colitur.git/test/test_colitur.ml, branch v0.1.0

test(oracle): a second oracle year, 2038, closing step 4's blind spot

2026-08-17T14:00:23Z

Register section 6.7 recorded that step 4 of the reading chain -- the Common route -- had no external witness of any kind, and proved it by mutation: corrupting a Common citation left both the 16801-day differential and the 730-day oracle green. This closes that. 2038 is the only year in 2005-2050 in which two of the five Common-routed saints are the observed office (Perpetua and Felicitas on 6 March, Frances of Rome on 9 March). 365 days were captured live, one request per day, zero failures. A separate fixture, not more rows on the existing one. sources.md already records that the live endpoint has drifted from lectio's archived 2026-2027 snapshot; merging two versions of one source behind a single set of expectations would make any future disagreement unattributable -- calendar or drift, no way to tell. Nothing in the 2038 comparison is checked against the older fixture, and its provenance file says so, along with the fact that a live capture is not reproducible byte-for-byte on demand the way a snapshot-derived one is. 338 of 365 days match. 27 differ, every one of them in a named, already-adjudicated class, none unexplained: 13 the BVM Saturday votive Mass (M26 shape 1), 7 colour (M21), 2 Judith (M28), 2 Christ the King's week (M26 shape 2b), 1 Christmas multi-Mass (M27), 1 Perpetua's Common, 1 new. Each was decided in a different year, so 2038 re-confirms them independently -- the point of a second window is not new rulings but evidence the old ones are not artefacts of their own year. Two findings came out of it. Perpetua and Felicitas: missalemeum serves the Common of Virgins and calls them "Virgins and Martyrs" in its own oration, where the Missal directs "Missa Me exspectaverunt, de Communi non Virginum I loco" (scan1:27634-27635). Both Commons share the Introit Ps 118:95-96, which is exactly why the calendarium qualifies its direction, and both women were mothers. Verdict colitur; the mechanism is located on missalemeum's side, not merely asserted. Passion Tuesday: the Missal prints "Dan. 14, 27 et 28-42" (scan1:11106) and colitur reproduces that two-part form where missalemeum collapses it to 27-42. Same verses; the convention is deliberate, appearing also in the Seven Sorrows and the Common of Non-Virgins. Verdict colitur, cosmetic. Citations are notation-normalised before comparing, as layer 3 already does: Ecclus/Sir, Joann/John, Luc/Luke, Matth/Matt. Each pair was added because a real row needed it, and no target contains its own source as a substring, so the set is idempotent. Not compared, stated rather than left to be found: commemorations and observed-identity. That machinery is built around a date-literal 28-entry allow-list specific to 2026-2027, and re-deriving it for a second year is its own task. 2038 compares rank, colour, Epistle, Gospel. The extractor's day count was hardcoded to 730, which silently forbade any other window. It is now a parameter defaulting to 730, so the existing documented command keeps its guard and a partial fetch still fails loudly instead of producing a short fixture that passes a comparison it never ran. Teeth, by re-running section 6.7's own experiment: corrupting the Common of Non-Virgins II now reddens four tests, two of them external-oracle, where the same mutation previously reddened neither oracle layer. Residual: isidore-of-seville is still unwitnessed and needs 2035 or 2046; gregory-the-great and patrick are never the observed office in any year 2005-2050, so no fixture in that range can reach them. Register section 6.8 and 6.9.

kernel+ef: resolve readings, chain steps 1 and 2

2026-08-14T22:16:35Z

Liturgical_day.citations has read "always empty until Plan 4" since Plan 3; it is now filled. Rite.t gains a readings function, rite-supplied for the same reason transfer_target is: what a day with no proper falls back to is a rubric, not a universal. Calendar calls it and passes its own temporal function as the callback the rite needs to reach another date. Steps 1 and 2 only: the observed celebration's own proper, else the day's own temporal slug. Nothing encodes "Lent has daily propers" -- the presence of an entry is the discriminator.

kernel(lectionary): slug-keyed reading citations

2026-08-14T14:43:00Z

Data only, the same shape and discipline as Layer: slug-canonical, duplicates rejected at construction naming the offending slug, sexp round-trips. Which slug a day falls back to is a rubric and belongs to the rite module, so nothing here knows about ferias or Sundays.

test: golden pins for the known-tricky years

2026-08-12T04:13:14Z

Validation layer 5: fourteen hand-verified pins in test/test_golden.ml, wired into the suite via test/test_colitur.ml. Each pinned day was computed and checked against its RG citation before being written down, never transcribed from `colitur day` output: - Easter extremes 1598 (earliest, 22 Mar), 1666 (latest, 25 Apr) and 2038 (late-modern instance of the same latest date) -- all three independently computed by hand via the Gauss/Meeus Gregorian Easter algorithm, not read off Computus.gregorian_easter. extreme_years in test_validate.ml already established 1598/1666 as the true 1583-2500 extremes (correcting a stale 1818/2038 comment); this file pins the resolved DAY there, not just the date. - Annunciation transfer, 25 March inside Holy Week (2016: 25 March is literally Good Friday) and the double transfer with St Joseph (2008: RG96's Attamen(a) claims Easter+8 for the Annunciation first, Joseph's own RG96 walk continues past it to the next day). - The RG96 Attamen(a) exception's own CONDITION pinned on both sides: 2057, 2007, 2012 (general walk suffices, lands before Easter, no exception) vs 2016/2024 (walk would cross Easter, Easter+8 fires). - 2011-07-04, the Precious Blood transfer (Sacred Heart outranks it outright on 1 July; the RG96 walk skips Visitation and a Sunday before landing on 4 July). - All Souls falling on a Sunday (2025, RG96 Attamen(b)) and Christmas falling on a Sunday (2022, RG91 entry 1 -- no contest, since 25 December is never an ordinary Sunday candidate in the EF temporal cycle). - Holy Thursday's white amid violet Passiontide (2026, RG128(b)/RG122) -- the case colitur and lectio previously agreed was violet, so the differential could never have caught it; only a golden pin or the missalemeum oracle can. - Advent/Lent Ember ferias commemorated when impeded (1900, 1902, RG24 + RG109(e)) contrasted with IV-class ferias never commemorated (2026, RG26) and RG111(b)'s Sunday rank floor (2009, 2026) -- all three assert the actual displaced/excluded candidate is present in `omitted`, not just that `commemorations` is empty, so a day with no losing candidate at all could not pass vacuously. 2038 is deliberately NOT pinned day-by-day: register item F4 confirms 2038-03-06/08/09 are wrong (a lectio bootstrap-generator defect, not fixable here). Only the Easter-week days, untouched by that bug, are pinned; the exclusion is stated in the test's own comment, not silent. Every weekday asserted was independently cross-checked against `date -d +%A` (glibc, wholly outside this codebase) before being written down. Precedence outcomes were traced against precedence_ef.ml's own band/ disposition/admit/transfer_target, not merely observed to look plausible; where the primary text alone doesn't fully settle an outcome (the 2008 Annunciation/Joseph tie-break, both landing at RG91 table entry 11), the test's own comment says so rather than overclaiming a citation. Perturbation-tested: temporarily broke Holy Thursday's white-colour special case in temporal_ef.ml, confirmed the golden test failed with a clear day-and-field diff, reverted.

test: oracle vs missalemeum 2026-2027; audit the sanctoral

2026-08-12T02:36:36Z

Validation layer 4: an oracle harness against missalemeum (Divinum Officium data), independent of the lectio bootstrap chain colitur's own sanctoral data comes from -- the only layer that can catch an error inherited from that bootstrap, and the only one that can validate commemorations at all (the lectio differential explicitly excludes them, per its own header comment). tools/extract_missalemeum_oracle.py shapes the fixture from lectio's sources/snapshot.tar.gz outside the test (no JSON library in this project's frozen deps, same reasoning test_differential.ml's own fixture already documents). test/fixtures/missalemeum-ef-2026-2027 .txt (730 days, SHA-256 pinned and asserted) + its own .provenance note record exactly how to regenerate it. test_oracle.ml compares three axes the oracle actually supports: rank, colour (SET MEMBERSHIP -- 14 of 730 days carry two colours, e.g. rose+violet on Gaudete/Laetare, which independently vindicates this project's own rose reading against lectio's violet-only one, recorded in the register), and commemoration presence/count. Slug identity is deliberately out of scope (needs a title->slug mapping, the data audit's own business, not the automated comparator's). Of 730 days, 688 matched cleanly outright. The remaining 42 are all named in data/ef/expected-divergences-missalemeum.sexp (14 cited entries, M1-M14): most are genuine primary-source-confirmed findings this task adjudicated and fixed in the two preceding commits (RG 33, RG 109/111, Holy Thursday's colour); the rest are real, cited, deferred feature/data gaps (RG 91 entry 27's BVM-Saturday office, RG 110's inseparable Peter/Paul commemoration, four sanctoral entries missing from lectio's own source) or genuine oracle-side artifacts -- honestly verdicted against whichever side this task's own primary-source research actually backs, never defaulted to colitur. One entry (M13, St Joseph vs the Friday of Passion Week 2027) is verdict open: adjudicated as unresolved after real search effort, not guessed past. The data audit: every sanctoral entry the comparison flagged was hand-checked against the 1962 calendarium, plus a 20-entry deterministic random control sample (seed 20260812) drawn independently of the flagged set. The control sample caught two entries (benedict, frances-rome) marked Commemoration_only in the bootstrapped data when the primary calendarium lists them as plain III-class feasts with their own Office -- traced to lectio's own source, not fixable here, and reported as a signal (10% of a random sample) rather than a blanket claim. Coverage recorded honestly in docs/research/rules-register.md's own three buckets: confirmed by oracle (200/322), confirmed by hand (23/322, 2 of them wrong), unverified (115/322) -- the unverified bucket stated explicitly rather than left implicit. Harness teeth demonstrated and reverted (not committed): a fixture rank/colour edit on a previously-clean day fails both the checksum pin and the no-unexplained-differences assertion independently; an expected_rows drift on the allow-list fails the citation-count assertion. Both captured with their exact failure messages, both reverted before this commit.

test: differential vs lectio 2005-2050 with a cited allow-list

2026-08-12T01:18:15Z

Validation layer 3 (design spec's five): compares colitur's real day-by-day EF output against lectio (sibling project, Go), 2005-2050, one line per civil day. Of 16801 day-pairs, 11206 already agree on the seven leading columns; the 5595 that don't resolve into exactly 25 distinct field-diff signatures, all triaged. Three strictly separate layers, per the controller's ruling (the brief's single flat allow-list assumed a handful of differences, not 5595): - Layer A (test_differential.ml, norm_season/norm_slug): vocabulary. An explicit, closed table of naming synonyms with no liturgical substance (lectio's easter/christmas vs colitur's paschaltide/christmastide; a handful of slugs that are two names for the identical office). No wildcards -- every entry is a literal string pair. - Layer B (strip_epiphany_index): numbering. The one slug family whose lectio/colitur index offset is not a constant (Time-after-Epiphany week numbering, register 3c#5) has its embedded digit stripped to a common form on both sides before comparing; rank and colour stay fully compared. - Layer C (data/ef/expected-divergences.sexp): the cited allow-list. Ten genuine liturgical disagreements, each citing its RG paragraph and naming which engine is right (always colitur, verified against the Missal/register, never against lectio's own behaviour). This is the only layer permitted to cover a difference in rank, colour, or which celebration is observed. Five extend or restate register 3c's already- documented divergences (season boundary, Sunday I-class, Advent Ember ferias, Rogations); five are new, found and adjudicated in this task (Lent Ember days, the Nativity Octave, Ember-day-vs-saint precedence, the St Joseph transfer off a Lent Sunday, and the 2011 Sacred Heart / Precious Blood / Visitation collision). expected_rows on each entry is an exact regression pin, asserted by the test, not documentation. 13 January (register 6's long-open "Baptism of the Lord" item) is confirmed empirically fixed already -- Task 11's sanctoral wiring closed it before this task started -- so it is not allow-listed; the only residual difference there is the season boundary already covered by C1. Two stated limits carried from the brief (commemorations are not comparable; lectio's own EF oracle asserts season only, 2025-2026 only, so a rank/colour difference is not presumptive evidence against colitur) plus a third found during this task (the week column is a display convention on both sides, not a liturgical fact, and is not compared at all) are documented in the test file's own doc comment. Fixture: test/fixtures/lectio-ef-2005-2050.txt, committed as plain text (1.4 MB), generated by lectio commit 2386a45; provenance recorded in the sibling .provenance file. Colitur's side is recomputed fresh from the library on every run, through the same Calendar/Rite_ef pipeline `colitur day` uses, not the compiled binary. Proved the harness has teeth by two reverted perturbations: a genuine colour difference injected into a fully-covered fixture row fails the "no unexplained differences" check with the exact mismatched row printed; a one-row drift in an allow-list entry's expected_rows fails the count check independently, showing it is not merely a duplicate of the first assertion. 236/236 tests green, clean-build verified, deterministic across OCAMLRUNPARAM=R.

rite(ef): clamp the RG96 search at the domain ceiling

2026-08-11T23:16:22Z

search_from could walk up to 400 days past origin before Calendar's own ~start ~stop clamp is ever consulted, and nothing stopped it probing occupant on a date past 31 December 9999 -- occupant chains through the real EF rite's temporal, which calls Computus.gregorian_easter, not total outside 1583..9999 (it Date.makes and failwiths on Error). Not reachable with the shipped sanctoral data alone, but reachable through the project's own primary extension path: an overlay adding an I-class feast on 25 December leaves nothing but Class2 Nativity-octave days for the rest of civil year 9999, so the unguarded search reached 1 January of year 10000 and crashed there with 'computus: year 10000 out of range 1583..9999'. 9999 is an in-range year and the kernel's contract is 'never raises on in-range input'. search_from now also stops, without probing occupant again, once it passes Date's own domain ceiling -- the same 'return a finite date, let Calendar's own out-of-range handling record it, never pretend to have found something admissible' contract the existing step-count guard already follows. Two new tests, both mutation-verified to actually reproduce the crash when the guard is removed (see the task report): a precedence_ef.ml unit test using the real Temporal_ef.temporal as occupant (a synthetic occupant can never discriminate this, since it never calls Computus itself), and a Calendar-level integration test reproducing the exact overlay-based scenario the review found.

data(ef): bootstrap the 1962 sanctoral from lectio

2026-08-11T22:05:57Z

Convert lectio's tridentine-calendar.ini (322 entries) into data/ef/sanctoral.sexp via a validating OCaml converter, tools/ bootstrap_sanctoral.ml, rather than a hand-written script: every field is built through Slug.of_string, Colour.of_string and Vocab_ef.rank_of_string, so the emitted sexp is valid by construction. Two conversion decisions, both documented rather than buried: - subject defaults to Subject.Saint, overriding Celebration.make's kernel default of Subject.Temporal, for the 316 entries with no explicit class; - rank = commemoration maps to status = Commemoration_only with an inferred Class3 (not a citation -- it is what the 1960 reform reduced most simple feasts from), recorded as an open item in the rules register for the oracle to adjudicate. Every celebration is tagged layer = Precedence_ef.universal_layer, the provenance id RG 91's band classifier reads to tell the universal calendar from proper/indult data. The generated file carries a provenance header: source path, its SHA-256, and the UTC conversion date, so re-bootstrapping against a newer lectio is reproducible and diffable. Output is byte-identical across runs. test/test_sanctoral_ef.ml loads the file through Layer.load and checks the counts independently derived from the source INI (322 entries, 114 Commemoration_only, 12 Class1, no Subject.Temporal, every date resolves in a leap year), plus two named spot-checks against the INI's own text -- one entry with an explicit class field, one commemoration -- so a passing count cannot hide the wrong 322 entries having been converted.

rite(ef): RG 91 Table of Precedence

2026-08-11T20:04:59Z

Precedence_ef.band transcribes RG 91's 28-entry Table of Precedence (rules-register.md §4) for the EF rite: given a day's context and a candidate celebration, returns the table's own entry number, 1-28 (I class 1-13, II class 14-21, III class 22-26, IV class 27-28); lower wins. Every branch carries its entry number and register citation in a comment, checked in the table's own numeric order. Two entries are transcribed as the register states them even though they invert the pattern the rest of the table follows: at III class, 23 (particular calendars) outranks 24 (universal), the reverse of how 11/12 and 14/16/19/20 rank a universal feast ahead of a proper one at I and II class. Sanctoral-origin, layer-decided entries (11-13, 14/16/19/20, 23/24) follow the brief's structural insight: a celebration whose layer is not the universal base is an overlay -- proper, or indult if its layer id also carries the indult prefix. Neither the universal-layer id nor the indult prefix is an RG citation; both are colitur's own data-modelling convention, exposed from the module so whichever task loads the real EF sanctoral overlays can align to them. Vigils (21, 26) are read off the temporal cycle's own -vigil slug suffix rather than gated on origin, since a II/III-class vigil can be either temporal-origin (Ascension, already produced by temporal_ef) or sanctoral-origin (a saint's vigil, no task has loaded yet); Ember days (part of entry 18) are read off temporal_ef's own ember slug prefixes rather than re-derived, since the September anchor is independently flagged there as one of the more contested dates in the calendar. A candidate shape the table has no row for (e.g. a Class1 vigil that is not Nativity or Pentecost, or a Class4 candidate marked as a vigil -- RG 91 has no IV-class vigil either) returns a dedicated unclassified sentinel (max_int) rather than being folded into a same-rank entry it does not belong to. test_precedence_ef.ml is table-driven: one Alcotest.test_case per RG 91 entry (55 rows total, several entries covered by more than one named day so a single missed offset cannot hide behind a passing sibling), each date computed from Computus.gregorian_easter rather than hand-typed, so an arithmetic slip cannot pass by accident.

kernel(calendar): the year is the primitive, the day is derived

2026-08-11T18:01:54Z

Transfers make per-date resolution impossible to do correctly: resolving 25 March can push a feast onto 26 March, and RG 97-98 has coinciding I-class feasts transfer in table order, which needs global knowledge. So year computes a whole liturgical year in one pass and day indexes into it. Pure, no cache, no mutable state. This commit resolves each day but does not yet place deferred transfers; they are recorded with a reason. Task 6 adds the placement pass.