From 2a005692eee62bfe8425ebf800e758728775bbfe Mon Sep 17 00:00:00 2001
From: Leo Vasanko <git@vasanko.com>
Date: Fri, 23 Jan 2026 21:31:54 +0000
Subject: [PATCH] Fixes to JSONL management, starting from empty state rather
 than default DB.

---
 paskia/db/jsonl.py      | 43 ++++++++++++++++++++++-------------------
 paskia/db/operations.py | 19 +++++++++++++-----
 2 files changed, 37 insertions(+), 25 deletions(-)

diff --git a/paskia/db/jsonl.py b/paskia/db/jsonl.py
index a009321..2dff383 100644
--- a/paskia/db/jsonl.py
+++ b/paskia/db/jsonl.py
@@ -32,7 +32,7 @@ class _ChangeRecord(msgspec.Struct):
 _change_encoder = msgspec.json.Encoder()
 
 
-async def load_jsonl(db_path: Path, empty_data: dict) -> dict:
+async def load_jsonl(db_path: Path) -> dict:
     """Load data from disk by applying change log.
 
     Replays all changes from JSONL file using plain dicts (to handle
@@ -40,29 +40,32 @@ async def load_jsonl(db_path: Path, empty_data: dict) -> dict:
 
     Args:
         db_path: Path to the JSONL database file
-        empty_data: Empty data structure to start with (as dict)
 
     Returns:
         The final state after applying all changes
+
+    Raises:
+        ValueError: If file doesn't exist or cannot be loaded
     """
-    data_dict = empty_data.copy()
-    if db_path.exists():
-        try:
-            # Read entire file at once and split into lines
-            async with aiofiles.open(db_path, "rb") as f:
-                content = await f.read()
-            for line_num, line in enumerate(content.split(b"\n"), 1):
-                line = line.strip()
-                if not line:
-                    continue
-                try:
-                    change = msgspec.json.decode(line)
-                    # Apply the diff to current state (marshal=True for $-prefixed keys)
-                    data_dict = jsondiff.patch(data_dict, change["diff"], marshal=True)
-                except Exception as e:
-                    raise ValueError(f"Error parsing line {line_num}: {e}")
-        except (OSError, ValueError, msgspec.DecodeError) as e:
-            raise ValueError(f"Failed to load database: {e}")
+    if not db_path.exists():
+        raise ValueError(f"Database file not found: {db_path}")
+    data_dict: dict = {}
+    try:
+        # Read entire file at once and split into lines
+        async with aiofiles.open(db_path, "rb") as f:
+            content = await f.read()
+        for line_num, line in enumerate(content.split(b"\n"), 1):
+            line = line.strip()
+            if not line:
+                continue
+            try:
+                change = msgspec.json.decode(line)
+                # Apply the diff to current state (marshal=True for $-prefixed keys)
+                data_dict = jsondiff.patch(data_dict, change["diff"], marshal=True)
+            except Exception as e:
+                raise ValueError(f"Error parsing line {line_num}: {e}")
+    except (OSError, ValueError, msgspec.DecodeError) as e:
+        raise ValueError(f"Failed to load database: {e}")
     return data_dict
 
 
diff --git a/paskia/db/operations.py b/paskia/db/operations.py
index fff966a..db0fbc5 100644
--- a/paskia/db/operations.py
+++ b/paskia/db/operations.py
@@ -73,13 +73,22 @@ class DB:
         self._current_actor: str = "system"
 
     async def load(self, db_path: str | None = None) -> None:
-        """Load data from JSONL change log."""
+        """Load data from JSONL change log.
+
+        If file doesn't exist, keeps the initialized empty structure and
+        sets _previous_builtins to {} for creating a new database.
+        """
         if db_path is not None:
             self.db_path = Path(db_path)
-        empty = msgspec.to_builtins(self._data)
-        data_dict = await load_jsonl(self.db_path, empty)
-        self._data = _json_decoder.decode(_json_encoder.encode(data_dict))
-        self._previous_builtins = msgspec.to_builtins(self._data)
+        try:
+            data_dict = await load_jsonl(self.db_path)
+            self._data = _json_decoder.decode(_json_encoder.encode(data_dict))
+            # Track the JSONL file state directly - this is what we diff against
+            self._previous_builtins = data_dict
+        except ValueError:
+            if self.db_path.exists():
+                raise  # File exists but failed to load - re-raise
+            # File doesn't exist: keep initialized _data, _previous_builtins stays {}
 
     def _queue_change(self) -> None:
         current = msgspec.to_builtins(self._data)