fix: database lock

mapillary_tools/history.py

@@ -1,24 +1,17 @@
 from __future__ import annotations
 
-import contextlib
-import dbm
 import json
 import logging
+import os
+import sqlite3
 import string
 import threading
 import time
 import typing as T
+from functools import wraps
 from pathlib import Path
 
-# dbm modules are dynamically imported, so here we explicitly import dbm.sqlite3 to make sure pyinstaller include it
-# Otherwise you will see: ImportError: no dbm clone found; tried ['dbm.sqlite3', 'dbm.gnu', 'dbm.ndbm', 'dbm.dumb']
-try:
-    import dbm.sqlite3  # type: ignore
-except ImportError:
-    pass
-
-
-from . import constants, types
+from . import constants, store, types
 from .serializer.description import DescriptionJSONSerializer
 
 JSONDict = T.Dict[str, T.Union[str, int, float, None]]
@@ -85,103 +78,140 @@ def write_history(
         fp.write(json.dumps(history))
 
 
+def _retry_on_database_lock_error(fn):
+    """
+    Decorator to retry a function if it raises a sqlite3.OperationalError with
+    "database is locked" in the message.
+    """
+
+    @wraps(fn)
+    def wrapper(*args, **kwargs):
+        while True:
+            try:
+                return fn(*args, **kwargs)
+            except sqlite3.OperationalError as ex:
+                if "database is locked" in str(ex).lower():
+                    LOG.warning(f"{str(ex)}")
+                    LOG.info("Retrying in 1 second...")
+                    time.sleep(1)
+                else:
+                    raise ex
+
+    return wrapper
+
+
 class PersistentCache:
-    _lock: contextlib.nullcontext | threading.Lock
+    _lock: threading.Lock
 
     def __init__(self, file: str):
-        # SQLite3 backend supports concurrent access without a lock
-        if dbm.whichdb(file) == "dbm.sqlite3":
-            self._lock = contextlib.nullcontext()
-        else:
-            self._lock = threading.Lock()
         self._file = file
+        self._lock = threading.Lock()
 
     def get(self, key: str) -> str | None:
+        if not self._db_existed():
+            return None
+
         s = time.perf_counter()
 
-        with self._lock:
-            with dbm.open(self._file, flag="c") as db:
-                value: bytes | None = db.get(key)
+        with store.KeyValueStore(self._file, flag="r") as db:
+            try:
+                raw_payload: bytes | None = db.get(key)  # data retrieved from db[key]
+            except Exception as ex:
+                if self._table_not_found(ex):
+                    return None
+                raise ex
 
-        if value is None:
+        if raw_payload is None:
             return None
 
-        payload = self._decode(value)
+        data: JSONDict = self._decode(raw_payload)  # JSON dict decoded from db[key]
 
-        if self._is_expired(payload):
+        if self._is_expired(data):
             return None
 
-        file_handle = payload.get("file_handle")
+        cached_value = data.get("value")  # value in the JSON dict decoded from db[key]
 
         LOG.debug(
             f"Found file handle for {key} in cache ({(time.perf_counter() - s) * 1000:.0f} ms)"
         )
 
-        return T.cast(str, file_handle)
+        return T.cast(str, cached_value)
 
-    def set(self, key: str, file_handle: str, expires_in: int = 3600 * 24 * 2) -> None:
+    @_retry_on_database_lock_error
+    def set(self, key: str, value: str, expires_in: int = 3600 * 24 * 2) -> None:
         s = time.perf_counter()
 
-        payload = {
+        data = {
             "expires_at": time.time() + expires_in,
-            "file_handle": file_handle,
+            "value": value,
         }
 
-        value: bytes = json.dumps(payload).encode("utf-8")
+        payload: bytes = json.dumps(data).encode("utf-8")
 
         with self._lock:
-            with dbm.open(self._file, flag="c") as db:
-                db[key] = value
+            with store.KeyValueStore(self._file, flag="c") as db:
+                db[key] = payload
 
         LOG.debug(
             f"Cached file handle for {key} ({(time.perf_counter() - s) * 1000:.0f} ms)"
         )
 
+    @_retry_on_database_lock_error
     def clear_expired(self) -> list[str]:
-        s = time.perf_counter()
-
         expired_keys: list[str] = []
 
-        with self._lock:
-            with dbm.open(self._file, flag="c") as db:
-                if hasattr(db, "items"):
-                    items: T.Iterable[tuple[str | bytes, bytes]] = db.items()
-                else:
-                    items = ((key, db[key]) for key in db.keys())
+        s = time.perf_counter()
 
-                for key, value in items:
-                    payload = self._decode(value)
-                    if self._is_expired(payload):
+        with self._lock:
+            with store.KeyValueStore(self._file, flag="c") as db:
+                for key, raw_payload in db.items():
+                    data = self._decode(raw_payload)
+                    if self._is_expired(data):
                         del db[key]
                         expired_keys.append(T.cast(str, key))
 
-        if expired_keys:
-            LOG.debug(
-                f"Cleared {len(expired_keys)} expired entries from the cache ({(time.perf_counter() - s) * 1000:.0f} ms)"
-            )
+        LOG.debug(
+            f"Cleared {len(expired_keys)} expired entries from the cache ({(time.perf_counter() - s) * 1000:.0f} ms)"
+        )
 
         return expired_keys
 
-    def keys(self):
-        with self._lock:
-            with dbm.open(self._file, flag="c") as db:
-                return db.keys()
+    def keys(self) -> list[str]:
+        if not self._db_existed():
+            return []
 
-    def _is_expired(self, payload: JSONDict) -> bool:
-        expires_at = payload.get("expires_at")
+        try:
+            with store.KeyValueStore(self._file, flag="r") as db:
+                return [key.decode("utf-8") for key in db.keys()]
+        except Exception as ex:
+            if self._table_not_found(ex):
+                return []
+            raise ex
+
+    def _is_expired(self, data: JSONDict) -> bool:
+        expires_at = data.get("expires_at")
         if isinstance(expires_at, (int, float)):
             return expires_at is None or expires_at <= time.time()
         return False
 
-    def _decode(self, value: bytes) -> JSONDict:
+    def _decode(self, raw_payload: bytes) -> JSONDict:
         try:
-            payload = json.loads(value.decode("utf-8"))
+            data = json.loads(raw_payload.decode("utf-8"))
         except json.JSONDecodeError as ex:
             LOG.warning(f"Failed to decode cache value: {ex}")
             return {}
 
-        if not isinstance(payload, dict):
-            LOG.warning(f"Invalid cache value format: {payload}")
+        if not isinstance(data, dict):
+            LOG.warning(f"Invalid cache value format: {raw_payload!r}")
             return {}
 
-        return payload
+        return data
+
+    def _db_existed(self) -> bool:
+        return os.path.exists(self._file)
+
+    def _table_not_found(self, ex: Exception) -> bool:
+        if isinstance(ex, sqlite3.OperationalError):
+            if "no such table" in str(ex):
+                return True
+        return False

mapillary_tools/store.py

@@ -0,0 +1,128 @@
+"""
+This module provides a persistent key-value store based on SQLite.
+
+This implementation is mostly copied from dbm.sqlite3 in the Python standard library,
+but works for Python >= 3.9, whereas dbm.sqlite3 is only available for Python 3.13.
+
+Source: https://github.com/python/cpython/blob/3.13/Lib/dbm/sqlite3.py
+"""
+
+import os
+import sqlite3
+import sys
+from collections.abc import MutableMapping
+from contextlib import closing, suppress
+from pathlib import Path
+
+BUILD_TABLE = """
+  CREATE TABLE IF NOT EXISTS Dict (
+    key BLOB UNIQUE NOT NULL,
+    value BLOB NOT NULL
+  )
+"""
+GET_SIZE = "SELECT COUNT (key) FROM Dict"
+LOOKUP_KEY = "SELECT value FROM Dict WHERE key = CAST(? AS BLOB)"
+STORE_KV = "REPLACE INTO Dict (key, value) VALUES (CAST(? AS BLOB), CAST(? AS BLOB))"
+DELETE_KEY = "DELETE FROM Dict WHERE key = CAST(? AS BLOB)"
+ITER_KEYS = "SELECT key FROM Dict"
+
+
+def _normalize_uri(path):
+    path = Path(path)
+    uri = path.absolute().as_uri()
+    while "//" in uri:
+        uri = uri.replace("//", "/")
+    return uri
+
+
+class KeyValueStore(MutableMapping):
+    def __init__(self, path, /, *, flag="r", mode=0o666):
+        """Open a key-value database and return the object.
+
+        The 'path' parameter is the name of the database file.
+
+        The optional 'flag' parameter can be one of ...:
+            'r' (default): open an existing database for read only access
+            'w': open an existing database for read/write access
+            'c': create a database if it does not exist; open for read/write access
+            'n': always create a new, empty database; open for read/write access
+
+        The optional 'mode' parameter is the Unix file access mode of the database;
+        only used when creating a new database. Default: 0o666.
+        """
+        path = os.fsdecode(path)
+        if flag == "r":
+            flag = "ro"
+        elif flag == "w":
+            flag = "rw"
+        elif flag == "c":
+            flag = "rwc"
+            Path(path).touch(mode=mode, exist_ok=True)
+        elif flag == "n":
+            flag = "rwc"
+            Path(path).unlink(missing_ok=True)
+            Path(path).touch(mode=mode)
+        else:
+            raise ValueError(f"Flag must be one of 'r', 'w', 'c', or 'n', not {flag!r}")
+
+        # We use the URI format when opening the database.
+        uri = _normalize_uri(path)
+        uri = f"{uri}?mode={flag}"
+
+        if sys.version_info >= (3, 12):
+            # This is the preferred way, but only available in Python 3.10 and newer.
+            self._cx = sqlite3.connect(uri, autocommit=True, uri=True)
+        else:
+            self._cx = sqlite3.connect(uri, uri=True)
+
+        # This is an optimization only; it's ok if it fails.
+        with suppress(sqlite3.OperationalError):
+            self._cx.execute("PRAGMA journal_mode = wal")
+
+        if flag == "rwc":
+            self._execute(BUILD_TABLE)
+
+    def _execute(self, *args, **kwargs):
+        if sys.version_info >= (3, 12):
+            return closing(self._cx.execute(*args, **kwargs))
+        else:
+            # Use a context manager to commit the changes
+            with self._cx:
+                return closing(self._cx.execute(*args, **kwargs))
+
+    def __len__(self):
+        with self._execute(GET_SIZE) as cu:
+            row = cu.fetchone()
+        return row[0]
+
+    def __getitem__(self, key):
+        with self._execute(LOOKUP_KEY, (key,)) as cu:
+            row = cu.fetchone()
+        if not row:
+            raise KeyError(key)
+        return row[0]
+
+    def __setitem__(self, key, value):
+        self._execute(STORE_KV, (key, value))
+
+    def __delitem__(self, key):
+        with self._execute(DELETE_KEY, (key,)) as cu:
+            if not cu.rowcount:
+                raise KeyError(key)
+
+    def __iter__(self):
+        with self._execute(ITER_KEYS) as cu:
+            for row in cu:
+                yield row[0]
+
+    def close(self):
+        self._cx.close()
+
+    def keys(self):
+        return list(super().keys())
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *args):
+        self.close()

mapillary_tools/uploader.py

@@ -1311,7 +1311,7 @@ def _is_uuid(key: str) -> bool:
 
 
 def _build_upload_cache_path(upload_options: UploadOptions) -> Path:
-    # Different python/CLI versions use different cache (dbm) formats.
+    # Different python/CLI versions use different cache formats.
     # Separate them to avoid conflicts
     py_version_parts = [str(part) for part in sys.version_info[:3]]
     version = f"py_{'_'.join(py_version_parts)}_{VERSION}"