optimize channel_db:

- use python objects mirrored by sql database
 - write sql to file asynchronously
 - the sql decorator is awaited in sweepstore, not in channel_db

electrum/sql_db.py

@@ -2,6 +2,7 @@ import os
 import concurrent
 import queue
 import threading
+import asyncio
 
 from sqlalchemy import create_engine
 from sqlalchemy.pool import StaticPool
@@ -18,28 +19,32 @@ def sql(func):
     """wrapper for sql methods"""
     def wrapper(self, *args, **kwargs):
         assert threading.currentThread() != self.sql_thread
-        f = concurrent.futures.Future()
+        f = asyncio.Future()
         self.db_requests.put((f, func, args, kwargs))
-        return f.result(timeout=10)
+        return f
     return wrapper
 
 class SqlDB(Logger):
     
-    def __init__(self, network, path, base):
+    def __init__(self, network, path, base, commit_interval=None):
         Logger.__init__(self)
         self.base = base
         self.network = network
         self.path = path
+        self.commit_interval = commit_interval
         self.db_requests = queue.Queue()
         self.sql_thread = threading.Thread(target=self.run_sql)
         self.sql_thread.start()
 
     def run_sql(self):
+        #return
+        self.logger.info("SQL thread started")
         engine = create_engine('sqlite:///' + self.path, pool_reset_on_return=None, poolclass=StaticPool)#, echo=True)
         DBSession = sessionmaker(bind=engine, autoflush=False)
-        self.DBSession = DBSession()
         if not os.path.exists(self.path):
             self.base.metadata.create_all(engine)
+        self.DBSession = DBSession()
+        i = 0
         while self.network.asyncio_loop.is_running():
             try:
                 future, func, args, kwargs = self.db_requests.get(timeout=0.1)
@@ -50,7 +55,14 @@ class SqlDB(Logger):
             except BaseException as e:
                 future.set_exception(e)
                 continue
-            future.set_result(result)
+            if not future.cancelled():
+                future.set_result(result)
+            # note: in sweepstore session.commit() is called inside
+            # the sql-decorated methods, so commiting to disk is awaited
+            if self.commit_interval:
+                i = (i + 1) % self.commit_interval
+                if i == 0:
+                    self.DBSession.commit()
         # write
         self.DBSession.commit()
         self.logger.info("SQL thread terminated")