polytypo 1.0.0

data/lib/polytypo/data/fixtures/de-CH.json

@@ -0,0 +1,501 @@
+{
+  "spec": "1.0.0",
+  "locale": "de-CH",
+  "cases": [
+    {
+      "id": "de-ch-spaces-collapse-run",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "Hallo   Welt.",
+      "out": "Hallo Welt.",
+      "note": "spaces.md §6 case 1."
+    },
+    {
+      "id": "de-ch-spaces-before-punctuation",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "See p. 12 .",
+      "out": "See p. 12.",
+      "note": "spaces.md §6 case 10."
+    },
+    {
+      "id": "de-ch-spaces-checkbox",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "- [ ] Milch kaufen",
+      "out": "- [ ] Milch kaufen",
+      "note": "spaces.md §6 case 4: the empty-bracket guard keeps the GFM task-list marker intact."
+    },
+    {
+      "id": "de-ch-spaces-nobreak-not-collapsed",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "a  b",
+      "out": "a  b",
+      "note": "spaces.md §6 case 8: two no-break spaces are never collapsed — this rule only ever removes U+0020."
+    },
+    {
+      "id": "de-ch-ellipsis-three-dots",
+      "rule": "ellipsis",
+      "mode": "text",
+      "in": "Warte... was?",
+      "out": "Warte… was?",
+      "note": "ellipsis.md §6 case 1."
+    },
+    {
+      "id": "de-ch-ellipsis-mixed-run",
+      "rule": "ellipsis",
+      "mode": "text",
+      "in": "Warte…… was",
+      "out": "Warte… was",
+      "note": "ellipsis.md §6 case 2."
+    },
+    {
+      "id": "de-ch-ellipsis-two-dots-after-terminal",
+      "rule": "ellipsis",
+      "mode": "text",
+      "in": "Wirklich?..",
+      "out": "Wirklich?…",
+      "note": "ellipsis.md §6 case 3: abbreviatedAfterTerminal = false."
+    },
+    {
+      "id": "de-ch-ellipsis-existing",
+      "rule": "ellipsis",
+      "mode": "text",
+      "in": "Er ging…",
+      "out": "Er ging…",
+      "note": "ellipsis.md §6 case 6."
+    },
+    {
+      "id": "de-ch-dashes-parenthetical-tight-input",
+      "rule": "dashes",
+      "mode": "text",
+      "in": "Der Plan--falls es einen gibt--scheitert.",
+      "out": "Der Plan – falls es einen gibt – scheitert.",
+      "note": "dash.parenthetical = en-spaced — Bundeskanzlei Rz. 231/237/238: der Gedankenstrich ist der Halbgeviertstrich, mit Leerzeichen davor und danach."
+    },
+    {
+      "id": "de-ch-dashes-authored-em-dash-converts",
+      "rule": "dashes",
+      "mode": "text",
+      "in": "Das Jahr — ein langes — endet.",
+      "out": "Das Jahr – ein langes – endet.",
+      "note": "dashes.md §6 case 41, spec 0.2.0. Spec 0.1.0's authored-dash guard (and a briefly-shipped 0.2.0 narrowing that preserved only length) is fully retired: de-CH's parenthetical is `en-spaced`, so the author's em dash is promoted to it exactly as a hyphen-typed one would be — length included. See dashes.md §7 item 14 for why length stopped being treated as reliably authorial. Bundeskanzlei Rz. 236 states the Geviertstrich (U+2014) is not permitted in Swiss usage; the author's em dash is now corrected to the prescribed en dash rather than merely respaced or left alone. This fixture holds only for U+2014 — see `de-ch-dashes-authored-en-dash-tight-declined` for why U+2013 does not follow the same path since spec 0.6.0."
+    },
+    {
+      "id": "de-ch-dashes-authored-en-dash-tight-declined",
+      "rule": "dashes",
+      "mode": "text",
+      "in": "Das Jahr–ein langes–endet.",
+      "out": "Das Jahr–ein langes–endet.",
+      "note": "dashes.md §3.4 P5 (spec 0.6.0): a pure single authored U+2013 is declined unconditionally, even though de-CH's `dash.parenthetical` target glyph is already `en-spaced` (U+2013) and Bundeskanzlei Rz. 236 forbids the Geviertstrich outright. Before P5 this tight authored en-dash would have been re-spaced; P5 protects the whole token, spacing included (§8.8)."
+    },
+    {
+      "id": "de-ch-ranges-years",
+      "rule": "ranges",
+      "mode": "text",
+      "in": "Die Jahre 1939-1945",
+      "out": "Die Jahre 1939⁠–⁠1945",
+      "note": "Moved from `dashes` to `ranges` (spec 0.5.0, opt-in — `rules: {ranges: true}` explicit here since `ranges` defaults off). dash.range = en-tight — Bundeskanzlei Rz. 234: der Bis-Strich steht ohne Leerzeichen, „die Jahre 1939–1945“. The hyphen-typed range converts and is bound with U+2060 word joiners (ranges.md §3.3.1), which are zero-width and visible only in spec/fixtures/.escaped/.",
+      "rules": {
+        "ranges": true
+      }
+    },
+    {
+      "id": "de-ch-ranges-hours",
+      "rule": "ranges",
+      "mode": "text",
+      "in": "16-17 Uhr",
+      "out": "16⁠–⁠17 Uhr",
+      "note": "Moved from `dashes` to `ranges` (spec 0.5.0, opt-in — `rules: {ranges: true}` explicit here since `ranges` defaults off). Rz. 234, second cited example, with the same U+2060 binding (ranges.md §3.3.1).",
+      "rules": {
+        "ranges": true
+      }
+    },
+    {
+      "id": "de-ch-ranges-roundtrip",
+      "rule": "ranges",
+      "mode": "text",
+      "in": "die Artikel 10–12",
+      "out": "die Artikel 10–12",
+      "note": "spec 0.5.0 ownership correction: moved from `dashes` to `ranges` (`rules: {ranges: true}` explicit — `ranges` defaults off) — this case tests G1-G5/cluster/binding behaviour that `dashes` no longer evaluates at all for a digit-flanked token; it was vacuous under `rule: \"dashes\"` after the split. An AUTHORED range: byte-identical out, and specifically with no U+2060 joiners, where the hyphen-typed `1939-1945` above gets them (§3.3.1, §7.14). Both forms look the same on screen, so nothing in the rendered document explains why one may break across lines and the other may not — the asymmetry is recorded rather than hidden. It follows from one principle: the rule never makes an edit whose entire content is invisible, because an invisible-only change to already-correct text cannot be reviewed by hand. ranges.md §3-§4 is the current normative source for this rule's own behaviour here.",
+      "rules": {
+        "ranges": true
+      }
+    },
+    {
+      "id": "de-ch-ranges-iso-date",
+      "rule": "ranges",
+      "mode": "text",
+      "in": "Erschienen 2026-08-15",
+      "out": "Erschienen 2026-08-15",
+      "note": "spec 0.5.0 ownership correction: moved from `dashes` to `ranges` (`rules: {ranges: true}` explicit — `ranges` defaults off) — this case tests G1-G5/cluster/binding behaviour that `dashes` no longer evaluates at all for a digit-flanked token; it was vacuous under `rule: \"dashes\"` after the split. Guard G2 and the cluster guard. ranges.md §3.2's G2 (chain) is the current normative guard rejecting this.",
+      "rules": {
+        "ranges": true
+      }
+    },
+    {
+      "id": "de-ch-ranges-decreasing-range",
+      "rule": "ranges",
+      "mode": "text",
+      "in": "Resultat 20-10",
+      "out": "Resultat 20-10",
+      "note": "spec 0.5.0 ownership correction: moved from `dashes` to `ranges` (`rules: {ranges: true}` explicit — `ranges` defaults off) — this case tests G1-G5/cluster/binding behaviour that `dashes` no longer evaluates at all for a digit-flanked token; it was vacuous under `rule: \"dashes\"` after the split. dashes.md §6 case 10: guard G5, the left value must not exceed the right. ranges.md §3.2's G5 (non-decreasing) is the current normative guard rejecting this.",
+      "rules": {
+        "ranges": true
+      }
+    },
+    {
+      "id": "de-ch-dashes-compound-hyphen",
+      "rule": "dashes",
+      "mode": "text",
+      "in": "Ein wohl-bekannter E-Mail-Adressat",
+      "out": "Ein wohl-bekannter E-Mail-Adressat",
+      "note": "Guard P1."
+    },
+    {
+      "id": "de-ch-dashes-composition-guard-t2",
+      "rule": "dashes",
+      "mode": "text",
+      "in": ".--.",
+      "out": ".--.",
+      "note": "dashes.md §6 case 28: guard T2, since de-CH is also an en-spaced locale."
+    },
+    {
+      "id": "de-ch-hyphen-noop-compounds",
+      "rule": "hyphen",
+      "mode": "text",
+      "in": "Ein E-Mail-Konto und der Nord-Süd-Verkehr",
+      "out": "Ein E-Mail-Konto und der Nord-Süd-Verkehr",
+      "note": "hyphen.md §6 case 13: all three hyphen lists are empty (Bundeskanzlei Rz. 223/224 — the only documented protected hyphen is the open-ended Ergänzungsstrich, which no literal list can express). A provable total no-op; a port that implements the rule anyway must fail here."
+    },
+    {
+      "id": "de-ch-hyphen-noop-supplementary",
+      "rule": "hyphen",
+      "mode": "text",
+      "in": "Papierproduktion und -handel",
+      "out": "Papierproduktion und -handel",
+      "note": "The Ergänzungsstrich of Rz. 224 itself: normatively unbreakable and deliberately still untouched."
+    },
+    {
+      "id": "de-ch-quotes-primary-no-inner-space",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Er sagte \"hallo\" zweimal.",
+      "out": "Er sagte «hallo» zweimal.",
+      "note": "The headline Swiss case, and it contradicts what most references and most tools assume: Swiss German uses guillemets « », and there is NO space inside them. Bundeskanzlei Rz. 201–203 prescribes « » (and forbids the German » « for official texts); Rz. 202 regulates only the space BEFORE the opening and AFTER the closing mark, and none of its examples («Zukunft für Schweizer Fahrende») sets an inner space. Hence quotes.primary.innerSpace = none and N8 does nothing — unlike fr, where the same glyphs take U+00A0 inside."
+    },
+    {
+      "id": "de-ch-quotes-primary-roundtrip",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Er sagte «hallo» zweimal.",
+      "out": "Er sagte «hallo» zweimal.",
+      "note": "The fixed point of the previous case: no space is inserted on a second pass either."
+    },
+    {
+      "id": "de-ch-quotes-inner-space-removed",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Er sagte « hallo » zweimal.",
+      "out": "Er sagte «hallo» zweimal.",
+      "note": "quotes.md §3.7, updated for spec 0.3.0 mandate 2: a space touching a quote mark is sloppiness, not evidence. Under 0.1.0 an already-correctly-guillemetted pair with a stray inner space was left untouched by `quotes` (no straight marks to act on). Under mandate 2 the pair is recognised on the first application and, since de-CH's innerSpace is none, the inner space is deleted by `quotes` itself rather than surviving as a typo. Renamed from de-ch-quotes-inner-space-not-removed, whose old id described 0.1.0 behaviour that mandate 2 reverses."
+    },
+    {
+      "id": "de-ch-quotes-nested",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "\"Er sagte 'nein' zu mir\", notierte sie.",
+      "out": "«Er sagte ‹nein› zu mir», notierte sie.",
+      "note": "Depth 2 takes the halbe Anführungszeichen ‹ › (Rz. 201–203), not the German ‚ ‘."
+    },
+    {
+      "id": "de-ch-quotes-unbalanced",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "Er sagte \"hallo",
+      "out": "Er sagte \"hallo",
+      "note": "quotes.md §3.6."
+    },
+    {
+      "id": "de-ch-apostrophe-medial",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "Ich hab's gesehen",
+      "out": "Ich hab’s gesehen",
+      "note": "apostrophe.md §6 case 1."
+    },
+    {
+      "id": "de-ch-apostrophe-existing",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "Ich hab’s gesehen",
+      "out": "Ich hab’s gesehen",
+      "note": "U+2019 is never examined."
+    },
+    {
+      "id": "de-ch-apostrophe-isolated-mark",
+      "rule": "apostrophe",
+      "mode": "text",
+      "in": "a ' b",
+      "out": "a ' b",
+      "note": "apostrophe.md §6 case 9 and §7.4: nothing can be inferred from an isolated mark, so it stays straight — output text may legitimately still contain U+0027."
+    },
+    {
+      "id": "de-ch-symbols-copyright",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Copyright (c) 2026 Iurii Rogulia",
+      "out": "Copyright © 2026 Iurii Rogulia",
+      "note": "symbols.md §6 case 1."
+    },
+    {
+      "id": "de-ch-symbols-adjacent-signs",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "(c)(r)",
+      "out": "©(r)",
+      "note": "symbols.md §5: guard S1 lists ©, ® and ™ precisely so that `(c)(r)` converges in one run instead of two."
+    },
+    {
+      "id": "de-ch-symbols-multiplication-spaced",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "A 3x5 card, 10 x 20 grid",
+      "out": "A 3×5 card, 10 × 20 grid",
+      "note": "symbols.md §6 case 5 (verbatim): the tight and the symmetric-spaced form in one string, with nothing after either chain that another rule would claim."
+    },
+    {
+      "id": "de-ch-symbols-multiplication-followed-by-unit",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Format 10 x 20 cm",
+      "out": "Format 10 × 20 cm",
+      "note": "The cross-rule case: the U+00D7 and its two ordinary spaces are `symbols` (§3.3 step 8 carries each side's space code point across unchanged), while the U+00A0 before `cm` is `nbsp` N5, because de-CH lists `cm` in beforeUnits. Two rules, two different space decisions in one string."
+    },
+    {
+      "id": "de-ch-symbols-chain-converted",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Eine 2x3x4 Schachtel",
+      "out": "Eine 2×3×4 Schachtel",
+      "note": "symbols.md §6 case 5e: the whole chain converts in ONE pass. The branch is a scan of `DIGIT+ (X DIGIT+)+` (§3.3), not a pairwise test — a pairwise implementation converts one operator per pass and never converges, which is the non-idempotency §7.10 measures in Lebedev's Typograf. The former chain guard M5 was removed rather than extended, because M2/M3 are what made a pairwise chain reject itself."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-tight",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Tolerance +/-5 points",
+      "out": "Tolerance ±5 points",
+      "note": "symbols.md §6 case 5k (verbatim): the literal three code points `+/-` become U+00B1 when a digit follows."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-spaced",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Tolerance +/- 5 points",
+      "out": "Tolerance ± 5 points",
+      "note": "symbols.md §6 case 5l (verbatim): one intervening U+0020 is allowed by guard F2, and the space itself is left exactly as it was — it stays U+0020 here, because nothing else claims it."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-followed-by-unit",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Tolerance +/-5 mm",
+      "out": "Tolerance ±5 mm",
+      "note": "The cross-rule case for the tight form: U+00B1 from `symbols`, U+00A0 from `nbsp` N5 (beforeUnits `mm`). `symbols` never emits a no-break space."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-spaced-followed-by-unit",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Tolerance +/- 5 mm",
+      "out": "Tolerance ± 5 mm",
+      "note": "And for the spaced form. Note the two spaces get different verdicts in the same token: the one between ± and 5 is left as U+0020 by guard F2, while the one between 5 and mm becomes U+00A0 under nbsp N5."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-prose",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "lines marked +/- were edited",
+      "out": "lines marked +/- were edited",
+      "note": "symbols.md §6 case 5n, and the reason guard F2 requires a digit at all: prose that NAMES the characters rather than using them must survive. A changelog line or a legend is exactly the content this package runs on, and converting it would silently corrupt a sentence about the ASCII symbols themselves."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-character-class",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "match [+/-] once",
+      "out": "match [+/-] once",
+      "note": "symbols.md §6 case 5o: guard F1, a regular-expression character class — the one code-like context that survives into prose about code."
+    },
+    {
+      "id": "de-ch-symbols-plusminus-bare-not-converted",
+      "rule": "symbols",
+      "mode": "text",
+      "in": "Range 5+-3",
+      "out": "Range 5+-3",
+      "note": "symbols.md §6 case 5p, §3.4 and §7.11: bare `+-` is NEVER converted, in any context, and the asymmetry with `+/-` is the design rather than an oversight — `+-` occurs in diff listings, ASCII table borders, regex classes and adjacent source operators, none of which is skipped in text mode. Do not add it later for symmetry."
+    },
+    {
+      "id": "de-ch-nbsp-abbreviation-zb",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "z. B. so",
+      "out": "z. B. so",
+      "note": "N4 — Bundeskanzlei Rz. 438, mehrgliedrige Abkürzungen."
+    },
+    {
+      "id": "de-ch-nbsp-abbreviation-zb-place",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "z. B. Bern",
+      "out": "z. B. Bern",
+      "note": "nbsp.md §6 case 14 in de-CH: guard C1-a keeps the initials sub-rule from gluing `B.` to the following capitalised word. The same false positive existed here as in de-DE."
+    },
+    {
+      "id": "de-ch-nbsp-abbreviation-four-part",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "m. a. W. so",
+      "out": "m. a. W. so",
+      "note": "N4 binds every internal space of the listed form — Rz. 438 lists m. a. W. and m. w. H. explicitly."
+    },
+    {
+      "id": "de-ch-nbsp-before-word-sankt",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "St. Gallen und St. Moritz",
+      "out": "St. Gallen und St. Moritz",
+      "note": "N10 — Rz. 438 names exactly this case: die eingliedrige Abkürzung „St.“ für „Sankt“ in Ortsnamen. Three code points, so guard G-D does not apply."
+    },
+    {
+      "id": "de-ch-nbsp-units",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Es sind 20 km und 5 %",
+      "out": "Es sind 20 km und 5 %",
+      "note": "N5 twice — Rz. 251, 549–550 (Zahl und Einheit) and Rz. 554 (vor % und ‰)."
+    },
+    {
+      "id": "de-ch-nbsp-unit-never-inserts",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Es sind 20km",
+      "out": "Es sind 20km",
+      "note": "nbsp.md §7.2: N5 converts, never inserts."
+    },
+    {
+      "id": "de-ch-nbsp-before-number",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "siehe S. 17 und Ziff. 3",
+      "out": "siehe S. 17 und Ziff. 3",
+      "note": "N9 — Rz. 251 („S. 17 f.“) and Rz. 727 (abgekürzte Gliederungseinheiten Kap., Art., Abs., Ziff.)."
+    },
+    {
+      "id": "de-ch-nbsp-before-number-abs-kap",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Abs. 2 und Kap. 4",
+      "out": "Abs. 2 und Kap. 4",
+      "note": "N9 on the remaining two Gliederungseinheiten."
+    },
+    {
+      "id": "de-ch-nbsp-symbol-number",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Art. 35 und § 12 und Nr. 5",
+      "out": "Art. 35 und § 12 und Nr. 5",
+      "note": "Three abbreviation-and-number forms that look alike and are reached three different ways, which is the point of the case. `Art.` binds through N9 (beforeNumber): the Schreibweisungen define Begriffszeichen in Rz. 439 as a closed list — Ziffern, %, ‰, Gedankenstrich, Schrägstrich, §, Währungszeichen, mathematische Zeichen, Einheitenzeichen — and `Art.` is not in it, while Rz. 727 puts it on the abbreviation side beside Abs., Ziff. and Kap. `§` binds through N6 (afterSymbols), where Rz. 439 does list it. `Nr.` binds through neither: it was removed from this file for want of any source that states a spacing rule for it. The difference between the three is evidence, not how German looks."
+    },
+    {
+      "id": "de-ch-nbsp-nr-unbound-pending-citation",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "siehe Nr. 5",
+      "out": "siehe Nr. 5",
+      "note": "Pinned separately so a future contributor knows exactly which fixture to flip: if DIN 5008 (paywalled, unread) turns out to state the binding, `Nr.` goes into nbsp.beforeNumber — not afterSymbols, since Rz. 439's closed list excludes it and Duden classifies it as an abbreviation — and this `out` becomes `siehe Nr.` + U+00A0 + `5`."
+    },
+    {
+      "id": "de-ch-nbsp-before-number-left-boundary",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Fig.Art. 5 und Muster-Art. 5",
+      "out": "Fig.Art. 5 und Muster-Art. 5",
+      "note": "Why the afterSymbols → beforeNumber move for `Art.` is not cosmetic. N6 asks only that the code point before the entry is not ALNUM, so both of these bound while `Art.` was an afterSymbols entry. N9's left guard is much stronger — cp[a-1] must be NONE, SPACELIKE, OPENISH or SENTENCE-DASH (nbsp.md §3.12 step 2) — so a preceding full stop or hyphen now declines the match, and neither of these binds."
+    },
+    {
+      "id": "de-ch-nbsp-initials",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "J. K. Rowling",
+      "out": "J. K. Rowling",
+      "note": "N7 clause C1 twice, confirmed chain of two initials before the surname. initialBinding: \"chain\" (spec 0.6.0) is a derivation in this locale file, not a literal Weisung — Rz. 251 names „eine Abkürzung und ein Ortsname“ and Rz. 438 the parts of multi-token abbreviations."
+    },
+    {
+      "id": "de-ch-nbsp-no-french-spacing",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Wirklich ? Ja !",
+      "out": "Wirklich? Ja!",
+      "note": "Both punctuation lists are empty: Swiss German is not French, even though it shares the guillemets."
+    },
+    {
+      "id": "de-ch-nbsp-plain-sentence",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Er ging einfach nach Hause",
+      "out": "Er ging einfach nach Hause",
+      "note": "No short word list, no abbreviation, no unit, no initial."
+    },
+    {
+      "id": "de-ch-roundtrip-typeset-sentence",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "St. Gallen liegt 20 km entfernt – siehe S. 17.",
+      "out": "St. Gallen liegt 20 km entfernt – siehe S. 17.",
+      "note": "Byte-identical round trip: a bound place-name abbreviation, a bound unit, an en-spaced dash and a bound page reference."
+    },
+    {
+      "id": "de-ch-roundtrip-quotes-and-range",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "«Die Jahre 1939–1945», sagte er.",
+      "out": "«Die Jahre 1939–1945», sagte er.",
+      "note": "Correct Swiss guillemets without inner space, plus a tight en-dash range already exactly in de-CH's `en-tight` form. §3.3.1's invisible-edit test computes the unbound replacement first; it is identical to the input, so no joiners are added and nothing is emitted at all (§7.14) — this holds regardless of whether the dash was hyphen- or en-typed (spec 0.2.0 retired the authored-dash guard). Correctly typeset Swiss copy round-trips byte-for-byte, including the invisible layer."
+    },
+    {
+      "id": "de-ch-roundtrip-plain-prose",
+      "rule": "spaces",
+      "mode": "text",
+      "in": "Ein ganz gewöhnlicher Satz bleibt unverändert.",
+      "out": "Ein ganz gewöhnlicher Satz bleibt unverändert.",
+      "note": "No candidate for any rule."
+    },
+    {
+      "id": "de-ch-combined-sentence",
+      "rule": "nbsp",
+      "mode": "text",
+      "in": "Er sagte: \"Die Strecke -- rund 20 km -- war lang...\" und ging nach St. Gallen.",
+      "out": "Er sagte: «Die Strecke – rund 20 km – war lang…» und ging nach St. Gallen.",
+      "note": "A quotation, a parenthetical dash, an ellipsis and a number with a unit in one sentence: guillemets without inner space, en-spaced dashes (never an em dash), U+2026, U+00A0 between 20 and km, and St. bound to Gallen."
+    },
+    {
+      "id": "de-ch-quotes-041-cobug-apostrophe-v1",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "\"B'’\"‹",
+      "out": "«B’’»‹",
+      "note": "Implementation-discovered composition-obligation violation against I5 by apostrophe (R6), closed by quotes.md §3.2's V1ID normalisation (spec 0.4.1). Before the fix: on pass 1, quotes' raw pairing finds a crossing WIDE pair (the two U+0022) and a NARROW pair (U+2019/U+2039), both assigned depth 1 (primary), and the certification gate rejects the whole set down to the empty set because rendering both to the same primary glyph creates a new V1-adjacency the re-derivation cannot reproduce -- so quotes makes no edit, and apostrophe then converts the sole unmatched U+0027 to U+2019. On pass 2 the input has two literally-identical adjacent U+2019 where pass 1 had one U+0027 and one U+2019 (different code points); V1's old literal comparison newly fires and drops both from candidacy, removing the crossing NARROW pair and letting the WIDE pair certify cleanly -- a different, non-idempotent result. V1ID makes U+0027 and U+2019 the same identity for V1's own comparison on both passes, so pass 1 already reaches the fixed point this fixture asserts. Found by the fast-check idempotency property sweep; reproduces deterministically (not a flaky/seed-dependent finding). See quotes.md §5 Lemma A / Corollary A1 and apostrophe.md §5's Against I5 section."
+    },
+    {
+      "id": "de-ch-quotes-rock-n-roll",
+      "rule": "quotes",
+      "mode": "text",
+      "in": "rock 'n' roll",
+      "out": "rock 'n' roll",
+      "note": "spec 0.5.0: de-CH has no cited quotes.elisionIdioms entry for this idiom (quotes.md §7 item 8 — only en-US does, via Chicago Manual of Style Online and the American Heritage Dictionary). The general ambiguous-medial-span veto (quotes.md 3.2) preserves both marks as literal U+0027 rather than pairing them as an ordinary quotation, closing the false positive AUDIT_REMEDIATION_AND_RELEASE_PLAN.md 3.1 named for \"equivalent cases in several locales.\""
+    }
+  ]
+}