fix: Improve query performance by using a single connection

Use one connection for the lifetime of the process rather
than creating a new connection for each query.

This cuts execution time roughtly in half for queries
across large Things databases.

Author: mbhutton

things/database.py

@@ -168,6 +168,7 @@ class Database:
     """
 
     debug = False
+    connection: sqlite3.Connection
 
     # pylint: disable=R0913
     def __init__(self, filepath=None, print_sql=False):
@@ -181,6 +182,11 @@ def __init__(self, filepath=None, print_sql=False):
         if self.print_sql:
             self.execute_query_count = 0
 
+        # "ro" means read-only
+        # See: https://sqlite.org/uri.html#recognized_query_parameters
+        uri = f"file:{self.filepath}?mode=ro"
+        self.connection = sqlite3.connect(uri, uri=True)  # pylint: disable=E1101
+
         # Test for migrated database in Things 3.15.16+
         # --------------------------------
         assert self.get_version() > 21, (
@@ -495,16 +501,12 @@ def execute_query(self, sql_query, parameters=(), row_factory=None):
             print(prettify_sql(sql_query))
             print()
 
-        # "ro" means read-only
-        # See: https://sqlite.org/uri.html#recognized_query_parameters
-        uri = f"file:{self.filepath}?mode=ro"
-        connection = sqlite3.connect(uri, uri=True)  # pylint: disable=E1101
-        connection.row_factory = row_factory or dict_factory
-        cursor = connection.cursor()
+        self.connection.row_factory = row_factory or dict_factory
+        cursor = self.connection.cursor()
+
         cursor.execute(sql_query, parameters)
 
         result = cursor.fetchall()
-        connection.close()
         return result