At the same time GINO is kinda different in e.g. different layers of SQLAlchemy integration (vanilla SQLAlchemy + async execution, using GINO Model to build queries only, or full ORM) and the GINO Model loader system

Oh thanks! Still work to do there 😛

@fantixking do you have any benchmarks for GinoORM with comparing to over ORMs? for example, tortoise?

I tried to google it, but did not find anything

@fantixking Question from Dmitry: why do you use asyncio.sleep(0.1) in this example? According to Dmitry, it will work just fine and nothing should block

Unfortunately not yet - would be fun to do some benchmarking. I’m not expecting GINO to be the fastest as in execution, but asyncpg does bring some boost to cover the ORM overhead - this applies to Tortoise and Encode ORM too

yeah, if compare to aiopg - asyncpg is really fast)

Right, without the sleep it won’t block. This code here is to simulate doing async work in real life, and demonstrating how blocking ORM database access could ruin an asynchronous application if used inappropriately.

thank you!

@fantixking, probably some misunderstanding occurs, could you please explain what is the difference between 10th and 11th call in that code?

Вопросы Фантиксу

@fantixking can I ask off-top question? it is relative to support project on local languages

a lot of china open-sources packages has docs only in mandarian (sorry if I'm wrong in it, mean main China's language)

@fantixking thank you for the talk 👏🏻

Thank you all! Its’ been a great pleasure to be here. I’ll keep answering to the questions, please forgive my slow typing

What is the general attitude to English in IT in China? is this something optional? why am I asking - in Russia there are always disputes about whether it is necessary to localize documentation or projects? it is believed that this stimulates the isolation of development within the country, while English, on the contrary, allows you to expand the scope

I very hard formulate my question)) sorry

Сейчас в эфире @k0tb9g9m0t
С темой: Внедрение С++ расширений в Python

Это живой доклад, пишите в чате вопросы, мы их зададим спикеру после выступления :)

Oh sure, sorry for the confusion! So the 10th request could successfully grab a DB connection from the pool, it’s actually the last connection in the pool. Then when the 11th request came in - assuming all first 10 requests are awaiting on some asynchronous tasks (keeping 1 DB connectionn each!) - it will firstly try to acquire a new connection, but there isn’t any more to spare, so it blocks, and it’s blocking the thread. It’s like an infinite loop that keeps acquiring connections from the pool but never returns, it will block forever and won’t get unblocked just by itself.