polytypo 1.1.0 → 1.3.0

data/lib/polytypo/data/fixtures/es.json

@@ -0,0 +1,193 @@
+{
+  "spec": "1.3.0",
+  "locale": "es",
+  "cases": [
+    {
+      "id": "es-quotes-primary",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Ha dicho \"buenos días\" y salió.",
+      "out": "Ha dicho «buenos días» y salió.",
+      "note": "quotes.primary = U+00AB/U+00BB con innerSpace \"none\" (DPD «comillas» 1: «se recomienda utilizar en primera instancia las comillas angulares»; «pegadas a la primera y la última palabra»)."
+    },
+    {
+      "id": "es-quotes-nested",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Escribió: \"Me dijo 'llego enseguida' y desapareció.\"",
+      "out": "Escribió: «Me dijo “llego enseguida” y desapareció.»",
+      "note": "Nivel 1 angulares, nivel 2 inglesas, según el orden que fija el DPD. El tercer nivel del DPD (simples) no es expresable en el esquema."
+    },
+    {
+      "id": "es-quotes-inner-space-deleted",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Ha dicho « buenos días » y salió.",
+      "out": "Ha dicho «buenos días» y salió.",
+      "note": "innerSpace \"none\": N8 elimina el espacio interior que un texto copiado de una fuente francesa trae consigo."
+    },
+    {
+      "id": "es-ellipsis-three-dots",
+      "rule": "ellipsis",
+      "mode": "text",
+      "in": "Espera... ¿qué?",
+      "out": "Espera… ¿qué?",
+      "note": "Tres U+002E pasan a U+2026 (DPD «puntos suspensivos» 1: «tres puntos consecutivos ―y solo tres―»)."
+    },
+    {
+      "id": "es-ellipsis-not-abbreviated-after-terminal",
+      "rule": "ellipsis",
+      "mode": "text",
+      "in": "¿De verdad?..",
+      "out": "¿De verdad?…",
+      "note": "abbreviatedAfterTerminal = false: la forma de dos puntos tras «?» que usa el ruso no existe en español, así que los dos puntos se completan a U+2026 en vez de conservarse."
+    },
+    {
+      "id": "es-dashes-parenthetical-none",
+      "rule": "dashes",
+      "mode": "text",
+      "in": "El plan - si existe - fracasa.",
+      "out": "El plan - si existe - fracasa.",
+      "note": "dash.parenthetical = \"none\": el DPD prescribe raya con espacios FUERA del par y ninguno dentro, asimetría que el enum no expresa; mientras tanto la regla no toca nada. Véase la voz dashes de es.json."
+    },
+    {
+      "id": "es-ranges-guion-none",
+      "rule": "ranges",
+      "mode": "text",
+      "in": "las páginas 23-45",
+      "out": "las páginas 23-45",
+      "note": "ranges.md §3: el token no se convierte porque dash.range es \"none\" — el signo español del intervalo es el guion U+002D, que ya está escrito (DPD «guion» 3 b), y el enum no tiene un valor para él. El caso se genera con ranges activado explícitamente, ya que la regla está desactivada por defecto.",
+      "rules": {
+        "ranges": true
+      }
+    },
+    {
+      "id": "es-hyphen-noop",
+      "rule": "hyphen",
+      "mode": "text",
+      "in": "un estudio teórico-práctico",
+      "out": "un estudio teórico-práctico",
+      "note": "Las tres listas de hyphen están vacías, así que la regla es un no-op total demostrable para el español (hyphen.md §2)."
+    },
+    {
+      "id": "es-nbsp-inverted-marks-untouched",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "¿Qué hora es? ¡Vaya!",
+      "out": "¿Qué hora es? ¡Vaya!",
+      "note": "beforePunctuation y narrowBeforePunctuation vacías: el español no pone espacio antes de «?» ni de «!», y U+00BF/U+00A1 no son miembros de CLOSEISH, así que la restricción Q-P impediría expresarlo aunque la fuente lo pidiera."
+    },
+    {
+      "id": "es-nbsp-degree-tight",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Hacía 27° a la sombra",
+      "out": "Hacía 27° a la sombra",
+      "note": "El DPD «símbolo» 5.4 pega «°» a la cifra cuando no se especifica la escala, por eso «°» a secas NO está en beforeUnits y aquí no se inserta nada."
+    },
+    {
+      "id": "es-nbsp-percent",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Aprobó un 8 % de los alumnos",
+      "out": "Aprobó un 8 % de los alumnos",
+      "note": "N5 beforeUnits: el espacio pasa a U+00A0. RAE, Duda lingüística: «El símbolo del porcentaje se separa con espacio de la cifra que lo precede: 50 %»; la indivisibilidad la impone el DPD «símbolo» 5.6."
+    },
+    {
+      "id": "es-nbsp-celsius",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "La temperatura es de 27 °C",
+      "out": "La temperatura es de 27 °C",
+      "note": "«°C» sí está en beforeUnits, citado literalmente por el DPD «símbolo» 5.4 («27° […] pero 27 °C»). Contrasta con es-nbsp-degree-tight."
+    },
+    {
+      "id": "es-nbsp-abbreviation-internal-space",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Viajó a los EE. UU. el lunes",
+      "out": "Viajó a los EE. UU. el lunes",
+      "note": "N4 abbreviations: el espacio interior de una abreviatura de varios elementos pasa a U+00A0 (DPD «abreviatura»: «Cuando la abreviatura se compone de varios elementos, estos no deben separarse en líneas diferentes»)."
+    },
+    {
+      "id": "es-nbsp-abbreviation-p-ej",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "algunos casos, p. ej. este",
+      "out": "algunos casos, p. ej. este",
+      "note": "La misma subrregla sobre «p. ej.», forma atestiguada en la Lista de abreviaturas de El buen uso del español."
+    },
+    {
+      "id": "es-nbsp-before-number",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "véase la pág. 23",
+      "out": "véase la pág. 23",
+      "note": "N9 beforeNumber: «⊗15 / págs.» es el ejemplo del propio DPD «abreviatura» para este caso."
+    },
+    {
+      "id": "es-nbsp-before-word",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "El Sr. Pérez llegó tarde",
+      "out": "El Sr. Pérez llegó tarde",
+      "note": "N10 beforeWord: «⊗Sr. / Pérez» es el segundo ejemplo de la misma entrada. La lista se limita a los tratamientos de cortesía, por el riesgo de falsos positivos de esta subrregla."
+    },
+    {
+      "id": "es-nbsp-initials-chain",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Lo escribió J. A. Pérez ayer",
+      "out": "Lo escribió J. A. Pérez ayer",
+      "note": "initialBinding = \"chain\": la cadena completa se liga, incluido el espacio que precede a la primera inicial, igual que en en-US y ru. Decisión del operador (18.09.2026) ante una fuente que no distingue entre una inicial y dos; véase la voz nbsp correspondiente."
+    },
+    {
+      "id": "es-nbsp-lone-initial-not-bound",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "La ley N. Roma decidió",
+      "out": "La ley N. Roma decidió",
+      "note": "El reverso del caso anterior y la razón de elegir \"chain\": una sola mayúscula con punto seguida de un nombre propio NO se liga, de modo que un final de frase no se confunde con una inicial."
+    },
+    {
+      "id": "es-nbsp-ac-guard",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "en el año 33 a. C. Roma era una república",
+      "out": "en el año 33 a. C. Roma era una república",
+      "note": "N4 liga el espacio interior de «a. C.» y la guarda C1-a (nbsp.md §3.9) impide que C1 lea «C.» como inicial y «Roma» como apellido. Equivalente español del alemán «z. B. Berlin»."
+    },
+    {
+      "id": "es-apostrophe-foreign-name",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "L'Hospitalet de Llobregat",
+      "out": "L’Hospitalet de Llobregat",
+      "note": "U+0027 pasa a U+2019. El DPD «apóstrofo» limita el signo en español a textos antiguos, habla reproducida y nombres de otras lenguas: este es el tercer caso, y es el que un texto español encuentra de verdad."
+    },
+    {
+      "id": "es-roundtrip-noop",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Ha dicho «buenos días» y salió…",
+      "out": "Ha dicho «buenos días» y salió…",
+      "note": "modes.md §4: un texto ya compuesto vuelve byte a byte."
+    },
+    {
+      "id": "es-spaces-brackets-and-final-stop",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "Dijo ( sin dudar ) que sí  .",
+      "out": "Dijo (sin dudar) que sí.",
+      "note": "spaces.md §3: los espacios interiores de los paréntesis se eliminan y la serie anterior al punto final se colapsa. Regla independiente de la locale, presente aquí para que es tenga un caso por cada regla canónica."
+    },
+    {
+      "id": "es-symbols-copyright-and-dimensions",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Copyright (c) 2026, tamaño 40x60 cm",
+      "out": "Copyright © 2026, tamaño 40×60 cm",
+      "note": "symbols.md: «(c)» pasa a U+00A9 y la «x» entre cifras a U+00D7. El U+00A0 ante «cm» lo pone nbsp N5, no symbols; se deja en el mismo caso porque así se ve que las dos reglas componen."
+    }
+  ]
+}

data/lib/polytypo/data/fixtures/fi.json

@@ -1,5 +1,5 @@
 {
-  "spec": "1.1.0",
+  "spec": "1.3.0",
   "locale": "fi",
   "cases": [
     {
@@ -1309,6 +1309,14 @@
       "in": "'90s were fun,' he said",
       "out": "”90s were fun,” he said",
       "note": "quotes.md §6 row E3s: the straight-input control for E3."
+    },
+    {
+      "id": "fi-apostrophe-elision-before-shared-quote-glyph",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "lehti l'\"Humanité\"",
+      "out": "lehti l’”Humanité”",
+      "note": "apostrophe.md §3.3 cases 3 and 3a (spec 1.2.0): fi opens and closes with U+201D, which is in CLOSEISH, so case 3 already converted this before 1.2.0. Pinned so that case 3a is not read as changing a shared-glyph locale."
     }
   ]
 }

data/lib/polytypo/data/fixtures/fr-CA.json

@@ -1,5 +1,5 @@
 {
-  "spec": "1.1.0",
+  "spec": "1.3.0",
   "locale": "fr-CA",
   "cases": [
     {
@@ -263,6 +263,55 @@
         "ranges": true
       },
       "note": "spec 0.5.0: fr-CA's dash.range is \"none\" (no verified convention), so `ranges` substitutes nothing even when explicitly enabled (ranges.md §2, §3.3) — the same invariant the pre-0.5.0 `dashes`-tagged no-op fixtures for this locale already proved for the combined rule."
+    },
+    {
+      "id": "fr-ca-nbsp-span-boundary-strong-colon",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "<strong>Résistant au gel :</strong> il reste souple",
+      "out": "<strong>Résistant au gel :</strong> il reste souple",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0): same as the `fr` case; `fr-CA` lists `:` in its own punctuation data."
+    },
+    {
+      "id": "fr-ca-nbsp-span-boundary-markdown-strong-colon",
+      "rule": "nbsp",
+      "mode": "markdown",
+      "dialect": "commonmark",
+      "in": "**Note :** voir plus bas\n",
+      "out": "**Note :** voir plus bas\n",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0): the colon ends the strong span; the −1 marker after it is in CLOSEISH, so N1 restores the U+00A0 that `spaces` deleted. (fr-CA lists no punctuation for N2, so `?` would take nothing back in any mode.)"
+    },
+    {
+      "id": "fr-ca-apostrophe-elision-before-straight-quote",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "l'\"idée\" est simple",
+      "out": "l’« idée » est simple",
+      "note": "apostrophe.md §3.3 case 3a (spec 1.2.0): same as `fr`."
+    },
+    {
+      "id": "fr-ca-nbsp-numero-binds-number",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "voir n° 5",
+      "out": "voir n° 5",
+      "note": "nbsp.md §3.11 N9 : « n° » figure dans nbsp.beforeNumber depuis la spec 1.3.0 et se lie au nombre qui suit. La forme retenue est « n » + U+00B0, l'usage le plus répandu ; voir l'entrée sources de la locale, la source (OQLF) prescrivant un tracé et non un code point."
+    },
+    {
+      "id": "fr-ca-nbsp-numero-capital-is-its-own-entry",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "N° 5 de la revue",
+      "out": "N° 5 de la revue",
+      "note": "nbsp.md §3.11 N9 étape 1 : aucune tolérance de casse, « N° » est donc une entrée distincte de « n° » dans nbsp.beforeNumber."
+    },
+    {
+      "id": "fr-ca-nbsp-numero-needs-a-digit",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "le n° cinq",
+      "out": "le n° cinq",
+      "note": "nbsp.md §3.11 N9 étape 4 : il faut un DIGIT après le séparateur. Un nombre écrit en toutes lettres n'en est pas un, l'espace reste sécable."
     }
   ]
 }

data/lib/polytypo/data/fixtures/fr.json

@@ -1,5 +1,5 @@
 {
-  "spec": "1.1.0",
+  "spec": "1.3.0",
   "locale": "fr",
   "cases": [
     {
@@ -598,6 +598,287 @@
         "ranges": true
       },
       "note": "spec 0.5.0: fr's dash.range is \"none\" (no verified convention), so `ranges` substitutes nothing even when explicitly enabled (ranges.md §2, §3.3) — the same invariant the pre-0.5.0 `dashes`-tagged no-op fixtures for this locale already proved for the combined rule."
+    },
+    {
+      "id": "fr-spaces-dot-word-start-dotfile",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "Voir le fichier .htaccess .",
+      "out": "Voir le fichier .htaccess.",
+      "note": "spaces.md §3.4 word-start clause (spec 1.2.0): the space before `.htaccess` survives, the space before the terminal dot is deleted. `.` is in no `nbsp` list for `fr`, so nothing is put back."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-strong-colon",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "<strong>Résistant au gel :</strong> il reste souple",
+      "out": "<strong>Résistant au gel :</strong> il reste souple",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0), §6 row 6c; modes.md §3.3: `spaces` deletes the typed space before `:`, and the −1 span boundary marker after the mark is in `nbsp`'s CLOSEISH, so N1 inserts U+00A0 at the mark's index, inside the span. Spec 1.1.0 runtimes produced `gel:</strong>`."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-br-question",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "Et la réglementation ?<br>Oui",
+      "out": "Et la réglementation ?<br>Oui",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0), §6 row 6d: `<br>` leaves no line terminator in the gap, so the marker is −1; it is in CLOSEISH and N2 restores U+202F."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-link-question",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "<a href=\"/faq\">Comment ?</a> Oui",
+      "out": "<a href=\"/faq\">Comment ?</a> Oui",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0): the mark ends the link text, the marker follows it, and N2 restores U+202F instead of leaving `Comment?`."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-markdown-strong-colon",
+      "rule": "nbsp",
+      "mode": "markdown",
+      "dialect": "commonmark",
+      "in": "**Résistant au gel :** il reste souple\n",
+      "out": "**Résistant au gel :** il reste souple\n",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0): the Markdown form of the French definition pattern; the closing `**` is a skipped region with no line terminator, so the marker is −1 and N1 restores U+00A0."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-not-openish-em",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "Il dit <em>non</em> ! Oui",
+      "out": "Il dit <em>non</em> ! Oui",
+      "note": "nbsp.md §3.3 step 3 (spec 1.2.0), §6 row 6e; modes.md §3.3: the marker is not in `nbsp`'s OPENISH, so the quote-glyph guard does not decline and N2 converts the space. A port that applied the quote-glyph guard to the marker would leave a plain U+0020 here."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-not-openish-link",
+      "rule": "nbsp",
+      "mode": "markdown",
+      "dialect": "commonmark",
+      "in": "Voir [ceci](http://example.org) : oui\n",
+      "out": "Voir [ceci](http://example.org) : oui\n",
+      "note": "nbsp.md §3.3 step 3 (spec 1.2.0): the second witness that the quote-glyph guard does not decline at a span boundary. The link text ends two places left of `:`, and N1 still converts the space. The marker's absence from the rest of `nbsp`'s OPENISH is pinned by ru-nbsp-span-boundary-not-openish-short-word."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-mark-alone-in-span",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "mot<em>!</em>",
+      "out": "mot<em>!</em>",
+      "note": "nbsp.md §3.3 step 2; modes.md §3.4 edge-growth rule: the mark is alone in its span, so the insertion point is the span edge and the insertion is discarded, as it was before spec 1.2.0."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-colon-before-digit-span",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "Il est 12:<b>30</b>",
+      "out": "Il est 12 :<b>30</b>",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0), §6 row 6f: the stated cost of accepting the marker. Step 2 cannot see past the span boundary, so a colon that ends one span while a digit starts the next is treated as sentence punctuation."
+    },
+    {
+      "id": "fr-apostrophe-elision-before-guillemet",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "un mélange d'« urine diluée »",
+      "out": "un mélange d’« urine diluée »",
+      "note": "apostrophe.md §3.3 case 3a (spec 1.2.0), §6 row 15: letter left, opening quotation glyph right. Spec 1.1.0 left the apostrophe straight, because no case matched a letter followed by an OPENISH-only glyph."
+    },
+    {
+      "id": "fr-apostrophe-elision-before-straight-quote",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "l'\"idée\" est simple",
+      "out": "l’« idée » est simple",
+      "note": "apostrophe.md §3.2, §3.3 case 3a (spec 1.2.0): `quotes` (order 40) turns the straight `\"` into `«` first, so the mark reaches this rule next to an opening glyph and case 3a converts it."
+    },
+    {
+      "id": "fr-apostrophe-elision-qu-before-guillemet",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "Il faut qu'« il vienne »",
+      "out": "Il faut qu’« il vienne »",
+      "note": "apostrophe.md §3.3 case 3a (spec 1.2.0): `qu'` before a quotation."
+    },
+    {
+      "id": "fr-nbsp-span-boundary-colon-before-url-span",
+      "rule": "nbsp",
+      "mode": "html",
+      "in": "Voir http:<a href=\"//x.org\">//x.org</a>",
+      "out": "Voir http :<a href=\"//x.org\">//x.org</a>",
+      "note": "nbsp.md §3.3 step 2 (spec 1.2.0): the second stated cost of accepting the marker. Step 2 cannot see the `/` in the next span, so the colon is treated as sentence punctuation and N1 inserts U+00A0."
+    },
+    {
+      "id": "fr-nbsp-numero-binds-number",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "voir n° 5",
+      "out": "voir n° 5",
+      "note": "nbsp.md §3.11 N9 : « n° » figure dans nbsp.beforeNumber depuis la spec 1.3.0 et se lie au nombre qui suit. La forme retenue est « n » + U+00B0, l'usage le plus répandu ; voir l'entrée sources de la locale, la source (OQLF) prescrivant un tracé et non un code point."
+    },
+    {
+      "id": "fr-nbsp-numero-capital-is-its-own-entry",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "N° 5 de la revue",
+      "out": "N° 5 de la revue",
+      "note": "nbsp.md §3.11 N9 étape 1 : aucune tolérance de casse, « N° » est donc une entrée distincte de « n° » dans nbsp.beforeNumber."
+    },
+    {
+      "id": "fr-nbsp-numero-needs-a-digit",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "le n° cinq",
+      "out": "le n° cinq",
+      "note": "nbsp.md §3.11 N9 étape 4 : il faut un DIGIT après le séparateur. Un nombre écrit en toutes lettres n'en est pas un, l'espace reste sécable."
+    },
+    {
+      "id": "fr-markdown-commonmark-frontmatter-nbsp",
+      "rule": "nbsp",
+      "mode": "markdown",
+      "dialect": "commonmark",
+      "in": "---\ntitle: Une note ; suite\n---\n\nEst-ce vrai ?\n",
+      "out": "---\ntitle: Une note ; suite\n---\n\nEst-ce vrai ?\n",
+      "note": "modes.md §3.7.3 gives this exact reason for skipping frontmatter: without it `fr` would put U+00A0 before the colon of a machine-read field and U+202F before its semicolon. The body sentence shows the rule is otherwise active in the same document, so the untouched field proves the skip rather than an inactive rule."
+    },
+    {
+      "id": "fr-markdown-mdx-frontmatter-nbsp",
+      "rule": "nbsp",
+      "mode": "markdown",
+      "dialect": "mdx",
+      "in": "---\ntitle: Une note ; suite\n---\n\n<Callout>Est-ce vrai ?</Callout>\n",
+      "out": "---\ntitle: Une note ; suite\n---\n\n<Callout>Est-ce vrai ?</Callout>\n",
+      "note": "modes.md §3.7.3: the same guarantee under \"mdx\", where frontmatter plus JSX is the ordinary document shape. This is the case the polytypo-js source comment above its frontmatter skip entry described as a spec gap."
+    },
+    {
+      "id": "fr-nbsp-character-reference-numeric",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Bonjour&#160;: oui",
+      "out": "Bonjour&#160;: oui",
+      "note": "nbsp.md §3.3 step 4 (spec 1.3.0): the `;` closes `&#160;`, so N2 emits nothing and the reference survives. The `:` takes nothing either — step 1's run guard sees a mark to its left. Before 1.3.0 this produced `Bonjour&#160` + U+202F + `: oui`, which is no longer a character reference."
+    },
+    {
+      "id": "fr-nbsp-character-reference-named",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Tom &amp; Jerry",
+      "out": "Tom &amp; Jerry",
+      "note": "nbsp.md §3.3 step 4 (spec 1.3.0): the same guard on a named reference. Before 1.3.0 `text` mode destroyed every character reference in French input, and only in locales whose nbsp data puts a space before `;`."
+    },
+    {
+      "id": "fr-nbsp-character-reference-shape-not-table",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "a &notaname; b",
+      "out": "a &notaname; b",
+      "note": "nbsp.md §3.3 step 4 (spec 1.3.0): the guard tests the shape of a reference, not membership of the HTML named-reference table, which five runtimes would otherwise have to carry identically. Declining here costs nothing."
+    },
+    {
+      "id": "fr-nbsp-ordinary-semicolon-still-binds",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Oui ; non",
+      "out": "Oui ; non",
+      "note": "nbsp.md §3.3 step 4 (spec 1.3.0) is not a blanket refusal of `;`: an ordinary semicolon still takes U+202F under N2. Pinned next to the reference cases so the guard cannot be widened into one."
+    },
+    {
+      "id": "fr-nbsp-semicolon-after-digit-binds",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Section 4; suite",
+      "out": "Section 4 ; suite",
+      "note": "nbsp.md §3.3 step 4 (spec 1.3.0): the left walk finds the digit run but no `&` before it, so the guard does not fire."
+    },
+    {
+      "id": "fr-nbsp-narrow-substituted",
+      "rule": "nbsp",
+      "mode": "text",
+      "narrowNbsp": "nbsp",
+      "in": "Un délai ? Vraiment ! Et puis ; voilà.",
+      "out": "Un délai ? Vraiment ! Et puis ; voilà.",
+      "note": "nbsp.md §3.1a (spec 1.3.0): `narrowNbsp: \"nbsp\"` moves N2's target from U+202F to U+00A0. The same indices are claimed by the same sub-rule under the same guards — only the character written changes. Compare fr-nbsp-narrow-default, which is this input with the option absent."
+    },
+    {
+      "id": "fr-nbsp-narrow-default",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Un délai ? Vraiment ! Et puis ; voilà.",
+      "out": "Un délai ? Vraiment ! Et puis ; voilà.",
+      "note": "nbsp.md §3.1a: the control for fr-nbsp-narrow-substituted. No `narrowNbsp` key means \"narrow\", so every case written before spec 1.3.0 keeps its meaning and a runtime that ignores the option unknowingly still passes this one — which is why the substituted case above is the one that proves the feature."
+    },
+    {
+      "id": "fr-nbsp-narrow-substituted-authored",
+      "rule": "nbsp",
+      "mode": "text",
+      "narrowNbsp": "nbsp",
+      "in": "Oui ? Non !",
+      "out": "Oui ? Non !",
+      "note": "nbsp.md §3.1a: an authored U+202F at an index N2 claims is normalised to the substituted target, exactly as an authored U+00A0 is normalised to U+202F in the default configuration — the rule normalises a claimed index to its target, and the target has moved. This is the case that separates a target rewrite from post-processing: a caller's replaceAll would also produce this output, but would not be a fixed point when its own output is fed back in."
+    },
+    {
+      "id": "fr-nbsp-narrow-substituted-authored-default",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Oui ? Non !",
+      "out": "Oui ? Non !",
+      "note": "nbsp.md §3.1a: the control for the case above — with the default target an authored U+202F is already correct and nothing is emitted."
+    },
+    {
+      "id": "fr-nbsp-narrow-substituted-with-quotes",
+      "rule": "nbsp",
+      "mode": "text",
+      "narrowNbsp": "nbsp",
+      "in": "Il a dit : « oui » ; puis ?",
+      "out": "Il a dit : « oui » ; puis ?",
+      "note": "nbsp.md §3.1a with N1, N2 and N8 all claiming indices in one string. N1 (the colon) and N8 (`fr`'s primary pair, whose innerSpace is \"nbsp\") already wrote U+00A0 and are untouched by the option; N2 (`;` and `?`) moves. With the substitution on, all three want the same character — the option can only ever make N2's target EQUAL to N1's, never different, which is why §5's clause 0 needs no new case."
+    },
+    {
+      "id": "fr-nbsp-narrow-substituted-guards-unchanged",
+      "rule": "nbsp",
+      "mode": "text",
+      "narrowNbsp": "nbsp",
+      "in": "12:30 et http://x ; oui",
+      "out": "12:30 et http://x ; oui",
+      "note": "nbsp.md §3.1a: the option changes what the rule writes, never what it reads or which indices it claims. N1's right-context guard still protects the time and the URL — the colon in `12:30` and in `http://` takes nothing, under the substitution exactly as under the default."
+    },
+    {
+      "id": "fr-nbsp-quote-inner-space-authored-narrow",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Il a dit « mot ».",
+      "out": "Il a dit « mot ».",
+      "note": "nbsp.md §3.10 and §6 rows 3-4, corrected in spec 1.3.0. `fr.json` sets the primary pair's innerSpace to \"nbsp\", so N8's target is U+00A0 and an authored U+202F inside the quotation marks is CONVERTED, not already correct. Through spec 1.2.0 the worked-example table claimed both the narrow space and a fixed point here — a claim about a locale file the file never supported, and nothing pinned it. This case pins it."
+    },
+    {
+      "id": "fr-yaml-narrow-space-in-quoted",
+      "rule": "nbsp",
+      "mode": "yaml",
+      "keys": [
+        "description"
+      ],
+      "in": "description: \"Une note !\"\n",
+      "out": "description: \"Une note !\"\n",
+      "note": "modes.md §3.8.6: a quoted scalar's content is one span, and French spacing applies inside it exactly as in text mode."
+    },
+    {
+      "id": "fr-yaml-no-space-across-colon-split",
+      "rule": "nbsp",
+      "mode": "yaml",
+      "keys": [
+        "description"
+      ],
+      "in": "description: prose:more prose\n",
+      "out": "description: prose:more prose\n",
+      "note": "modes.md §3.8.6: the `:` is outside every span, so the insertion `nbsp` would make before it lands on a span edge and is discarded by §3.4. This is the accepted cost of the split, and it is what keeps the scalar parseable."
+    },
+    {
+      "id": "fr-yaml-guillemets-across-block-lines",
+      "rule": "quotes",
+      "mode": "yaml",
+      "keys": [
+        "description"
+      ],
+      "in": "description: |\n  Il a dit \"bonjour\n  monsieur\" et il est parti\n",
+      "out": "description: |\n  Il a dit « bonjour\n  monsieur » et il est parti\n",
+      "note": "modes.md §3.2 model C in a locale whose quotation output is guillemets plus U+00A0 rather than curly quotes — structurally different from the en-US case of the same shape."
     }
   ]
 }