viewer: raw-bytes prefilter to skip JSON parsing on non-matching lines

viewer.py

@@ -219,17 +219,25 @@ class Handler(http.server.BaseHTTPRequestHandler):
         need = (page + 1) * size
         rows, scanned = [], 0
         path = self.server.journal
+        # Cheap raw-bytes prefilter: skip JSON parsing for lines that cannot match (event/q/peer/payload).
+        event_token = ('"event": "%s"' % kind).encode() if kind else None
+        payload_token = b'"evidence"' if payload_only else None
         if path.is_file():
-            # ponytail: reverse scan is bounded by page depth, not file size, for unfiltered paging;
-            # add a sqlite FTS index only if filtered deep paging on multi-GB journals gets slow.
+            # ponytail: still O(file) for a rare term on a multi-GB journal; add a sqlite FTS5 index
+            # (stdlib, incremental) only once the journal reaches a few hundred MB.
             for raw in reverse_lines(path):
                 scanned += 1
-                line = raw.decode('utf-8', 'replace')
-                low = line.lower()
-                if any(term not in low for term in terms):
+                low = raw.lower() if (terms or peer) else None
+                if terms and any(term.encode() not in low for term in terms):
+                    continue
+                if peer and peer.encode() not in low:
+                    continue
+                if event_token and event_token not in raw:
+                    continue
+                if payload_token and payload_token not in raw:
                     continue
                 try:
-                    row = json.loads(line)
+                    row = json.loads(raw)
                 except ValueError:
                     continue
                 if kind and row.get('event') != kind: