From 9a0fb96c0c28540493a205b85d1b89d2c9b50f2b Mon Sep 17 00:00:00 2001 From: Jose Colon Rodriguez Date: Fri, 1 Mar 2024 18:51:43 -0400 Subject: Normalizer.eqlIgnoreCase compatibility caseless matching --- build.zig | 20 ++++++++++- codegen/fold.zig | 76 +++++++++++++++++++++++++++++++++++++++ src/FoldData.zig | 48 +++++++++++++++++++++++++ src/NormData.zig | 4 +++ src/Normalizer.zig | 103 ++++++++++++++++++++++++++++++++++++++++++++++++++--- src/main.zig | 17 ++++++--- 6 files changed, 258 insertions(+), 10 deletions(-) create mode 100644 codegen/fold.zig create mode 100644 src/FoldData.zig diff --git a/build.zig b/build.zig index 09edceb..3d4d1c4 100644 --- a/build.zig +++ b/build.zig @@ -88,6 +88,15 @@ pub fn build(b: *std.Build) void { const run_gencat_gen_exe = b.addRunArtifact(gencat_gen_exe); const gencat_gen_out = run_gencat_gen_exe.addOutputFileArg("gencat.bin.z"); + const fold_gen_exe = b.addExecutable(.{ + .name = "fold", + .root_source_file = .{ .path = "codegen/fold.zig" }, + .target = b.host, + .optimize = .Debug, + }); + const run_fold_gen_exe = b.addRunArtifact(fold_gen_exe); + const fold_gen_out = run_fold_gen_exe.addOutputFileArg("fold.bin.z"); + // Modules we provide // Code points const code_point = b.addModule("code_point", .{ @@ -174,6 +183,14 @@ pub fn build(b: *std.Build) void { }); normp_data.addAnonymousImport("normp", .{ .root_source_file = normp_gen_out }); + // Case folding + const fold_data = b.createModule(.{ + .root_source_file = .{ .path = "src/FoldData.zig" }, + .target = target, + .optimize = optimize, + }); + fold_data.addAnonymousImport("fold", .{ .root_source_file = fold_gen_out }); + const norm_data = b.createModule(.{ .root_source_file = .{ .path = "src/NormData.zig" }, .target = target, @@ -184,6 +201,7 @@ pub fn build(b: *std.Build) void { norm_data.addImport("CompatData", compat_data); norm_data.addImport("HangulData", hangul_data); norm_data.addImport("NormPropsData", normp_data); + norm_data.addImport("FoldData", fold_data); const norm = b.addModule("Normalizer", .{ .root_source_file = .{ .path = "src/Normalizer.zig" }, @@ -195,7 +213,7 @@ pub fn build(b: *std.Build) void { norm.addImport("NormData", norm_data); // General Category - const gencat_data = b.createModule(.{ + const gencat_data = b.addModule("GenCatData", .{ .root_source_file = .{ .path = "src/GenCatData.zig" }, .target = target, .optimize = optimize, diff --git a/codegen/fold.zig b/codegen/fold.zig new file mode 100644 index 0000000..7977e61 --- /dev/null +++ b/codegen/fold.zig @@ -0,0 +1,76 @@ +const std = @import("std"); +const builtin = @import("builtin"); +const fmt = std.fmt; +const mem = std.mem; + +pub fn main() !void { + var arena = std.heap.ArenaAllocator.init(std.heap.page_allocator); + defer arena.deinit(); + const allocator = arena.allocator(); + + // Process DerivedEastAsianWidth.txt + var in_file = try std.fs.cwd().openFile("data/unicode/CaseFolding.txt", .{}); + defer in_file.close(); + var in_buf = std.io.bufferedReader(in_file.reader()); + const in_reader = in_buf.reader(); + + 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!"); + + const compressor = std.compress.deflate.compressor; + var out_file = try std.fs.cwd().createFile(output_path, .{}); + defer out_file.close(); + var out_comp = try compressor(allocator, out_file.writer(), .{ .level = .best_compression }); + defer out_comp.deinit(); + const writer = out_comp.writer(); + + const endian = builtin.cpu.arch.endian(); + var line_buf: [4096]u8 = undefined; + + lines: while (try in_reader.readUntilDelimiterOrEof(&line_buf, '\n')) |line| { + if (line.len == 0 or line[0] == '#') continue; + + const no_comment = if (mem.indexOfScalar(u8, line, '#')) |octo| line[0..octo] else line; + + var field_iter = mem.tokenizeSequence(u8, no_comment, "; "); + var cps: [4]u24 = undefined; + var len: usize = 2; + + var i: usize = 0; + while (field_iter.next()) |field| : (i += 1) { + switch (i) { + 0 => cps[0] = try fmt.parseInt(u24, field, 16), + + 1 => { + if (!mem.eql(u8, field, "C") and !mem.eql(u8, field, "F")) continue :lines; + if (mem.eql(u8, field, "F")) len = 3; + }, + + 2 => { + if (len == 3) { + // Full case fold + // std.debug.print("-->{s} {s}\n", .{ line, field }); + var cp_iter = mem.tokenizeScalar(u8, field, ' '); + len = 1; + while (cp_iter.next()) |cp_str| : (len += 1) { + cps[len] = try fmt.parseInt(u24, cp_str, 16); + } + } else { + // Common case fold + cps[1] = try fmt.parseInt(u24, field, 16); + } + }, + + else => {}, + } + } + + try writer.writeInt(u8, @intCast(len), endian); + for (cps[0..len]) |cp| try writer.writeInt(u24, cp, endian); + } + + try writer.writeInt(u16, 0, endian); + try out_comp.flush(); +} diff --git a/src/FoldData.zig b/src/FoldData.zig new file mode 100644 index 0000000..139c677 --- /dev/null +++ b/src/FoldData.zig @@ -0,0 +1,48 @@ +const std = @import("std"); +const builtin = @import("builtin"); +const compress = std.compress; +const mem = std.mem; + +allocator: mem.Allocator, +fold: [][]u21 = undefined, + +const Self = @This(); + +pub fn init(allocator: mem.Allocator) !Self { + const decompressor = compress.deflate.decompressor; + const in_bytes = @embedFile("fold"); + 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, + .fold = try allocator.alloc([]u21, 0x110000), + }; + + @memset(self.fold, &.{}); + + while (true) { + const len: u8 = try reader.readInt(u8, endian); + if (len == 0) break; + const cp = try reader.readInt(u24, endian); + self.fold[cp] = try allocator.alloc(u21, len - 1); + for (0..len - 1) |i| { + self.fold[cp][i] = @intCast(try reader.readInt(u24, endian)); + } + } + + return self; +} + +pub fn deinit(self: *Self) void { + for (self.fold) |slice| self.allocator.free(slice); + self.allocator.free(self.fold); +} + +/// Returns the case fold for `cp`. +pub inline fn caseFold(self: Self, cp: u21) []const u21 { + return self.fold[cp]; +} diff --git a/src/NormData.zig b/src/NormData.zig index 7c2a09b..3c2f614 100644 --- a/src/NormData.zig +++ b/src/NormData.zig @@ -4,6 +4,7 @@ const mem = std.mem; const CanonData = @import("CanonData"); const CccData = @import("CombiningData"); const CompatData = @import("CompatData"); +const FoldData = @import("FoldData"); const HangulData = @import("HangulData"); const NormPropsData = @import("NormPropsData"); @@ -12,6 +13,7 @@ ccc_data: CccData, compat_data: CompatData, hangul_data: HangulData, normp_data: NormPropsData, +fold_data: FoldData, const Self = @This(); @@ -20,6 +22,7 @@ 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), + .fold_data = try FoldData.init(allocator), .hangul_data = try HangulData.init(allocator), .normp_data = try NormPropsData.init(allocator), }; @@ -30,5 +33,6 @@ pub fn deinit(self: *Self) void { self.ccc_data.deinit(); self.compat_data.deinit(); self.hangul_data.deinit(); + self.fold_data.deinit(); self.normp_data.deinit(); } diff --git a/src/Normalizer.zig b/src/Normalizer.zig index abe35e5..c68b2ec 100644 --- a/src/Normalizer.zig +++ b/src/Normalizer.zig @@ -343,7 +343,102 @@ test "nfkd !ASCII / alloc" { try testing.expectEqualStrings("He\u{301}llo World! \u{3a5}\u{301}", result.slice); } -// Composition utilities. +fn caseFold( + self: Self, + allocator: mem.Allocator, + cps: []const u21, +) ![]const u21 { + var cfcps = std.ArrayList(u21).init(allocator); + defer cfcps.deinit(); + + for (cps) |cp| { + const cf = self.norm_data.fold_data.caseFold(cp); + + if (cf.len == 0) { + try cfcps.append(cp); + } else { + try cfcps.appendSlice(cf); + } + } + + return try cfcps.toOwnedSlice(); +} + +fn nfkdCodePoints( + self: Self, + allocator: mem.Allocator, + cps: []const u21, +) ![]u21 { + var dcp_list = std.ArrayList(u21).init(allocator); + defer dcp_list.deinit(); + + var dc_buf: [18]u21 = undefined; + + for (cps) |cp| { + const dc = self.decompose(cp, .nfkd, &dc_buf); + + if (dc.form == .same) { + try dcp_list.append(cp); + } else { + try dcp_list.appendSlice(dc.cps); + } + } + + self.canonicalSort(dcp_list.items); + + return try dcp_list.toOwnedSlice(); +} + +pub fn eqlIgnoreCase( + self: Self, + allocator: mem.Allocator, + a: []const u8, + b: []const u8, +) !bool { + if (ascii.isAsciiOnly(a) and ascii.isAsciiOnly(b)) return std.ascii.eqlIgnoreCase(a, b); + + const nfd_a = try self.nfxdCodePoints(allocator, a, .nfd); + defer allocator.free(nfd_a); + const cf_nfd_a = try self.caseFold(allocator, nfd_a); + defer allocator.free(cf_nfd_a); + const nfkd_cf_nfd_a = try self.nfkdCodePoints(allocator, cf_nfd_a); + defer allocator.free(nfkd_cf_nfd_a); + const cf_nfkd_cf_nfd_a = try self.caseFold(allocator, nfkd_cf_nfd_a); + defer allocator.free(cf_nfkd_cf_nfd_a); + const nfkd_cf_nfkd_cf_nfd_a = try self.nfkdCodePoints(allocator, cf_nfkd_cf_nfd_a); + defer allocator.free(nfkd_cf_nfkd_cf_nfd_a); + + const nfd_b = try self.nfxdCodePoints(allocator, b, .nfd); + defer allocator.free(nfd_b); + const cf_nfd_b = try self.caseFold(allocator, nfd_b); + defer allocator.free(cf_nfd_b); + const nfkd_cf_nfd_b = try self.nfkdCodePoints(allocator, cf_nfd_b); + defer allocator.free(nfkd_cf_nfd_b); + const cf_nfkd_cf_nfd_b = try self.caseFold(allocator, nfkd_cf_nfd_b); + defer allocator.free(cf_nfkd_cf_nfd_b); + const nfkd_cf_nfkd_cf_nfd_b = try self.nfkdCodePoints(allocator, cf_nfkd_cf_nfd_b); + defer allocator.free(nfkd_cf_nfkd_cf_nfd_b); + + return mem.eql(u21, nfkd_cf_nfkd_cf_nfd_a, nfkd_cf_nfkd_cf_nfd_b); +} + +test "eqlIgnoreCase" { + const allocator = testing.allocator; + var data = try NormData.init(allocator); + defer data.deinit(); + var n = Self{ .norm_data = &data }; + + try testing.expect(try n.eqlIgnoreCase(allocator, "ascii only!", "ASCII Only!")); + + const a = "Héllo World! \u{3d3}"; + const b = "He\u{301}llo World! \u{3a5}\u{301}"; + try testing.expect(try n.eqlIgnoreCase(allocator, a, b)); + + const c = "He\u{301}llo World! \u{3d2}\u{301}"; + try testing.expect(try n.eqlIgnoreCase(allocator, a, c)); +} + +// Composition (NFC, NFKC) fn isHangul(self: Self, cp: u21) bool { return cp >= 0x1100 and self.norm_data.hangul_data.syllable(cp) != .none; @@ -504,11 +599,11 @@ test "nfkc" { try testing.expectEqualStrings("Complex char: \u{038E}", result.slice); } -/// Tests for equality of `a` and `b` after normalizing to NFD. +/// Tests for equality of `a` and `b` after normalizing to NFC. pub fn eql(self: Self, allocator: mem.Allocator, a: []const u8, b: []const u8) !bool { - var norm_result_a = try self.nfd(allocator, a); + var norm_result_a = try self.nfc(allocator, a); defer norm_result_a.deinit(); - var norm_result_b = try self.nfd(allocator, b); + var norm_result_b = try self.nfc(allocator, b); defer norm_result_b.deinit(); return mem.eql(u8, norm_result_a.slice, norm_result_b.slice); diff --git a/src/main.zig b/src/main.zig index 59a0fbc..a5afa66 100644 --- a/src/main.zig +++ b/src/main.zig @@ -49,21 +49,28 @@ pub fn main() !void { var iter = std.mem.splitScalar(u8, input, '\n'); var result: usize = 0; + var prev_line: []const u8 = ""; // var result: isize = 0; var timer = try std.time.Timer.start(); // while (iter.next()) |cp| result += codePointWidth(@intCast(cp.code)); // while (iter.next()) |_| result += 1; // while (iter.next()) |line| result += strWidth(line, &data); - while (iter.next()) |line| { - const nfc = try n.nfkc(allocator, line); - result += nfc.slice.len; - // nfc.deinit(); - } + // while (iter.next()) |line| { + // const nfc = try n.nfkc(allocator, line); + // result += nfc.slice.len; + // // nfc.deinit(); + // } // while (iter.next()) |cp| { // if (cp.code == 'É') std.debug.print("`{u}` Gc: {s}\n", .{ cp.code, @tagName(gencat_data.gc(cp.code)) }); // result += 1; // } + while (iter.next()) |line| { + if (try n.eqlIgnoreCase(allocator, prev_line, line)) { + result += line.len; + } + prev_line = line; + } std.debug.print("result: {}, took: {}\n", .{ result, timer.lap() / std.time.ns_per_ms }); } -- cgit v1.2.3