On Wednesday, 25 May 2016 19:15:38 UTC+1, Ian wrote:
> On Wed, May 25, 2016 at 10:52 AM, wrote:
> > Link: https://github.com/JimChengLin/AsyncDB
> >
> > I always wonder why people do not make an async DB, when they are crazy
> > with async web framework. Hard disks are faster than the Internet, but
> > still pretty slow compared to CPU/RAM.
>
> In-process dbm-style databases aren't really all that popular though,
> are they? At least, I never hear much about them. It seems to me that
> most people use out-of-process SQL or NoSQL databases. For those, the
> details of the database implementation aren't very important to the
> user, and all that is really needed is an async client library, such
> as aiopg.
>
> I wonder if there's a need for a version 3 of the Python DBAPI spec
> including async operations?
>
> > My implementation is not pure async though. The insert and del actions are
> > half async due to there is not async __setitem__. I prefer a neat API over
> > performance.
>
> Would it be going too far if we had async versions of all the special
> methods: __ainit__, __asetitem__, __aiadd__, etc.?
We do not have to have all special methods. Indeed, we just need an async
__setitem__. At least, it is my proposal. Would the core dev team consider it?
I think in-process DB is quite popular in less serious development, e.g. SQLite.
A sync DB with async connections can be treated as "async DB". But still, it is
"evil" :). You know what I mean? It is about purity. Haha
--
https://mail.python.org/mailman/listinfo/python-list
