1 files changed, 40 insertions, 0 deletions
diff --git a/lib/rites/rite_of/precedence_of.mli b/lib/rites/rite_of/precedence_of.mli
new file mode 100644
index 0000000..dfee504
--- /dev/null
+++ b/lib/rites/rite_of/precedence_of.mli
@@ -0,0 +1,40 @@
+(** The Tabula dierum liturgicorum of the 2002 Missale Romanum -- 13 numbered
+    entries in three parts -- plus Normae n. 60's occurrence and transfer
+    rules. The OF counterpart of {!Rite_ef.Precedence_ef}, and deliberately
+    the same shape.
+
+    Every citation names its document ("Normae n. N" or "IGMR n. N"): the
+    2002 Missal carries both, independently numbered from 1. *)
+
+open Colitur_kernel
+
+(** Returned for a candidate shape the Tabula has no row for. [max_int], so an
+    unclassified candidate always loses. Task 4's property asserts this is
+    never reached by anything {!Rite_of.Temporal_of} constructs. *)
+val unclassified : int
+
+(** The universal (General Roman Calendar) layer id. Not a Normae citation:
+    the Tabula distinguishes universal entries (3, 7, 10) from proper ones
+    (4, 8, 11), and [Celebration.t.layer] is the only field carrying that
+    distinction. *)
+val universal_layer : string
+
+(** Prefix marking a candidate as belonging to a PROPER (diocesan, national or
+    religious) calendar -- Tabula entries 4, 8 and 11.
+
+    NO SHIPPED DATA PRODUCES THIS. Phase 3 transcribes the General Roman
+    Calendar only (the spec's own universal-calendar-only discipline), so
+    entries 4, 8 and 11 are unreachable on shipped data and are exercised in
+    tests by synthetic candidates alone. They are transcribed anyway for two
+    reasons: the transcription of a 13-entry table should be complete rather
+    than silently 10 entries long, and a user [--overlay] carrying a local
+    patron becomes correct without a later edit here. *)
+val proper_prefix : string
+
+(** [band ctx c]: the Tabula's own entry number for [c], TIMES TEN (entry 1 ->
+    10, entry 13 -> 130); lower wins. {!unclassified} for a shape with no row.
+
+    The ×10 scaling matches {!Rite_ef.Precedence_ef.band} and exists for the
+    same reason: a future sub-rank fits between neighbours without rescaling
+    the table. *)
+val band : Vocab_of.season Precedence.context -> Vocab_of.rank Precedence.candidate -> int