CI: build + validate ui_datadict.db

.github/workflows/build.yml

@@ -26,6 +26,10 @@ jobs:
       - name: Regenerate data model + coverage badge
         run: python3 tools/gen_diagram.py
 
+      # Rebuild the UI / governance projection from the committed datadict.
+      - name: Build UI projection
+        run: python3 tools/build_ui_dict.py
+
       # Assert semantic invariants (not git-diff byte-identity, which is noisy
       # because of timestamps in the DB and the report's generated-date line).
       - name: Verify invariants

CHANGELOG.md

@@ -21,6 +21,12 @@ chronological development log see [`PROGRESS.md`](PROGRESS.md).
   source mask (e.g. GS1) or generated from type/allowed-values/scale.
   `datadict.db` is read-only input; the projection is rebuilt from scratch.
 
+### Changed
+
+- CI now rebuilds the UI projection and `tools/ci_check.py` validates
+  `ui_datadict.db` (completeness, FK integrity, `ByteLength = CharLength*4`,
+  positive `CharLength`, unique `(GroupID, Name)`).
+
 ## [1.2.1] - 2026-06-27
 
 ### Added

tools/ci_check.py

@@ -24,6 +24,55 @@
 from curated_descriptions import CURATED  # noqa: E402
 
 DB = os.path.join(ROOT, "datadict.db")
+UI_DB = os.path.join(ROOT, "ui_datadict.db")
+
+
+def check_ui(datadict_items):
+    """Invariants for the derived UI projection (ui_datadict.db)."""
+    fail = []
+    if not os.path.exists(UI_DB):
+        return [f"{UI_DB} not found (run build_ui_dict.py)"]
+    u = sqlite3.connect(UI_DB)
+    u.row_factory = sqlite3.Row
+    items = u.execute("SELECT COUNT(*) n FROM UI_DataItems").fetchone()["n"]
+    groups = u.execute("SELECT COUNT(*) n FROM Groups").fetchone()["n"]
+    orphan_group = u.execute(
+        "SELECT COUNT(*) n FROM UI_DataItems x WHERE NOT EXISTS "
+        "(SELECT 1 FROM Groups g WHERE g.GroupID = x.GroupID)").fetchone()["n"]
+    bad_groupcat = u.execute(
+        "SELECT COUNT(*) n FROM Groups g WHERE NOT EXISTS "
+        "(SELECT 1 FROM Categories c WHERE c.CategoryID = g.CategoryID)"
+    ).fetchone()["n"]
+    bad_bytes = u.execute(
+        "SELECT COUNT(*) n FROM UI_DataItems "
+        "WHERE ByteLength IS NOT CharLength * 4").fetchone()["n"]
+    bad_charlen = u.execute(
+        "SELECT COUNT(*) n FROM UI_DataItems "
+        "WHERE CharLength IS NULL OR CharLength <= 0").fetchone()["n"]
+    dupes = u.execute(
+        "SELECT COUNT(*) n FROM (SELECT GroupID, Name FROM UI_DataItems "
+        "GROUP BY GroupID, Name HAVING COUNT(*) > 1)").fetchone()["n"]
+
+    print(f"ui_items={items}  groups={groups}  orphan_groups={orphan_group}  "
+          f"bad_group_category={bad_groupcat}  byte!=char*4={bad_bytes}  "
+          f"bad_charlength={bad_charlen}  dup_(group,name)={dupes}")
+
+    if groups < 1:
+        fail.append("UI: no groups")
+    if items != datadict_items:
+        fail.append(f"UI: item count {items} != datadict {datadict_items} "
+                    "(projection should keep every item)")
+    if orphan_group:
+        fail.append(f"UI: {orphan_group} items reference a missing group")
+    if bad_groupcat:
+        fail.append(f"UI: {bad_groupcat} groups reference a missing category")
+    if bad_bytes:
+        fail.append(f"UI: {bad_bytes} rows where ByteLength != CharLength*4")
+    if bad_charlen:
+        fail.append(f"UI: {bad_charlen} rows with null/non-positive CharLength")
+    if dupes:
+        fail.append(f"UI: {dupes} duplicate (GroupID, Name) pairs")
+    return fail
 
 
 def main():
@@ -68,6 +117,8 @@ def main():
         fail.append(f"{orphan_curated} curated descriptions reference a "
                     "non-existent item Name")
 
+    fail += check_ui(total)
+
     if fail:
         print("FAIL: " + "; ".join(fail))
         return 1