Skip to main content

The case-status vocabulary has three coexisting generations

There is no single source of truth for what a case's status can be. Three naming generations coexist in the live system:

WhereValues used
src/lib/workflow-config.ts (CaseStatus)INTAKE, PRELIMINARY_INQUIRY, CHARGES_ISSUED, CONSULTATION_PORTAL, RESPONSE_WINDOW, EVIDENCE_PHASE, BODIES_OPINION, FINAL_REPORT, CLOSED, SUSPENDED, ARCHIVED (8 operational + 3 terminal)
useWorkflow.ts runtime guardsadditionally special-cases legacy aliases INQUERITO_PREVIO and NOTA_CULPA
seed-test-data edge functionDRAFT, INTAKE, PRELIMINARY_INQUIRY, FORMAL_PROCEEDING, CHARGES_NOTICE, CLOSED
Postgres case_status enum~24 values accreted across three migrations (see below)

The database case_status enum was:

  • v1: new, in_progress, under_review, closed
  • v2 (destructive DROP+CREATE with a data remap): DRAFT, INTAKE, PRELIMINARY_INQUIRY, FORMAL_PROCEEDING, CHARGES_NOTICE, EMPLOYEE_RESPONSE, EVIDENCE_PHASE, DECISION_DRAFT, DECISION_ISSUED, CLOSED, SUSPENDED, CANCELLED
  • v3 (ADD VALUE ×11): FORMAL_OPEN, PREPARE_CHARGES, CHARGES_REVIEW, CHARGES_ISSUED, CONSULTATION_PORTAL, RESPONSE_WINDOW, BODIES_CHECK, BODIES_OPINION, FINAL_REPORT, DECISION_REVIEW, ARCHIVED

Postgres enums cannot drop values, so every legacy value is still a legal state in the column. The column DEFAULT is 'DRAFT' — a value workflow-config.ts does not model at all.

The tacit assumption: that a case's status is one of the 8 the UI knows about. In reality the DB will happily hold FORMAL_PROCEEDING, new, or DECISION_ISSUED, and the UI's STATUS_LABELS[status] lookup returns undefined for those — blank badges, broken timelines, getFilteredTransitions returning [] (a stuck case).

What breaks: any rebuild that models status as a clean 8-value enum will choke on real historical rows. Migration of existing data requires an explicit mapping table across all three generations, and a decision on the orphan states.

Rebuild implication​

Define one canonical status set, write an explicit legacy→canonical map, and migrate the column with a CHECKed conversion. Do not assume the data is clean.