feat(cli): lectio --corpus-check validator

1 files changed, 93 insertions, 0 deletions

diff --git a/internal/bible/validate.go b/internal/bible/validate.go
new file mode 100644
index 0000000..27ec519
--- /dev/null
+++ b/internal/bible/validate.go
@@ -0,0 +1,93 @@
+package bible
+
+import (
+	"fmt"
+	"sort"
+)
+
+// CorpusReport is the outcome of validating one corpus's sidecar + text (see
+// CheckCorpus). Errors mean the corpus is unusable/malformed and fail
+// --corpus-check (exit 1); Warnings flag coverage gaps against the reference
+// Vulgate ("vul") and never fail the check -- a corpus may legitimately be
+// incomplete (the built-in wuj is) and still be a valid drop-in.
+type CorpusReport struct {
+	Code     string
+	Errors   []string
+	Warnings []string
+}
+
+// OK reports whether the corpus is usable: no errors. Warnings never affect it.
+func (r CorpusReport) OK() bool { return len(r.Errors) == 0 }
+
+var validPsalmSystems = map[string]bool{"vulgate": true, "hebrew": true, "drb": true}
+
+// CheckCorpus validates code's sidecar and text and reports coverage gaps vs
+// "vul" as warnings.
+//
+// Sidecar: lang, name and psalm_system are required (psalm_system must be
+// vulgate/hebrew/drb); sigla and autoselect are optional -- their absence, or
+// autoselect=false, is never flagged (the built-in wuj sets autoselect=false
+// and must pass clean).
+//
+// Text: every Book column value must be one of CanonicalBooks' 73 keys
+// (error). A chapter missing entirely versus vul, a verse-number gap within a
+// chapter, or a repeated verse number within a chapter, is a warning only --
+// real source texts legitimately do this (e.g. the LXX's lettered doublet
+// verses in 3 Kingdoms, or a scanned translation's occasional merged verse),
+// and it must never turn a corpus that otherwise loads fine into a failure.
+func CheckCorpus(code string) CorpusReport {
+	r := CorpusReport{Code: code}
+
+	m, ok := Meta(code)
+	if !ok || m.Lang == "" || m.Name == "" || m.PsalmSystem == "" {
+		r.Errors = append(r.Errors, "missing or incomplete sidecar (need lang, name, psalm_system)")
+	}
+	if m.PsalmSystem != "" && !validPsalmSystems[m.PsalmSystem] {
+		r.Errors = append(r.Errors, fmt.Sprintf("invalid psalm_system %q (want vulgate|hebrew|drb)", m.PsalmSystem))
+	}
+
+	c := load(code)
+	if len(c.books) == 0 {
+		r.Errors = append(r.Errors, "no verses parsed (empty or malformed .tsv)")
+		sort.Strings(r.Errors)
+		return r
+	}
+
+	canon := CanonicalBooks()
+	for book := range c.books {
+		if !canon[book] {
+			r.Errors = append(r.Errors, "unknown book name: "+book)
+		}
+	}
+
+	ref := load("vul")
+	for book, chaps := range c.books {
+		for ch, verses := range chaps {
+			seen := map[int]bool{}
+			maxV := 0
+			for _, v := range verses {
+				if seen[v.Verse] {
+					r.Warnings = append(r.Warnings, fmt.Sprintf("%s %d: duplicate verse %d", book, ch, v.Verse))
+				}
+				seen[v.Verse] = true
+				if v.Verse > maxV {
+					maxV = v.Verse
+				}
+			}
+			if maxV > len(seen) {
+				r.Warnings = append(r.Warnings, fmt.Sprintf("%s %d: verse gap (have %d verse(s), highest numbered %d)", book, ch, len(seen), maxV))
+			}
+		}
+		if refChaps, ok := ref.books[book]; ok {
+			for ch := range refChaps {
+				if _, have := chaps[ch]; !have {
+					r.Warnings = append(r.Warnings, fmt.Sprintf("%s: missing chapter %d (present in vul)", book, ch))
+				}
+			}
+		}
+	}
+
+	sort.Strings(r.Errors)
+	sort.Strings(r.Warnings)
+	return r
+}