Enhance OCR timestamp extraction with flexible regex matching and improved error handling

main.py

@@ -38,6 +38,9 @@ CAT_RE = re.compile(r"^([0-9a-fA-F-]{36})_categories\.json$")
 OCR_DT_RE = re.compile(
     r"(?P<h>\d{2}):(?P<m>\d{2}):(?P<s>\d{2})\s+(?P<d>\d{2})-(?P<mo>\d{2})-(?P<y>\d{4})"
 )
+OCR_DT_RE_FLEX = re.compile(
+    r"(?P<h>\d{2}):(?P<m>\d{2}):(?P<s>\d{2})\s*(?P<d>\d{2})[-.](?P<mo>\d{2})[-.](?P<y>\d{4})"
+)
 
 
 @dataclass
@@ -156,27 +159,57 @@ def ocr_extract_timestamp(jpg_bytes: bytes, crop_w_frac: float, crop_h_frac: flo
     cfg = r'--oem 3 --psm 6 -c tessedit_char_whitelist=0123456789:- '
     text = pytesseract.image_to_string(bw, config=cfg).strip()
 
-    m = OCR_DT_RE.search(text)
+    # retry without thresholding
-    if not m:
+    text2 = pytesseract.image_to_string(gray, config=cfg).strip()
-        # retry without thresholding
-        text2 = pytesseract.image_to_string(gray, config=cfg).strip()
-        m = OCR_DT_RE.search(text2)
-        if not m:
-            raise ValueError(f"OCR timestamp not found. OCR1='{text}' OCR2='{text2}'")
 
-    dt_local_naive = datetime(
+    dt_local_naive = _parse_ocr_datetime(text, text2)
-        int(m.group("y")),
-        int(m.group("mo")),
-        int(m.group("d")),
-        int(m.group("h")),
-        int(m.group("m")),
-        int(m.group("s")),
-    )
 
     # Keep camera-local timezone (no UTC conversion requested)
     return dt_local_naive.replace(tzinfo=customer_tzinfo)
 
 
+def _parse_ocr_datetime(*texts: str) -> datetime:
+    def build_dt(h, m, s, d, mo, y):
+        return datetime(int(y), int(mo), int(d), int(h), int(m), int(s))
+
+    def valid_dt(h, m, s, d, mo, y) -> bool:
+        try:
+            build_dt(h, m, s, d, mo, y)
+            return True
+        except Exception:
+            return False
+
+    # 1) direct regex matches
+    for t in texts:
+        for rx in (OCR_DT_RE, OCR_DT_RE_FLEX):
+            m = rx.search(t)
+            if m and valid_dt(m.group("h"), m.group("m"), m.group("s"), m.group("d"), m.group("mo"), m.group("y")):
+                return build_dt(m.group("h"), m.group("m"), m.group("s"), m.group("d"), m.group("mo"), m.group("y"))
+
+    # 2) find time and date separately (handles missing space between time and date)
+    time_rx = re.compile(r"(?P<h>\d{2}):(?P<m>\d{2}):(?P<s>\d{2})")
+    date_rx = re.compile(r"(?P<d>\d{2})[-.](?P<mo>\d{2})[-.](?P<y>\d{4})")
+    for t in texts:
+        tm = time_rx.search(t)
+        dm = date_rx.search(t)
+        if tm and dm and valid_dt(tm.group("h"), tm.group("m"), tm.group("s"), dm.group("d"), dm.group("mo"), dm.group("y")):
+            return build_dt(tm.group("h"), tm.group("m"), tm.group("s"), dm.group("d"), dm.group("mo"), dm.group("y"))
+
+    # 3) digits-only fallback: scan for HHMMSSDDMMYYYY
+    for t in texts:
+        digits = re.sub(r"\D", "", t)
+        for i in range(0, max(0, len(digits) - 13)):
+            chunk = digits[i:i + 14]
+            if len(chunk) < 14:
+                continue
+            h, m, s = chunk[0:2], chunk[2:4], chunk[4:6]
+            d, mo, y = chunk[6:8], chunk[8:10], chunk[10:14]
+            if valid_dt(h, m, s, d, mo, y):
+                return build_dt(h, m, s, d, mo, y)
+
+    raise ValueError(f"OCR timestamp not found. OCR1='{texts[0] if texts else ''}' OCR2='{texts[1] if len(texts) > 1 else ''}'")
+
+
 def exif_write_with_exiftool(jpg_in: bytes, dt_original_local: datetime, dt_digitized_utc: datetime) -> bytes:
     """
     DateTimeOriginal = OCR time (camera local) + OffsetTimeOriginal
@@ -280,11 +313,11 @@ def job_set_status(cur, uuid: str, status: str, *, err: Optional[str] = None, pr
     UPDATE remote_cam.import_job
        SET status = %s,
            attempts = attempts + CASE WHEN %s = 'ERROR' THEN 1 ELSE 0 END,
-           last_error = CASE WHEN %s IS NULL THEN last_error ELSE %s END,
+           last_error = COALESCE(%s::text, last_error),
            processed_jpg_key = COALESCE(%s, processed_jpg_key),
            thumbnail_key = COALESCE(%s, thumbnail_key)
      WHERE image_uuid = %s;
-    """, (status, status, err, err, processed_key, thumb_key, uuid))
+    """, (status, status, err, processed_key, thumb_key, uuid))
 
 
 def insert_resource(cur, uuid: str, typ: str):
@@ -314,7 +347,13 @@ def insert_metadata(cur, uuid: str, meta: dict):
     lat_f = float(lat) if lat not in (None, "") else None
     lon_f = float(lon) if lon not in (None, "") else None
 
-    # NOTE: DateTimeLastSettings looks like "05:45:35 30.06.2025" in sample; parse if needed later.
+    dls_raw = get("DateTimeLastSettings")
+    dls = None
+    if dls_raw:
+        try:
+            dls = datetime.strptime(dls_raw, "%H:%M:%S %d.%m.%Y").replace(tzinfo=timezone.utc)
+        except Exception:
+            dls = None
     cur.execute("""
     INSERT INTO remote_cam.metadata(
       metadata_uuid, import_ts, "ImageCreationDate", "SizeInByte", "ImageFormat", "ImageNight",
@@ -366,7 +405,7 @@ def insert_metadata(cur, uuid: str, meta: dict):
       servertime = EXCLUDED.servertime;
     """, (
         uuid, icd, get("SizeInByte"), get("ImageFormat"), get("ImageNight"),
-        get("DateTimeLastSettings"), get("BatteryStatus"), get("FirmwareVersion"),
+        dls, get("BatteryStatus"), get("FirmwareVersion"),
         get("SignalStrength"), get("Temperature"),
         get("CoordinateSwitch"), lat_f, lon_f, get("WorkPeriod"),
         get("WorkStart"), get("WorkEnd"), get("ThumbnailSize"),