trio
trio copied to clipboard
python-trio
Could this lib I made be added to the Awesome Libraries page?
https://pypi.org/project/trio-mongodb/ github.com/obnoxiousish/trio-mongodb
It uses multiprocess+trio.to_sync to process pymongo queries Or just trio.to_sync in the main process alternatively
Couldn't find any other nosql dbms option using trio, there is only *SQL or redis options, would be better if it was true async but would take 5ever to implement mongodb from scratch
By to_sync I guess you mean to_thread.run_sync()
Using threads only can potentially result in a massive slow down for your main process
Above is from your readme. Have you tested this? It could well be true but, then again, it might not be: Python releases the GIL during I/O so I would normally expect something I/O heavy like database client to scale very well across threads. Then again, if pymongo does a lot of processing to the data in pure Python (i.e., if it is poorly written) then it could be true. I'm just saying I wouldn't make a statement like that without some testing first.
A couple of other suggestions:
- Make the client an async context manager so that client will definitely be closed.
- For the threaded client, allow passing in a trio capacitylimiter, so that users can choose how many threads to use and also when some client connections share threads (and could block each other) and when not (by passing different capacity limiter objects).
I had similar thoughts with adding support for async context managers when I took a look at the project the other day, and I was also questioning if just moving everything to another thread would be helpful. Having some sort of comparison data would be very helpful. If there were a small script for generating comparison data, I would be happy to assist with your data collection!
Project page remains a 404, closing this for now. Please don't hesitate to reopen this once it's published.