From 32c68059a05dde8a57a330db6d14a32506081516 Mon Sep 17 00:00:00 2001
From: Jose Colon Rodriguez
Date: Tue, 27 Feb 2024 09:26:40 -0400
Subject: Using HangulData in NormData

---
 src/HangulData.zig | 52 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 src/NormData.zig   |  4 ++++
 src/Normalizer.zig | 23 +++++++++++------------
 3 files changed, 67 insertions(+), 12 deletions(-)
 create mode 100644 src/HangulData.zig

(limited to 'src')

diff --git a/src/HangulData.zig b/src/HangulData.zig
new file mode 100644
index 0000000..4d80c99
--- /dev/null
+++ b/src/HangulData.zig
@@ -0,0 +1,52 @@
+const std = @import("std");
+const builtin = @import("builtin");
+const compress = std.compress;
+const mem = std.mem;
+const testing = std.testing;
+
+pub const Syllable = enum {
+    none,
+    L,
+    LV,
+    LVT,
+    V,
+    T,
+};
+
+allocator: mem.Allocator,
+s1: []u16 = undefined,
+s2: []Syllable = undefined,
+
+const Self = @This();
+
+pub fn init(allocator: mem.Allocator) !Self {
+    const decompressor = compress.deflate.decompressor;
+    const in_bytes = @embedFile("hangul");
+    var in_fbs = std.io.fixedBufferStream(in_bytes);
+    var in_decomp = try decompressor(allocator, in_fbs.reader(), null);
+    defer in_decomp.deinit();
+    var reader = in_decomp.reader();
+
+    const endian = builtin.cpu.arch.endian();
+    var self = Self{ .allocator = allocator };
+
+    const stage_1_len: u16 = try reader.readInt(u16, endian);
+    self.s1 = try allocator.alloc(u16, stage_1_len);
+    for (0..stage_1_len) |i| self.s1[i] = try reader.readInt(u16, endian);
+
+    const stage_2_len: u16 = try reader.readInt(u16, endian);
+    self.s2 = try allocator.alloc(Syllable, stage_2_len);
+    for (0..stage_2_len) |i| self.s2[i] = @enumFromInt(try reader.readInt(u8, endian));
+
+    return self;
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.s1);
+    self.allocator.free(self.s2);
+}
+
+/// Returns the Hangul syllable type for `cp`.
+pub inline fn syllable(self: Self, cp: u21) Syllable {
+    return self.s2[self.s1[cp >> 8] + (cp & 0xff)];
+}
diff --git a/src/NormData.zig b/src/NormData.zig
index 83110f0..8923382 100644
--- a/src/NormData.zig
+++ b/src/NormData.zig
@@ -4,10 +4,12 @@ const mem = std.mem;
 const CanonData = @import("CanonData");
 const CccData = @import("CombiningData");
 const CompatData = @import("CompatData");
+const HangulData = @import("HangulData");
 
 canon_data: CanonData,
 ccc_data: CccData,
 compat_data: CompatData,
+hangul_data: HangulData,
 
 const Self = @This();
 
@@ -16,6 +18,7 @@ pub fn init(allocator: std.mem.Allocator) !Self {
         .canon_data = try CanonData.init(allocator),
         .ccc_data = try CccData.init(allocator),
         .compat_data = try CompatData.init(allocator),
+        .hangul_data = try HangulData.init(allocator),
     };
 }
 
@@ -23,4 +26,5 @@ pub fn deinit(self: *Self) void {
     self.canon_data.deinit();
     self.ccc_data.deinit();
     self.compat_data.deinit();
+    self.hangul_data.deinit();
 }
diff --git a/src/Normalizer.zig b/src/Normalizer.zig
index 1434043..0670cae 100644
--- a/src/Normalizer.zig
+++ b/src/Normalizer.zig
@@ -7,7 +7,6 @@ const testing = std.testing;
 
 const CodePointIterator = @import("code_point").Iterator;
 const case_fold_map = @import("ziglyph").case_folding;
-const hangul_map = @import("ziglyph").hangul;
 const norm_props = @import("ziglyph").normalization_props;
 
 pub const NormData = @import("NormData");
@@ -17,9 +16,9 @@ norm_data: *NormData,
 const Self = @This();
 
 // Hangul processing utilities.
-fn isHangulPrecomposed(cp: u21) bool {
-    if (hangul_map.syllableType(cp)) |kind| return kind == .LV or kind == .LVT;
-    return false;
+fn isHangulPrecomposed(self: Self, cp: u21) bool {
+    const kind = self.norm_data.hangul_data.syllable(cp);
+    return kind == .LV or kind == .LVT;
 }
 
 const SBase: u21 = 0xAC00;
@@ -117,7 +116,7 @@ pub fn decompose(self: Self, cp: u21, form: Form) Decomp {
     }
 
     // Hangul precomposed syllable full decomposition.
-    if (isHangulPrecomposed(cp)) {
+    if (self.isHangulPrecomposed(cp)) {
         const cps = decomposeHangul(cp);
         @memcpy(dc.cps[0..cps.len], &cps);
         return dc;
@@ -335,12 +334,12 @@ test "nfkd !ASCII / alloc" {
 
 // Composition utilities.
 
-fn isHangul(cp: u21) bool {
-    return cp >= 0x1100 and hangul_map.syllableType(cp) != null;
+fn isHangul(self: Self, cp: u21) bool {
+    return cp >= 0x1100 and self.norm_data.hangul_data.syllable(cp) != .none;
 }
 
 fn isNonHangulStarter(self: Self, cp: u21) bool {
-    return !isHangul(cp) and self.norm_data.ccc_data.isStarter(cp);
+    return !self.isHangul(cp) and self.norm_data.ccc_data.isStarter(cp);
 }
 
 /// Normalizes `str` to NFC.
@@ -395,7 +394,7 @@ fn nfxc(self: Self, allocator: std.mem.Allocator, str: []const u8, form: Form) !
                         for (d_list.items[(j + 1)..i]) |B| {
                             const cc_B = self.norm_data.ccc_data.ccc(B);
                             // Check for blocking conditions.
-                            if (isHangul(C)) {
+                            if (self.isHangul(C)) {
                                 if (cc_B != 0 or self.isNonHangulStarter(B)) continue :block_check;
                             }
                             if (cc_B >= cc_C) continue :block_check;
@@ -414,9 +413,9 @@ fn nfxc(self: Self, allocator: std.mem.Allocator, str: []const u8, form: Form) !
                 const L = d_list.items[sidx];
                 var processed_hangul = false;
 
-                if (isHangul(L) and isHangul(C)) {
-                    const l_stype = hangul_map.syllableType(L).?;
-                    const c_stype = hangul_map.syllableType(C).?;
+                if (self.isHangul(L) and self.isHangul(C)) {
+                    const l_stype = self.norm_data.hangul_data.syllable(L);
+                    const c_stype = self.norm_data.hangul_data.syllable(C);
 
                     if (l_stype == .LV and c_stype == .T) {
                         // LV, T
-- 
cgit v1.2.3