sparkles.base.text — conformance test cases

This is the tracked conformance ledger for the cell-splitting & width specification. Each row is a normative case: its want width is what kitty's algorithm prescribes. The executable report below measures the implementation via visibleWidth and prints PASS / FAIL per case, then a tally; it is verified by apps/ci against its [Output] block. All cases currently pass. The same assertions also live as library unittests in width.d and grapheme.d.

This is the curated, always-green ledger. For exhaustive differential testing (every code point, the official GraphemeBreakTest.txt / emoji-test.txt corpora, and cross-checks against the kitty and ghostty terminals), see the conformance harness.

Ledger

want = kitty-normative width (in cells).

Case	Input (code points)	want	status
ASCII letter	U+0041	1	✓
CJK ideograph (EAW W)	U+4E16	2	✓
Fullwidth letter (EAW F)	U+FF21	2	✓
base + combining mark (Mn)	U+0041 U+0301	1	✓
zero-width space (Cf)	U+200B	0	✓
flag (regional-indicator pair)	U+1F1FA U+1F1F8	2	✓
emoji + skin-tone modifier	U+1F44D U+1F3FE	2	✓
ZWJ family	U+1F469 U+200D U+1F467	2	✓
emoji base + VS16	U+2764 U+FE0F	2	✓
emoji base bare (ambiguous)	U+2764	1	✓
emoji base + VS15	U+2764 U+FE0E	1	✓
styled text (escapes width 0)	ESC[31m h i ESC[0m	2	✓
base + spacing mark (Mc)	U+0915 U+093E	1	✓
Devanagari syllable	U+0915 U+0940	1	✓
lone spacing mark (Mc)	U+093E	0	✓
lone regional indicator	U+1F1FA	2	✓
lone emoji modifier	U+1F3FE	2	✓
noncharacter	U+FDD0	0	✓

Notes on the trickier rows

Several rows were brought into conformance by the width-class fixes in width.d:

• Spacing marks (Mc) are now zero width (all Marks M* join Cf in the zero-width set), so a Brahmic syllable such as U+0915 U+093E is one 1-cell cluster — not two cells.
• A lone regional indicator is width 2 (EastAsianWidth.txt marks U+1F1E6..U+1F1FF as neutral, but kitty's flag rule overrides them to 2).
• Noncharacters (U+FDD0..U+FDEF, U+xxFFFE, U+xxFFFF) are discarded → 0.
• A lone emoji modifier is width 2, not 0. kitty's prose lists modifiers under zero-width Marks, but U+1F3FB..U+1F3FF have East_Asian_Width = W, and in kitty's priority order Doublewidth outranks Marks, so a modifier in isolation is 2 (it only ever contributes 0 inside a cluster, where the leading emoji already sets the width). sparkles matches this.

Executable report

#!/usr/bin/env dub
/+ dub.sdl:
    name "spec_text_conformance"
    dependency "sparkles:base" version="*"
+/
import std.stdio : writefln;
import sparkles.base.text.grapheme : visibleWidth;

struct Case { string label; string input; size_t want; }

void main()
{
    // `want` is the kitty-normative width, measured via the top-level visibleWidth.
    static immutable Case[] cases = [
        Case("ASCII 'A'",                   "A", 1),
        Case("CJK U+4E16",                  "世", 2),
        Case("Fullwidth U+FF21",            "Ａ", 2),
        Case("base+Mn (A+U+0301)",          "Á", 1),
        Case("ZWSP U+200B",                 "​", 0),
        Case("flag US",                     "\U0001F1FA\U0001F1F8", 2),
        Case("emoji+tone",                  "\U0001F44D\U0001F3FE", 2),
        Case("ZWJ family",                  "\U0001F469‍\U0001F467", 2),
        Case("heart+VS16",                  "❤️", 2),
        Case("heart bare (ambiguous)",      "❤", 1),
        Case("heart+VS15",                  "❤︎", 1),
        Case("styled (escapes=0)",          "\x1b[31mhi\x1b[0m", 2),
        Case("base+Mc (U+0915 U+093E)",     "का", 1),
        Case("syllable U+0915 U+0940",      "की", 1),
        Case("lone Mc U+093E",              "ा", 0),
        Case("lone RI U+1F1FA",             "\U0001F1FA", 2),
        Case("lone emoji modifier U+1F3FE", "\U0001F3FE", 2),
        Case("noncharacter U+FDD0",         "﷐", 0),
    ];

    size_t pass, fail;
    foreach (c; cases)
    {
        const got = visibleWidth(c.input);
        if (got == c.want) { writefln("PASS  %-32s width=%s", c.label, got); ++pass; }
        else { writefln("FAIL  %-32s got=%s want=%s", c.label, got, c.want); ++fail; }
    }
    writefln("---- %s passed, %s failed (of %s) ----", pass, fail, pass + fail);
}

PASS  ASCII 'A'                        width=1
PASS  CJK U+4E16                       width=2
PASS  Fullwidth U+FF21                 width=2
PASS  base+Mn (A+U+0301)               width=1
PASS  ZWSP U+200B                      width=0
PASS  flag US                          width=2
PASS  emoji+tone                       width=2
PASS  ZWJ family                       width=2
PASS  heart+VS16                       width=2
PASS  heart bare (ambiguous)           width=1
PASS  heart+VS15                       width=1
PASS  styled (escapes=0)               width=2
PASS  base+Mc (U+0915 U+093E)          width=1
PASS  syllable U+0915 U+0940           width=1
PASS  lone Mc U+093E                   width=0
PASS  lone RI U+1F1FA                  width=2
PASS  lone emoji modifier U+1F3FE      width=2
PASS  noncharacter U+FDD0              width=0
---- 18 passed, 0 failed (of 18) ----