unicode: delete props.zig and clean up symbols deps too (#8814)

Follow up to #8810

Same reasoning.

src/benchmark/IsSymbol.zig

@@ -10,7 +10,8 @@ const Allocator = std.mem.Allocator;
 const Benchmark = @import("Benchmark.zig");
 const options = @import("options.zig");
 const UTF8Decoder = @import("../terminal/UTF8Decoder.zig");
-const symbols = @import("../unicode/symbols.zig");
+const symbols = @import("../unicode/symbols_ziglyph.zig");
+const symbols_table = @import("../unicode/symbols_table.zig").table;
 
 const log = std.log.scoped(.@"is-symbol-bench");
 
@@ -127,7 +128,7 @@ fn stepTable(ptr: *anyopaque) Benchmark.Error!void {
             const cp_, const consumed = d.next(c);
             assert(consumed);
             if (cp_) |cp| {
-                std.mem.doNotOptimizeAway(symbols.table.get(cp));
+                std.mem.doNotOptimizeAway(symbols_table.get(cp));
             }
         }
     }

src/build/UnicodeTables.zig

@@ -15,7 +15,7 @@ pub fn init(b: *std.Build) !UnicodeTables {
     const props_exe = b.addExecutable(.{
         .name = "props-unigen",
         .root_module = b.createModule(.{
-            .root_source_file = b.path("src/unicode/props.zig"),
+            .root_source_file = b.path("src/unicode/props_ziglyph.zig"),
             .target = b.graph.host,
             .strip = false,
             .omit_frame_pointer = false,
@@ -26,7 +26,7 @@ pub fn init(b: *std.Build) !UnicodeTables {
     const symbols_exe = b.addExecutable(.{
         .name = "symbols-unigen",
         .root_module = b.createModule(.{
-            .root_source_file = b.path("src/unicode/symbols.zig"),
+            .root_source_file = b.path("src/unicode/symbols_ziglyph.zig"),
             .target = b.graph.host,
             .strip = false,
             .omit_frame_pointer = false,

src/main_ghostty.zig

@@ -191,6 +191,8 @@ test {
     _ = @import("simd/main.zig");
     _ = @import("synthetic/main.zig");
     _ = @import("unicode/main.zig");
+    _ = @import("unicode/props_ziglyph.zig");
+    _ = @import("unicode/symbols_ziglyph.zig");
 
     // Extra
     _ = @import("extra/bash.zig");

src/renderer/cell.zig

@@ -6,7 +6,7 @@ const terminal = @import("../terminal/main.zig");
 const renderer = @import("../renderer.zig");
 const shaderpkg = renderer.Renderer.API.shaders;
 const ArrayListCollection = @import("../datastruct/array_list_collection.zig").ArrayListCollection;
-const symbols = @import("../unicode/symbols.zig").table;
+const symbols = @import("../unicode/symbols_table.zig").table;
 
 /// The possible cell content keys that exist.
 pub const Key = enum {

src/unicode/grapheme.zig

@@ -1,7 +1,6 @@
 const std = @import("std");
-const props = @import("props.zig");
-const GraphemeBoundaryClass = props.GraphemeBoundaryClass;
-const table = props.table;
+const table = @import("props_table.zig").table;
+const GraphemeBoundaryClass = @import("Properties.zig").GraphemeBoundaryClass;
 
 /// Determines if there is a grapheme break between two codepoints. This
 /// must be called sequentially maintaining the state between calls.

src/unicode/main.zig

@@ -7,7 +7,5 @@ pub const graphemeBreak = grapheme.graphemeBreak;
 pub const GraphemeBreakState = grapheme.BreakState;
 
 test {
-    _ = @import("props_ziglyph.zig");
-    _ = @import("symbols.zig");
     @import("std").testing.refAllDecls(@This());
 }

src/unicode/props.zig

@@ -1,181 +0,0 @@
-const props = @This();
-const std = @import("std");
-const assert = std.debug.assert;
-const ziglyph = @import("ziglyph");
-const lut = @import("lut.zig");
-
-/// The lookup tables for Ghostty.
-pub const table = table: {
-    // This is only available after running main() below as part of the Ghostty
-    // build.zig, but due to Zig's lazy analysis we can still reference it here.
-    const generated = @import("unicode_tables").Tables(Properties);
-    const Tables = lut.Tables(Properties);
-    break :table Tables{
-        .stage1 = &generated.stage1,
-        .stage2 = &generated.stage2,
-        .stage3 = &generated.stage3,
-    };
-};
-
-/// Property set per codepoint that Ghostty cares about.
-///
-/// Adding to this lets you find new properties but also potentially makes
-/// our lookup tables less efficient. Any changes to this should run the
-/// benchmarks in src/bench to verify that we haven't regressed.
-pub const Properties = struct {
-    /// Codepoint width. We clamp to [0, 2] since Ghostty handles control
-    /// characters and we max out at 2 for wide characters (i.e. 3-em dash
-    /// becomes a 2-em dash).
-    width: u2 = 0,
-
-    /// Grapheme boundary class.
-    grapheme_boundary_class: GraphemeBoundaryClass = .invalid,
-
-    // Needed for lut.Generator
-    pub fn eql(a: Properties, b: Properties) bool {
-        return a.width == b.width and
-            a.grapheme_boundary_class == b.grapheme_boundary_class;
-    }
-
-    // Needed for lut.Generator
-    pub fn format(
-        self: Properties,
-        comptime layout: []const u8,
-        opts: std.fmt.FormatOptions,
-        writer: anytype,
-    ) !void {
-        _ = layout;
-        _ = opts;
-        try std.fmt.format(writer,
-            \\.{{
-            \\    .width= {},
-            \\    .grapheme_boundary_class= .{s},
-            \\}}
-        , .{
-            self.width,
-            @tagName(self.grapheme_boundary_class),
-        });
-    }
-};
-
-/// Possible grapheme boundary classes. This isn't an exhaustive list:
-/// we omit control, CR, LF, etc. because in Ghostty's usage that are
-/// impossible because they're handled by the terminal.
-pub const GraphemeBoundaryClass = enum(u4) {
-    invalid,
-    L,
-    V,
-    T,
-    LV,
-    LVT,
-    prepend,
-    extend,
-    zwj,
-    spacing_mark,
-    regional_indicator,
-    extended_pictographic,
-    extended_pictographic_base, // \p{Extended_Pictographic} & \p{Emoji_Modifier_Base}
-    emoji_modifier, // \p{Emoji_Modifier}
-
-    /// Gets the grapheme boundary class for a codepoint. This is VERY
-    /// SLOW. The use case for this is only in generating lookup tables.
-    pub fn init(cp: u21) GraphemeBoundaryClass {
-        // We special-case modifier bases because we should not break
-        // if a modifier isn't next to a base.
-        if (ziglyph.emoji.isEmojiModifierBase(cp)) {
-            assert(ziglyph.emoji.isExtendedPictographic(cp));
-            return .extended_pictographic_base;
-        }
-
-        if (ziglyph.emoji.isEmojiModifier(cp)) return .emoji_modifier;
-        if (ziglyph.emoji.isExtendedPictographic(cp)) return .extended_pictographic;
-        if (ziglyph.grapheme_break.isL(cp)) return .L;
-        if (ziglyph.grapheme_break.isV(cp)) return .V;
-        if (ziglyph.grapheme_break.isT(cp)) return .T;
-        if (ziglyph.grapheme_break.isLv(cp)) return .LV;
-        if (ziglyph.grapheme_break.isLvt(cp)) return .LVT;
-        if (ziglyph.grapheme_break.isPrepend(cp)) return .prepend;
-        if (ziglyph.grapheme_break.isExtend(cp)) return .extend;
-        if (ziglyph.grapheme_break.isZwj(cp)) return .zwj;
-        if (ziglyph.grapheme_break.isSpacingmark(cp)) return .spacing_mark;
-        if (ziglyph.grapheme_break.isRegionalIndicator(cp)) return .regional_indicator;
-
-        // This is obviously not INVALID invalid, there is SOME grapheme
-        // boundary class for every codepoint. But we don't care about
-        // anything that doesn't fit into the above categories.
-        return .invalid;
-    }
-
-    /// Returns true if this is an extended pictographic type. This
-    /// should be used instead of comparing the enum value directly
-    /// because we classify multiple.
-    pub fn isExtendedPictographic(self: GraphemeBoundaryClass) bool {
-        return switch (self) {
-            .extended_pictographic,
-            .extended_pictographic_base,
-            => true,
-
-            else => false,
-        };
-    }
-};
-
-pub fn get(cp: u21) Properties {
-    const zg_width = ziglyph.display_width.codePointWidth(cp, .half);
-
-    return .{
-        .width = @intCast(@min(2, @max(0, zg_width))),
-        .grapheme_boundary_class = .init(cp),
-    };
-}
-
-/// Runnable binary to generate the lookup tables and output to stdout.
-pub fn main() !void {
-    var arena_state = std.heap.ArenaAllocator.init(std.heap.page_allocator);
-    defer arena_state.deinit();
-    const alloc = arena_state.allocator();
-
-    const gen: lut.Generator(
-        Properties,
-        struct {
-            pub fn get(ctx: @This(), cp: u21) !Properties {
-                _ = ctx;
-                return props.get(cp);
-            }
-
-            pub fn eql(ctx: @This(), a: Properties, b: Properties) bool {
-                _ = ctx;
-                return a.eql(b);
-            }
-        },
-    ) = .{};
-
-    const t = try gen.generate(alloc);
-    defer alloc.free(t.stage1);
-    defer alloc.free(t.stage2);
-    defer alloc.free(t.stage3);
-    try t.writeZig(std.io.getStdOut().writer());
-
-    // Uncomment when manually debugging to see our table sizes.
-    // std.log.warn("stage1={} stage2={} stage3={}", .{
-    //     t.stage1.len,
-    //     t.stage2.len,
-    //     t.stage3.len,
-    // });
-}
-
-// This is not very fast in debug modes, so its commented by default.
-// IMPORTANT: UNCOMMENT THIS WHENEVER MAKING CODEPOINTWIDTH CHANGES.
-// test "unicode props: tables match ziglyph" {
-//     const testing = std.testing;
-//
-//     const min = 0xFF + 1; // start outside ascii
-//     for (min..std.math.maxInt(u21)) |cp| {
-//         const t = table.get(@intCast(cp));
-//         const zg = @min(2, @max(0, ziglyph.display_width.codePointWidth(@intCast(cp), .half)));
-//         if (t.width != zg) {
-//             std.log.warn("mismatch cp=U+{x} t={} zg={}", .{ cp, t, zg });
-//             try testing.expect(false);
-//         }
-//     }
-// }

src/unicode/props_ziglyph.zig

@@ -40,7 +40,7 @@ pub fn get(cp: u21) Properties {
     const zg_width = ziglyph.display_width.codePointWidth(cp, .half);
     return .{
         .width = @intCast(@min(2, @max(0, zg_width))),
-        .grapheme_boundary_class = .init(cp),
+        .grapheme_boundary_class = graphemeBoundaryClass(cp),
     };
 }
 

src/unicode/symbols_table.zig

@@ -0,0 +1,17 @@
+const lut = @import("lut.zig");
+
+/// The lookup tables for Ghostty.
+pub const table = table: {
+    // This is only available after running a generator as part of the Ghostty
+    // build.zig process, but due to Zig's lazy analysis we can still reference
+    // it here.
+    //
+    // An example process is the `main` in `symbols_ziglyph.zig`
+    const generated = @import("symbols_tables").Tables(bool);
+    const Tables = lut.Tables(bool);
+    break :table Tables{
+        .stage1 = &generated.stage1,
+        .stage2 = &generated.stage2,
+        .stage3 = &generated.stage3,
+    };
+};

src/unicode/symbols_ziglyph.zig

@@ -4,19 +4,6 @@ const assert = std.debug.assert;
 const ziglyph = @import("ziglyph");
 const lut = @import("lut.zig");
 
-/// The lookup tables for Ghostty.
-pub const table = table: {
-    // This is only available after running main() below as part of the Ghostty
-    // build.zig, but due to Zig's lazy analysis we can still reference it here.
-    const generated = @import("symbols_tables").Tables(bool);
-    const Tables = lut.Tables(bool);
-    break :table Tables{
-        .stage1 = &generated.stage1,
-        .stage2 = &generated.stage2,
-        .stage3 = &generated.stage3,
-    };
-};
-
 /// Returns true of the codepoint is a "symbol-like" character, which
 /// for now we define as anything in a private use area and anything
 /// in several unicode blocks:
@@ -82,6 +69,7 @@ test "unicode symbols: tables match ziglyph" {
     if (std.valgrind.runningOnValgrind() > 0) return error.SkipZigTest;
 
     const testing = std.testing;
+    const table = @import("symbols_table.zig").table;
 
     for (0..std.math.maxInt(u21)) |cp| {
         const t = table.get(@intCast(cp));