FDN-01: repository & projektgerüst

Git-Repository für bestehenden archivdms-Code initialisiert, Branch-/Commit-Konvention (feature/<ticket>-<slug>-Branches, Ticket-Prefix in Commit-Nachricht) etabliert.
2026-08-11 21:27:53 +02:00
parent 40ed80da71
commit 9a24ea29e1
274 changed files with 53708 additions and 0 deletions
@@ -0,0 +1,46 @@
+-- 024_ocr_words.sql
+-- Documentation only — applied via internal/storage/ocr_words.go
+-- (Store.initOCRWordsSchema), wired into Store.initSchema in documents.go.
+--
+-- Phase 2 of the OCR text-highlight/overlay feature (Phase 1: internal/ocr
+-- gained Result.Words []WordBox, coordinates already mapped back into the
+-- original uploaded file's coordinate space — see internal/ocr/coords.go).
+-- This migration adds the table that persists those word boxes per
+-- document, so a future Phase 3 read endpoint can serve them for a
+-- search-highlight/overlay UI without re-running OCR.
+--
+-- No tenant_id column: access is always mediated through document_id (join
+-- or verify against documents.tenant_id), never queried directly by tenant.
+--
+-- ON DELETE CASCADE on document_id: word boxes are a derived index over a
+-- document's OCR text/original file, not a GoBD document themselves, so
+-- cascading on hard-delete (trash workflow, see 007_trash.sql) is correct
+-- and does not affect WORM retention of the document's own storage_path/
+-- content_hash.
+--
+-- Every OCR (re)run (upload async job, manual reprocess endpoint, and the
+-- `documents reprocess-all` CLI, which shares ReprocessDocument with the
+-- endpoint) deletes this document's existing rows and re-inserts the fresh
+-- set (storage.ReplaceOCRWords), so re-OCR never accumulates duplicates.
+
+CREATE TABLE IF NOT EXISTS ocr_words (
+    id          BIGSERIAL PRIMARY KEY,
+    document_id BIGINT NOT NULL REFERENCES documents(id) ON DELETE CASCADE,
+    page        INTEGER NOT NULL DEFAULT 1,
+    block       INTEGER NOT NULL DEFAULT 0,
+    par         INTEGER NOT NULL DEFAULT 0,
+    line        INTEGER NOT NULL DEFAULT 0,
+    word_text   TEXT NOT NULL,
+    "left"      INTEGER NOT NULL,
+    top         INTEGER NOT NULL,
+    width       INTEGER NOT NULL,
+    height      INTEGER NOT NULL,
+    confidence  DOUBLE PRECISION NOT NULL DEFAULT 0
+);
+
+CREATE INDEX IF NOT EXISTS idx_ocr_words_document ON ocr_words (document_id);
+
+-- Plain btree, not pg_trgm/GIN: initSchema cannot assume CREATE EXTENSION
+-- privileges on the live server. Revisit if Phase 3 needs fuzzy/substring
+-- prefiltering (would require enabling pg_trgm out-of-band first).
+CREATE INDEX IF NOT EXISTS idx_ocr_words_word_text ON ocr_words (word_text);