""" SQL-related functions defined here are really pass the queries (or other SQL commands) to :class:`.threads.sqlThread` through `sqlSubmitQueue` queue and check or return the result got from `sqlReturnQueue`. This is done that way because :mod:`sqlite3` is so thread-unsafe that they won't even let you call it from different threads using your own locks. SQLite objects can only be used from one thread. .. note:: This actually only applies for certain deployments, and/or really old version of sqlite. I haven't actually seen it anywhere. Current versions do have support for threading and multiprocessing. I don't see an urgent reason to refactor this, but it should be noted in the comment that the problem is mostly not valid. Sadly, last time I checked, there is no reliable way to check whether the library is or isn't thread-safe. """ import Queue import threading sqlSubmitQueue = Queue.Queue() """the queue for SQL""" sqlReturnQueue = Queue.Queue() """the queue for results""" sql_lock = threading.Lock() """ lock to prevent queueing a new request until the previous response is available """ sql_available = False """set to True by `.threads.sqlThread` immediately upon start""" sql_ready = threading.Event() """set by `.threads.sqlThread` when ready for processing (after initialization is done)""" def sqlQuery(sql_statement, *args): """ Query sqlite and return results :param str sql_statement: SQL statement string :param list args: SQL query parameters :rtype: list """ assert sql_available sql_lock.acquire() sqlSubmitQueue.put(sql_statement) if args == (): sqlSubmitQueue.put('') elif isinstance(args[0], (list, tuple)): sqlSubmitQueue.put(args[0]) else: sqlSubmitQueue.put(args) queryreturn, _ = sqlReturnQueue.get() sql_lock.release() return queryreturn def sqlExecuteChunked(sql_statement, idCount, *args): """Execute chunked SQL statement to avoid argument limit""" # SQLITE_MAX_VARIABLE_NUMBER, # unfortunately getting/setting isn't exposed to python assert sql_available sqlExecuteChunked.chunkSize = 999 if idCount == 0 or idCount > len(args): return 0 total_row_count = 0 with sql_lock: for i in range( len(args) - idCount, len(args), sqlExecuteChunked.chunkSize - (len(args) - idCount) ): chunk_slice = args[ i:i + sqlExecuteChunked.chunkSize - (len(args) - idCount) ] sqlSubmitQueue.put( sql_statement.format(','.join('?' * len(chunk_slice))) ) # first static args, and then iterative chunk sqlSubmitQueue.put( args[0:len(args) - idCount] + chunk_slice ) ret_val = sqlReturnQueue.get() total_row_count += ret_val[1] sqlSubmitQueue.put('commit') return total_row_count def sqlExecute(sql_statement, *args): """Execute SQL statement (optionally with arguments)""" assert sql_available sql_lock.acquire() sqlSubmitQueue.put(sql_statement) if args == (): sqlSubmitQueue.put('') else: sqlSubmitQueue.put(args) _, rowcount = sqlReturnQueue.get() sqlSubmitQueue.put('commit') sql_lock.release() return rowcount def sqlStoredProcedure(procName): """Schedule procName to be run""" assert sql_available sql_lock.acquire() sqlSubmitQueue.put(procName) sql_lock.release() class SqlBulkExecute(object): """This is used when you have to execute the same statement in a cycle.""" def __enter__(self): sql_lock.acquire() return self def __exit__(self, exc_type, value, traceback): sqlSubmitQueue.put('commit') sql_lock.release() @staticmethod def execute(sql_statement, *args): """Used for statements that do not return results.""" assert sql_available sqlSubmitQueue.put(sql_statement) if args == (): sqlSubmitQueue.put('') else: sqlSubmitQueue.put(args) sqlReturnQueue.get()