface.print(
+            \\//! This file is auto-generated. Do not edit.
+            \\
+            \\pub const cutoff: u21 = {};
+            \\pub const cwcf_exceptions_min: u21 = {};
+            \\pub const cwcf_exceptions_max: u21 = {};
+            \\pub const cwcf_exceptions: [{}]u21 = .{{
+        , .{ codepoint_cutoff, std.mem.min(u21, changes_when_casefolded_exceptions.items), std.mem.max(u21, changes_when_casefolded_exceptions.items), changes_when_casefolded_exceptions.items.len });
+        for (changes_when_casefolded_exceptions.items) |cp| try writer.interface.print("{}, ", .{cp});
+
+        try writer.interface.print(
+            \\
+            \\}};
+            \\
+            \\pub const multiple_start: u21 = {};
+            \\pub const stage1: [{}]u8 = .{{
+        , .{ multiple_codepoint_start, meaningful_stage1.len });
+        for (meaningful_stage1) |entry| try writer.interface.print("{}, ", .{entry});
+
+        try writer.interface.print(
+            \\
+            \\}};
+            \\
+            \\pub const stage2: [{}]u8 = .{{
+        , .{stage2.len});
+        for (stage2) |entry| try writer.interface.print("{}, ", .{entry});
+
+        try writer.interface.print(
+            \\
+            \\}};
+            \\
+            \\pub const stage3: [{}]i24 = .{{
+        , .{stage3.len});
+        for (stage3) |entry| try writer.interface.print("{}, ", .{entry});
+
+        try writer.interface.writeAll(
+            \\};
+        );
 
         try writer.interface.flush();
     }
diff --git a/codegen/hangul.zig b/codegen/hangul.zig
index 2e4c175..d7504a9 100644
--- a/codegen/hangul.zig
+++ b/codegen/hangul.zig
@@ -120,12 +120,24 @@ pub fn main() anyerror!void {
     defer out_file.close();
     var writer = out_file.writer(&write_buf);
 
-    const endian = builtin.cpu.arch.endian();
-    try writer.interface.writeInt(u16, @intCast(stage1.items.len), endian);
-    for (stage1.items) |i| try writer.interface.writeInt(u16, i, endian);
-
-    try writer.interface.writeInt(u16, @intCast(stage2.items.len), endian);
-    for (stage2.items) |i| try writer.interface.writeInt(u8, i, endian);
+    try writer.interface.print(
+        \\//! This file is auto-generated. Do not edit.
+        \\
+        \\pub const s1: [{}]u16 = .{{
+    , .{stage1.items.len});
+    for (stage1.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.print(
+        \\
+        \\}};
+        \\
+        \\pub const s2: [{}]u3 = .{{
+    , .{stage2.items.len});
+    for (stage2.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.writeAll(
+        \\};
+    );
 
     try writer.interface.flush();
 }
diff --git a/codegen/normp.zig b/codegen/normp.zig
index eaf6989..343f03e 100644
--- a/codegen/normp.zig
+++ b/codegen/normp.zig
@@ -121,12 +121,24 @@ pub fn main() anyerror!void {
     defer out_file.close();
     var writer = out_file.writer(&write_buf);
 
-    const endian = builtin.cpu.arch.endian();
-    try writer.interface.writeInt(u16, @intCast(stage1.items.len), endian);
-    for (stage1.items) |i| try writer.interface.writeInt(u16, i, endian);
-
-    try writer.interface.writeInt(u16, @intCast(stage2.items.len), endian);
-    for (stage2.items) |i| try writer.interface.writeInt(u8, i, endian);
+    try writer.interface.print(
+        \\//! This file is auto-generated. Do not edit.
+        \\
+        \\pub const s1: [{}]u16 = .{{
+    , .{stage1.items.len});
+    for (stage1.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.print(
+        \\
+        \\}};
+        \\
+        \\pub const s2: [{}]u3 = .{{
+    , .{stage2.items.len});
+    for (stage2.items) |entry| try writer.interface.print("{}, ", .{entry});
+
+    try writer.interface.writeAll(
+        \\};
+    );
 
     try writer.interface.flush();
 }