fix: backfill district case_type from detail page

ngm/ngscrape/spiders/district_case_enrichment.py

@@ -102,27 +102,51 @@ class DistrictCaseEnrichmentSpider(scrapy.Spider):
         # "DOWNLOAD_DELAY": 2,  # 2 second delay between requests
     }
 
-    def __init__(self, *args, **kwargs):
+    def __init__(self, *args, backfill_case_type=False, **kwargs):
         super().__init__(*args, **kwargs)
+        # Opt-in one-off mode (`-a backfill_case_type=true`) that also revisits
+        # already-enriched rows missing a case_type. Off by default so nightly
+        # runs never re-fetch permanently-typeless rows in a loop.
+        self.backfill_case_type = str(backfill_case_type).lower() in (
+            "1",
+            "true",
+            "yes",
+        )
 
     def start_requests(self):
         """Generate requests for cases that need enrichment"""
         self.engine = get_engine()
         init_db(self.engine)
         self.session = get_session(self.engine)
 
+        needs_enrichment = or_(
+            CourtCase.status == "pending",
+            CourtCase.status.is_(None),
+        )
+        if self.backfill_case_type:
+            # Revisit already-enriched rows whose case_type never got populated:
+            # the daily cause-list only carries the (often blank) "मुद्दा विषय"
+            # subject column, so these need the detail page's "मुद्दाको किसिम".
+            needs_enrichment = or_(
+                needs_enrichment,
+                and_(
+                    CourtCase.status == "enriched",
+                    or_(
+                        CourtCase.case_type.is_(None),
+                        CourtCase.case_type == "",
+                    ),
+                ),
+            )
+
         # Query all district court cases that need enrichment in one go
-        # Priority: newer registration dates first, status = pending or NULL
+        # Priority: newer registration dates first.
         with self.session.begin():
             cases_to_enrich = (
                 self.session.query(CourtCase.case_number, CourtCase.court_identifier)
                 .filter(
                     and_(
                         CourtCase.court_identifier.like("%dc"),
-                        or_(
-                            CourtCase.status == "pending",
-                            CourtCase.status.is_(None),
-                        ),
+                        needs_enrichment,
                     )
                 )
                 .order_by(CourtCase.registration_date_ad.desc().nullslast())
@@ -259,7 +283,23 @@ def parse_case_detail(self, response):
                 return
 
             if case.status == "enriched":
-                self.logger.info(f"Case {case_number} already enriched, skipping")
+                if case.case_type:
+                    self.logger.info(f"Case {case_number} already enriched, skipping")
+                    return
+                # Already enriched but missing case_type: backfill just that field
+                # from the detail page without re-touching entities/hearings (whose
+                # rebuild deletes rows that may have downstream linkages).
+                case_type = self._extract_enrichment_data(soup).get("case_type")
+                if case_type:
+                    case.case_type = case_type[:200]
+                    case.updated_at = datetime.now(KATHMANDU_TZ).replace(tzinfo=None)
+                    self.logger.info(
+                        f"Backfilled case_type for {case_number} ({code_name}): {case_type}"
+                    )
+                else:
+                    self.logger.info(
+                        f"Case {case_number} ({code_name}) has no case_type on detail page"
+                    )
                 return
 
         # Extract enrichment data
@@ -296,6 +336,11 @@ def _extract_enrichment_data(self, soup: BeautifulSoup) -> Dict:
                     # Map Nepali labels to database fields
                     if label == "रजिष्ट्रेशन नं" and value:
                         data["registration_number"] = value[:100]
+                    elif label == "मुद्दाको किसिम" and value:
+                        # True case type, only available on the detail page; the
+                        # cause-list listing has no such column (it carries the
+                        # subject), which is why district case_type is often blank.
+                        data["case_type"] = value[:200]
                     elif label == "मुद्दाको बिषय" and value:
                         data["case_subject"] = value
                     elif label == "मुद्दाको स्थिति" and value: