Collapse README bench block to a per-language bar chart

The auto-generated block was a TS-only expanded breakdown (per-issue table,
21-fix list, token table) — that doesn't scale as Monogram adds languages.
Replace it with one compact ASCII bar chart, one row per language (Monogram
vs official, % of that language's documented official-grammar bugs handled),
so new languages slot in as single rows.

- buildBenchMarkdown takes a per-language list and renders bars; TypeScript is
  populated, JavaScript is a pending placeholder until it gets the oracle bench.
- The per-issue / token detail is dropped from the README (still printed when
  you run the bench directly).

Author: johnsoncodehk

README.md

@@ -56,32 +56,24 @@ The valid-code/bidirectional numbers are the grammar's correctness proof; the hi
 
 ### Head-to-head: highlighter correctness vs the official grammar
 
-The numbers below are **auto-generated** by [`test/highlight-bench.ts`](test/highlight-bench.ts) — it grades *both* the official TextMate grammar and Monogram's against a neutral `tsc` oracle, so the comparison is absolute (each has a real distance from 100%), not relative.
+**Auto-generated** by [`test/highlight-bench.ts`](test/highlight-bench.ts) — one row per language, more as Monogram gains them:
 
 <!-- bench:start -->
-<!-- generated by `node test/highlight-bench.ts --write-readme` — do not edit by hand -->
+<!-- generated by `npm run bench:readme` — do not edit by hand -->
 
-Both grammars are graded against a neutral **`tsc`** oracle (the parser only — never
-Monogram's own output, never the official grammar). Ceiling is 100%.
+Each bar = **% of that language's documented official-grammar bugs** the highlighter renders
+correctly, graded against a neutral `tsc` oracle (100% = all of them). Monogram derives its
+highlighter from its conformance-proven parser; the official one is hand-written regex.
 
-**On the documented official-grammar bug ledger** — the 49 [`microsoft/TypeScript-TmLanguage`](https://github.com/microsoft/TypeScript-TmLanguage/issues) issues in [`test/issue-cases.ts`](test/issue-cases.ts) (49 adjudicable by a syntactic oracle), graded per issue:
-
-| grammar | handles the bug correctly |
-|---|---|
-| Official TextMate | 38.8% (19/49) |
-| **Monogram** | **77.6% (38/49)** |
-
-Independently-verified **Monogram fixes** — official structurally wrong, Monogram right: **21** issues (#815, #853, #859, #873, #883, #891, #973, #978, #981, #983, #1014, #1020, #1025, #1040, #1041, #1043, #1050, #1051, #1053, #1056, #1059). Monogram regressions vs official: 2 (#819, #876).
-
-**Token-level role accuracy** (every parser-classifiable token):
-
-| corpus | Official | Monogram |
-|---|---|---|
-| documented bug ledger (534 tokens) | 83.5% | 94.2% |
-
-<sub>Parser-conformance row omitted — run with the TS corpus cloned to `/tmp/ts-repo` to include it.</sub>
+```
+TypeScript
+  Monogram  █████████████████░░░░░  78%  (38/49)
+  official  █████████░░░░░░░░░░░░░  39%  (19/49)
+JavaScript
+  pending — JS not yet on the neutral-oracle bench (see ROADMAP)
+```
 
-<sub>Regenerate: `node test/highlight-bench.ts --write-readme`. Neutral oracle + frozen scope→role table: [`test/scope-roles.ts`](test/scope-roles.ts).</sub>
+<sub>TypeScript = 49 oracle-adjudicable open [`microsoft/TypeScript-TmLanguage`](https://github.com/microsoft/TypeScript-TmLanguage/issues) issues ([`test/issue-cases.ts`](test/issue-cases.ts)) — 21 of Monogram's wins are fixes the official grammar gets *structurally* wrong. Per-issue breakdown: `node test/highlight-bench.ts`. Regenerate: `npm run bench:readme`.</sub>
 <!-- bench:end -->
 
 > **The other side of the ledger (honesty check).** On the *broad* TS parser-conformance corpus — not just the documented bugs — raw token-role accuracy flips: the official grammar's decade of hardening leads (~99% vs ~96%). Monogram's gap there is specific backlog (Unicode identifiers, qualified type names), **not** the ambiguity class above. Clone the TS corpus to `/tmp/ts-repo` and run `node test/highlight-bench.ts` to see both corpora.

test/highlight-bench.ts

@@ -572,42 +572,43 @@ function reportByIssue(perInput: InputResult[]): IssueStats {
   return { total: all.length, adj: D, offPass, monoPass, fixes, regress, bothFail };
 }
 
-// ── README auto-generation: build the marker block from the computed summaries ─
-function buildBenchMarkdown(parser: BenchSummary | null, adv: BenchSummary | null, iss: IssueStats | null): string {
-  const p = (n: number, d: number) => (d === 0 ? 'n/a' : ((n / d) * 100).toFixed(1) + '%');
+// ── README auto-generation: one compact bar chart, one row per language ────────
+// Deliberately NOT expanded (no per-issue tables): this scales to many languages.
+interface LangResult { name: string; issue: IssueStats | null; note?: string }
+
+const BAR_W = 22;
+function bar(frac: number): string {
+  const f = Math.max(0, Math.min(BAR_W, Math.round(frac * BAR_W)));
+  return '█'.repeat(f) + '░'.repeat(BAR_W - f);
+}
+
+function buildBenchMarkdown(langs: LangResult[]): string {
   const out: string[] = [];
-  out.push('<!-- generated by `node test/highlight-bench.ts --write-readme` — do not edit by hand -->');
+  out.push('<!-- generated by `npm run bench:readme` — do not edit by hand -->');
   out.push('');
-  out.push('Both grammars are graded against a neutral **`tsc`** oracle (the parser only — never');
-  out.push("Monogram's own output, never the official grammar). Ceiling is 100%.");
-  if (iss) {
-    out.push('');
-    out.push(`**On the documented official-grammar bug ledger** — the ${iss.total} [\`microsoft/TypeScript-TmLanguage\`](https://github.com/microsoft/TypeScript-TmLanguage/issues) issues in [\`test/issue-cases.ts\`](test/issue-cases.ts) (${iss.adj} adjudicable by a syntactic oracle), graded per issue:`);
-    out.push('');
-    out.push('| grammar | handles the bug correctly |');
-    out.push('|---|---|');
-    out.push(`| Official TextMate | ${p(iss.offPass, iss.adj)} (${iss.offPass}/${iss.adj}) |`);
-    out.push(`| **Monogram** | **${p(iss.monoPass, iss.adj)} (${iss.monoPass}/${iss.adj})** |`);
-    out.push('');
-    out.push(`Independently-verified **Monogram fixes** — official structurally wrong, Monogram right: **${iss.fixes.length}** issues (${iss.fixes.map((n) => '#' + n).join(', ')}).` + (iss.regress.length ? ` Monogram regressions vs official: ${iss.regress.length} (${iss.regress.map((n) => '#' + n).join(', ')}).` : ''));
-  }
-  const tokRows: string[] = [];
-  if (adv) tokRows.push(`| documented bug ledger (${adv.token.n} tokens) | ${p(adv.token.oRole, adv.token.n)} | ${p(adv.token.mRole, adv.token.n)} |`);
-  if (parser) tokRows.push(`| TS parser conformance (${parser.token.n} tokens) | ${p(parser.token.oRole, parser.token.n)} | ${p(parser.token.mRole, parser.token.n)} |`);
-  if (tokRows.length) {
-    out.push('');
-    out.push('**Token-level role accuracy** (every parser-classifiable token):');
-    out.push('');
-    out.push('| corpus | Official | Monogram |');
-    out.push('|---|---|---|');
-    out.push(...tokRows);
-  }
-  if (!parser) {
-    out.push('');
-    out.push('<sub>Parser-conformance row omitted — run with the TS corpus cloned to `/tmp/ts-repo` to include it.</sub>');
+  out.push("Each bar = **% of that language's documented official-grammar bugs** the highlighter renders");
+  out.push('correctly, graded against a neutral `tsc` oracle (100% = all of them). Monogram derives its');
+  out.push('highlighter from its conformance-proven parser; the official one is hand-written regex.');
+  out.push('');
+  out.push('```');
+  for (const L of langs) {
+    out.push(L.name);
+    if (L.issue && L.issue.adj > 0) {
+      const m = L.issue.monoPass / L.issue.adj;
+      const o = L.issue.offPass / L.issue.adj;
+      out.push(`  Monogram  ${bar(m)}  ${Math.round(m * 100)}%  (${L.issue.monoPass}/${L.issue.adj})`);
+      out.push(`  official  ${bar(o)}  ${Math.round(o * 100)}%  (${L.issue.offPass}/${L.issue.adj})`);
+    } else {
+      out.push(`  ${L.note ?? 'pending — not yet on the neutral-oracle bench'}`);
+    }
   }
+  out.push('```');
   out.push('');
-  out.push('<sub>Regenerate: `node test/highlight-bench.ts --write-readme`. Neutral oracle + frozen scope→role table: [`test/scope-roles.ts`](test/scope-roles.ts).</sub>');
+  const ts = langs.find((l) => l.issue && l.issue.adj > 0);
+  const ctx = ts
+    ? `TypeScript = ${ts.issue!.adj} oracle-adjudicable open [\`microsoft/TypeScript-TmLanguage\`](https://github.com/microsoft/TypeScript-TmLanguage/issues) issues ([\`test/issue-cases.ts\`](test/issue-cases.ts)) — ${ts.issue!.fixes.length} of Monogram's wins are fixes the official grammar gets *structurally* wrong. `
+    : '';
+  out.push(`<sub>${ctx}Per-issue breakdown: \`node test/highlight-bench.ts\`. Regenerate: \`npm run bench:readme\`.</sub>`);
   return out.join('\n');
 }
 
@@ -649,8 +650,6 @@ async function allTs(dir: string): Promise<string[]> {
 const WRITE_README = argv.includes('--write-readme');
 const WHICH = getFlag('--corpus') ?? 'both'; // --write-readme respects --corpus (CI uses adversarial = deterministic, no TS clone)
 
-let parserSummary: BenchSummary | null = null;
-let advSummary: BenchSummary | null = null;
 let issueStats: IssueStats | null = null;
 
 if (WHICH === 'parser' || WHICH === 'both') {
@@ -665,7 +664,7 @@ if (WHICH === 'parser' || WHICH === 'both') {
     for (const f of files) {
       try { inputs.push({ name: f, text: readFileSync(f, 'utf-8') }); } catch { /* skip unreadable */ }
     }
-    parserSummary = runBench('parser conformance corpus', 'tests/cases/conformance/parser (TS parser test suite)', inputs).summary;
+    runBench('parser conformance corpus', 'tests/cases/conformance/parser (TS parser test suite)', inputs);
   }
 }
 
@@ -683,10 +682,14 @@ if (WHICH === 'adversarial' || WHICH === 'both') {
     `${distinct} issues / ${advInputs.length} cases from test/issue-cases.ts`,
     advInputs,
   );
-  advSummary = res.summary;
   issueStats = reportByIssue(res.perInput);
 }
 
 if (WRITE_README) {
-  writeReadmeBlock(buildBenchMarkdown(parserSummary, advSummary, issueStats));
+  // One row per language. TypeScript has the full oracle bench today; add more
+  // languages here as each gets its grammar + documented-bug ledger.
+  writeReadmeBlock(buildBenchMarkdown([
+    { name: 'TypeScript', issue: issueStats },
+    { name: 'JavaScript', issue: null, note: 'pending — JS not yet on the neutral-oracle bench (see ROADMAP)' },
+  ]));
 }