Threads and async

Threads

Every function in jsonrpcclient can be called from several threads at once. That includes free-threaded Python (3.14t), where there is no GIL. The test suite runs on 3.14t. It calls the request functions from many threads at once and checks that no call fails and no id is handed out twice.

The request functions share one id sequence per style across the whole process (see Ids). Two threads calling request at the same time always get different ids. The iterators in jsonrpcclient.id_generators are thread-safe too, so you can share one between threads.

parse and parse_json keep no state. The only thing to watch is the lazy iterator that parse returns for a batch: don't consume one iterator from two threads. Call list() on it first if you need to share the responses.

New in 4.0.4

Before 4.0.4, request_hex, request_random and request_uuid could raise ValueError: generator already executing when called from several threads at once, and free-threaded Python could crash.

asyncio

jsonrpcclient does no I/O and never blocks, so you can call it directly in async code. Use an async transport for the sending part: httpx's AsyncClient, aiohttp or websockets. See Transports.

Requests made concurrently with asyncio.gather still get distinct ids. Their responses can arrive in any order, so match them by id as you would for a batch.